A method, device and medium for rapid estimation of LED chip lifetime

Abstract

The application discloses a kind of LED chip life quick estimation method, equipment and medium, it is related to electric digital data processing technical field, including, based on photoelectric thermal characteristic data, for calibration sample design current pulse stress sequence, photoelectric thermal response data is synchronously collected before and after pulse, and is checked and is completed with mapping storage, generates photoelectric thermal pulse response dataset;According to photoelectric thermal pulse response dataset, extract the photoelectric thermal change characteristic sensitive to the life of LED chip, and construct the photoelectric thermal pulse life fingerprint of LED chip;Build life quick estimation model, and according to photoelectric thermal pulse life fingerprint and working condition information, the life of LED chip is predicted, generates the predicted life of LED chip;The application is compressed into unified life representation carrier by constructing the photoelectric thermal pulse life fingerprint of LED chip, photoelectric thermal evolution process under pulse stress is as life modeling provides directly callable degradation feature basis.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of digital data processing, and particularly relates to an LED chip life rapid estimation method, device and medium. BACKGROUND

[0002] As a key device in solid-state lighting and display devices, the life and reliability of an LED chip directly affect the light flux maintenance and maintenance cost of a whole lamp. Existing life evaluation usually adopts an accelerated aging test under constant current and constant temperature or temperature and current, continuously monitors light flux, color coordinates, electrical parameters and junction temperature and other photoelectric and thermal characteristics of samples under rated or more severe working conditions, and combines pre-set failure criteria and a statistical life model to infer the life distribution and reliability level of the LED chip, which is used for device screening and product design, and is a common reliability evaluation idea in the industry.

[0003] However, from the perspective of conventional accelerated aging and life extrapolation methods, on the one hand, the light flux decay curve under constant stress is mainly used as the main basis, and it is difficult to fully utilize the photoelectric and thermal characteristics in short-time stress response to rapidly estimate the early degradation behavior of the LED chip; on the other hand, the joint modeling of historical aging data and in-service failure records is limited, and there is a lack of systematic calibration and closed-loop correction between the model and the actual application working condition. SUMMARY

[0004] In view of the above existing problems, the present application is proposed.

[0005] Therefore, the present application provides an LED chip life rapid estimation method to solve the problems that the prior art cannot rapidly estimate the life of the LED chip by utilizing the photoelectric and thermal response characteristics under short-time stress, and the joint modeling of historical aging data and in-service failure records is limited.

[0006] To solve the above technical problems, the present application provides the following technical solutions:

[0007] In a first aspect, the present application provides an LED chip life rapid estimation method, which comprises collecting photoelectric and thermal characteristic data of the LED chip under rated conditions, and verifying and screening the data in combination with failure criteria to generate a calibration sample set of the LED chip;

[0008] Based on the photoelectric and thermal characteristic data, a current pulse stress sequence is designed for the calibration sample, photoelectric and thermal response data are synchronously collected before and after the pulse, and verification, completion and mapping storage are performed to generate a photoelectric and thermal pulse response data set;

[0009] According to the photoelectric and thermal pulse response data set, photoelectric and thermal change characteristics sensitive to the life of the LED chip are extracted, and a photoelectric and thermal pulse life fingerprint of the LED chip is constructed;

[0010] A life quick estimation model is constructed, and life of the LED chip is predicted according to the photoelectric thermal pulse life fingerprint and working condition information, so as to generate the predicted life of the LED chip;

[0011] According to the photoelectric thermal pulse life fingerprint, clustering analysis is carried out on the LED chip, and representative samples of each type are extracted for accelerated aging verification, the deviation between the predicted life and the measured life is obtained, and the life quick estimation model is optimized.

[0012] As a preferred scheme of the LED chip life quick estimation method, the combination of the failure criterion for verification and screening generates a calibrated sample set of the LED chip, and the steps are as follows,

[0013] The photoelectric thermal characteristics data of the LED chip are collected and standardized under the rated condition, and the LED chip test data set is generated by summarizing;

[0014] Based on the historical aging data and the in-service failure record, the relationship between different light maintenance rates, color coordinate offset, upper limit of electrical parameters and actual failure probability is statistically fitted by using the multivariate Logistic regression analysis method, and the failure criterion of the LED chip is generated;

[0015] The photoelectric thermal characteristics data of each LED chip are verified and screened in combination with the LED chip test data set and the failure criterion, and the calibrated sample set of the LED chip is generated.

[0016] As a preferred scheme of the LED chip life quick estimation method, the current pulse stress sequence is designed for the calibrated sample, and the steps are as follows,

[0017] According to the difference between the photoelectric thermal characteristics data of each LED chip in the calibrated sample set of the LED chip and the failure criterion, the allowable space of each LED chip from the failure boundary is estimated;

[0018] The allowable space is counted, the low quantile value is selected as the safety margin of the stress level that the calibrated sample can withstand, and the safety margin is converted into the target degradation excitation strength according to the degradation excitation conversion coefficient;

[0019] Under the constraints of the safety margin and the target degradation excitation strength, the current amplitude pulse width and pulse interval are selected to generate the current pulse stress sequence.

[0020] As a preferred scheme of the LED chip life quick estimation method, the verification and mapping storage are carried out, and the photoelectric thermal pulse response data set is generated, and the steps are as follows,

[0021] Under the controlled ambient temperature, the current pulse stress sequence is applied to each LED chip in turn, the photoelectric thermal response data is synchronously collected before and after each current pulse, and is collected according to the LED chip number and the current pulse sequence, so as to form a photoelectric thermal response original data set;

[0022] The photoelectric thermal response original data set is subjected to missing value completion, format unification and standardization processing, so as to generate a photoelectric thermal pulse response data set.

[0023] As a preferred scheme of the LED chip life rapid estimation method, the LED chip life rapid estimation method comprises the following steps:

[0024] According to the photoelectric thermal pulse response data set, the photoelectric thermal feature change amount and the change rate per unit time of each LED chip before and after each current pulse relative to the last time are calculated, so as to obtain a photoelectric thermal change trajectory set.

[0025] According to the photoelectric thermal change trajectory set and the corresponding relationship between the photoelectric thermal change and the life performance in the historical aging data, the distinguishing ability of different photoelectric thermal feature change amounts and change rates to the degradation degree is analyzed, the photoelectric thermal change amount and the change rate capable of distinguishing different life levels are screened, and a photoelectric thermal change feature set is formed.

[0026] According to the photoelectric thermal change feature set, the photoelectric thermal change amount and the change rate of each LED chip under each current pulse are spliced according to a preset feature order and a fixed length, so as to construct a photoelectric thermal pulse life fingerprint of the LED chip.

[0027] As a preferred scheme of the LED chip life rapid estimation method, the LED chip life rapid estimation method comprises the following steps:

[0028] The life rapid estimation model is constructed based on an input fusion layer, a physical index extraction layer, a feature representation layer and a life regression layer.

[0029] The input fusion layer splices the photoelectric thermal pulse life fingerprint of the LED chip and the working condition information in a fixed order, and performs standardization processing, so as to generate a unified input feature vector.

[0030] The physical index extraction layer extracts the photoelectric thermal response rate information (the luminous flux change rate per unit time, the reverse leakage current increment and the junction temperature rise rate) according to the degradation rate rule, and converts the photoelectric thermal response rate information into a degradation rate index and a thermal stress index through a principal component analysis method.

[0031] The feature characterization layer fuses the unified input feature vector with the degradation rate index and the thermal stress index, and obtains a life characterization vector through nonlinear transformation compression and reorganization;

[0032] The life regression layer estimates the predicted life of the LED chip based on the life characterization vector through gradient boosting regression tree, and reversely optimizes the feature characterization layer and the input fusion layer in the training stage between the predicted life and the measured life, to form a complete life rapid estimation model.

[0033] The photoelectric thermal pulse life fingerprint and the working condition information are input into the life rapid estimation model to obtain the predicted life of the LED chip.

[0034] As a preferred scheme of the LED chip life rapid estimation method, when the photoelectric thermal pulse life fingerprint is used for cluster analysis on the LED chip, the photoelectric thermal pulse life fingerprint is divided into several fingerprint cluster categories according to the similarity degree, and a life fingerprint clustering result is generated.

[0035] According to the photoelectric thermal pulse life fingerprint of the LED chip, the photoelectric thermal pulse life fingerprint is divided into several fingerprint cluster categories according to the similarity degree, and a life fingerprint clustering result is generated.

[0036] According to the photoelectric thermal pulse life fingerprint of the LED chip, the photoelectric thermal pulse life fingerprint is divided into several fingerprint cluster categories according to the similarity degree, and a life fingerprint clustering result is generated.

[0037] The LED chips in the accelerated aging verification sample set are continuously run under the accelerated aging working condition until the failure criterion is met, the measured life of each LED chip is recorded, and the measured life is compared with the predicted life of the corresponding LED chip to generate a life deviation data set.

[0038] According to the life deviation data set, the parameters of the life regression layer in the life rapid estimation model and the photoelectric thermal change features in the feature characterization layer are optimized and updated.

[0039] As a preferred scheme of the LED chip life rapid estimation method, the photoelectric thermal feature data includes luminous flux, color coordinates, peak wavelength, forward voltage, reverse leakage current, dynamic on-resistance, junction temperature estimation value, junction shell thermal resistance and temperature rise amplitude under pulse current.

[0040] In a second aspect, the present application provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and wherein the computer program is executed by the processor to implement any step of the LED chip life rapid estimation method according to the first aspect of the present application.

[0041] In a third aspect, the present application provides a computer readable storage medium having stored thereon a computer program, wherein the computer program, when executed by a processor, implements any step of the LED chip lifetime fast estimation method according to the first aspect of the present application.

[0042] The present application has the following beneficial effects: by constructing the photo-electric-thermal pulse lifetime fingerprint of the LED chip, the photo-electric-thermal evolution process under pulse stress is compressed into a unified lifetime characterization carrier, providing a directly callable degradation feature basis for lifetime modeling; by constructing the lifetime fast estimation model and generating the predicted lifetime of the LED chip according to the photo-electric-thermal pulse lifetime fingerprint and the working condition information, the processing link of directly outputting the lifetime characterization vector and the lifetime result from the short-time photo-electric-thermal test result under the given use condition is realized, which facilitates the rapid acquisition of the lifetime information of the LED chip in production screening and reliability evaluation. BRIEF DESCRIPTION OF DRAWINGS

[0043] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiment description. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without creative labor.

[0044] Fig. 1 The flowchart of the LED chip lifetime fast estimation method.

[0045] Fig. 2 The flowchart of acquiring the lifetime deviation data set.

[0046] Fig. 3 The flowchart of generating the calibration sample set.

[0047] Fig. 4 The flowchart of constructing the photo-electric-thermal pulse lifetime fingerprint. DETAILED DESCRIPTION

[0048] In order to make the above-mentioned purposes, features and advantages of the present application more apparent and easy to understand, the specific embodiments of the present application will be described in detail below with reference to the drawings of the specification.

[0049] In the following description, many specific details are set forth in order to provide a thorough understanding of the present application, but the present application can also be implemented in other ways different from those described herein, and those skilled in the art can make similar generalizations without departing from the connotation of the present application, therefore the present application is not limited to the specific embodiments disclosed below.

[0050] Second, the "one embodiment" or "an embodiment" referred to herein can include a particular feature, structure, or characteristic. The various embodiments are not mutually exclusive, but a single embodiment can be selected from a plurality of mutually exclusive or other alternative embodiments.

[0051] With reference to Figs. 1-4 For one embodiment of the present application, the embodiment provides a method for rapid estimation of LED chip life, comprising the following steps:

[0052] S1, collect the photoelectric thermal characteristic data of the LED chip under the rated condition, and verify and screen in combination with the failure criterion to generate a calibration sample set of the LED chip.

[0053] Perform photoelectric thermal joint test on each LED chip in the batch under the rated condition, collect the photoelectric thermal characteristic data of the LED chip and perform standardization processing, and aggregate to generate an LED chip test data set.

[0054] The photoelectric thermal characteristic data includes luminous flux, color coordinates, peak wavelength, forward voltage, reverse leakage current, dynamic on-resistance, junction temperature estimate, junction shell thermal resistance, and temperature rise amplitude under pulse current.

[0055] Based on historical aging data and in-service failure records, the relationship between different luminous flux maintenance rates, color coordinate offset amounts, upper limits of electrical parameters, and actual failure probabilities is statistically fitted using a multivariate logistic regression analysis method to generate a failure criterion for the LED chip.

[0056] Further, based on historical aging data and in-service failure records, the historical aging data and in-service failure records are sorted by LED chip number, use condition, and time axis, the luminous flux maintenance rate, color coordinate offset amount, and upper limit of electrical parameters are calculated for each LED chip life record, and the actual failure mark of whether failure occurs is given to form a sample set containing luminous flux maintenance rate, color coordinate offset amount, upper limit of electrical parameters, and actual failure mark; based on the sample set, taking luminous flux maintenance rate, color coordinate offset amount, and upper limit of electrical parameters as independent variables, and taking actual failure mark as dependent variable, the actual failure probability under different combinations of luminous flux maintenance rate, color coordinate offset amount, and upper limit of electrical parameters is calculated using a multivariate logistic regression analysis method, the performance boundary parameters corresponding to the luminous flux maintenance rate, color coordinate offset amount, and upper limit of electrical parameters are back calculated according to a preset failure probability level (an actual failure probability value directly selected according to the maximum failure rate allowed in the target application, combined with relevant reliability standards or design specifications), and the performance boundary parameters are aggregated as a set of performance boundary parameters for judging whether the LED chip enters a failure state to generate a failure criterion for the LED chip.

[0057] It should be noted that the historical aging data refers to the complete test process data of the luminous flux change, electrical parameter change, junction temperature change and time to failure criterion of the LED chip of the previous production batch under controlled accelerated stress or long-term rated operating conditions. The in-service failure record refers to the field statistical information of the LED chip in the actual application scene, including the use of working conditions, the use of time, the failure phenomenon description and the confirmation conclusion of the returned sample data.

[0058] The failure criterion refers to a set of performance boundary parameters for judging whether the LED chip has entered a failure state, i.e., the allowable upper limit of each data in the optoelectronic thermal characteristic data.

[0059] The rated condition refers to the normal working state formed by the combination of the rated working current, the rated environment or the junction temperature and the corresponding driving voltage according to the LED chip product standard. The LED chip test data set includes the optoelectronic thermal characteristic data of each LED chip in the batch to be tested under the rated condition, and additional information such as the chip number and test time test working condition corresponding to the optoelectronic thermal characteristic data.

[0060] In combination with the LED chip test data set and the failure criterion, the optoelectronic thermal characteristic data of each LED chip is checked and screened, and the LED chip that does not meet the failure criterion is removed, and the LED chip that meets the failure criterion is screened, and a calibration sample set of the LED chip is generated.

[0061] Further, in combination with the LED chip test data set and the failure criterion, the corresponding optoelectronic thermal characteristic data of each LED chip in the LED chip test data set is read one by one, the optoelectronic thermal characteristic data is compared with each performance boundary parameter in the failure criterion, and a determination mark is given according to the comparison result whether each LED chip meets the failure criterion. After obtaining the determination mark, the LED chip with the determination mark not meeting the failure criterion is removed from the LED chip test data set, only the LED chip with the determination mark meeting the failure criterion is retained, and the LED chip meeting the failure criterion and the corresponding optoelectronic thermal characteristic data are re- summarized to form a calibration sample set of the LED chip for subsequent current pulse stress loading and life analysis.

[0062] S2, based on the optoelectronic thermal characteristic data, designing a current pulse stress sequence for the calibration sample, synchronously collecting optoelectronic thermal response data before and after the pulse, and performing verification, completion and mapping storage to generate an optoelectronic thermal pulse response data set.

[0063] According to the difference between the optoelectronic thermal characteristic data of each LED chip in the calibration sample set of the LED chip and the failure criterion, the allowable space of each LED chip from the failure boundary is estimated.

[0064] Further, according to the calibration sample set of the LED chip, the photoelectric thermal characteristic data of each LED chip in the calibration sample set of the LED chip is read item by item, and difference calculation is performed according to the photoelectric thermal characteristic data and the corresponding performance boundary parameter in the failure criterion, to obtain the remaining margin of each item of photoelectric thermal characteristic data from the failure criterion boundary; after obtaining the remaining margin of each item of photoelectric thermal characteristic data, the photoelectric thermal characteristic data of the same LED chip is comprehensively evaluated with the minimum remaining margin as the limiting condition, the corresponding minimum remaining margin is recorded as the allowable space of the LED chip from the failure boundary, and all the allowable spaces of the LED chips from the failure boundary are sorted into an allowable space set for subsequent determination of the safety margin.

[0065] The allowable spaces are counted, the low quantile value is selected as the safety margin of the stress level that the calibration sample can withstand, and the safety margin is converted into the target degradation excitation strength according to the degradation excitation conversion coefficient.

[0066] Further, the allowable space set is sorted according to the numerical value, and the allowable space value at the selected percentile position in the sorted allowable space set is calculated. The allowable space value at the selected percentile position in the sorted allowable space set is determined as the low quantile value of the allowable space; after obtaining the low quantile value of the allowable space, the low quantile value of the allowable space is taken as the safety margin of the stress level that the calibration sample can withstand, and the safety margin of the stress level that the calibration sample can withstand is converted into the target degradation excitation strength according to the corresponding relationship between the degradation excitation conversion coefficient and the safety margin of the stress level that the calibration sample can withstand, to provide a basis for subsequent selection of current amplitude pulse width and pulse interval under the constraints of safety margin and target degradation excitation strength.

[0067] It should be noted that the low quantile value refers to the percentile position of the allowable space of each LED chip from the failure boundary sorted from small to large, for example, the value at the 10% position of the smallest in all allowable spaces. The degradation excitation conversion coefficient is determined according to the fact that different conversion ratios of current pulses are sequentially applied to a small number of LED chip calibration samples, and it is compared whether the light flux drop amplitude under each conversion ratio falls into the degradation interval (such as 5% to 15% relative to the initial value) and does not trigger the failure criterion, and the conversion ratio that meets both the degradation interval requirement and the safety requirement is selected.

[0068] Under the constraints of safety margin and target degradation excitation strength, the current amplitude pulse width and pulse interval are selected so that the current pulse neither exceeds the safety margin of the LED chip nor causes measurable photoelectric thermal changes, and a current pulse stress sequence is generated.

[0069] Further, after the safety margin and the target degradation excitation intensity are determined, the target degradation excitation intensity is taken as the upper limit of the acceptable energy input of the LED chip under a single current pulse, and the candidate combination of the current amplitude, pulse width and pulse interval is converted according to the safety margin to obtain a candidate combination of the current amplitude, pulse width and pulse interval, the corresponding current pulse energy is calculated for each combination of the current amplitude, pulse width and pulse interval, and it is checked that the current pulse energy does not exceed the safety margin and can cause measurable changes in the luminous flux, reverse leakage current and junction temperature changes and other optoelectronic thermal characteristics according to the empirical sensitivity threshold (for example, the relative decrease in luminous flux after a single current pulse exceeds the measurement noise level, the reverse leakage current shows a sustained upward trend and is higher than the instrument noise band, and the junction temperature change exceeds the minimum resolvable temperature difference defined by the temperature measurement resolution), the combinations of the current amplitude, pulse width and pulse interval that meet the safety margin constraint and the measurable change constraint are arranged in a preset order, and the continuous arrangement obtains a current pulse stress sequence for subsequent application of current pulses and collection of optoelectronic thermal response data.

[0070] It should be noted that the empirical sensitivity threshold refers to the minimum relative change that can be stably detected under the conditions of optoelectronic thermal characteristic measurement noise and instrument resolution. Generally, the range of about 1% to 5% change in luminous flux relative to the initial value and a few degrees Celsius change in junction temperature can be determined as the value range of the empirical sensitivity threshold through pre-experiment. The preset order is arranged according to the principle that the current amplitude is from low to high, the pulse width is from short to long, the pulse interval meets the heat dissipation recovery requirement, and the stress is gradually increased to facilitate tracking of the degradation behavior.

[0071] Under the controlled ambient temperature, the current pulse stress sequence is applied to each LED chip in turn, and the optoelectronic thermal response data is collected synchronously before and after each current pulse, and is collected according to the LED chip number and the current pulse order to form an optoelectronic thermal response original data set.

[0072] Further, under the controlled ambient temperature, the calibration sample set of the LED chip is fixed on the test tool one by one, and the current pulse stress sequence is applied to each LED chip in turn according to the pre-determined current amplitude, pulse width and pulse interval, the optoelectronic thermal response data of the LED chip is collected before the start of each current pulse, and the same type of optoelectronic thermal response data as before the start of the current pulse is repeatedly collected immediately after the end of the corresponding current pulse, the optoelectronic thermal response data before and after each current pulse is used to generate a single-pulse optoelectronic thermal response record with LED chip number, current pulse order and time marker, and all single-pulse optoelectronic thermal response records are collected and stored according to the LED chip number and the current pulse order to form an optoelectronic thermal response original data set for subsequent missing value completion, format unification and standardization processing.

[0073] The photoelectric thermal response original data set is subjected to missing value completion, format unification and standardization processing to generate a photoelectric thermal pulse response data set for describing the degradation behavior of the calibration sample set of the LED chip under pulse conditions.

[0074] Further, the photoelectric thermal response original data set is subjected to integrity check according to the LED chip number, current pulse sequence and time mark, the missing photoelectric thermal response data in the photoelectric thermal response original data set is interpolated and completed using the adjacent current pulse under the same type record, the field name and unit of measurement of the photoelectric thermal response data are unified on the basis of completion, and the standardized photoelectric thermal response data are collected according to the LED chip number and the current pulse sequence to generate a photoelectric thermal pulse response data set for describing the degradation behavior of the calibration sample set of the LED chip under pulse conditions.

[0075] S3, according to the photoelectric thermal pulse response data set, extracting the photoelectric thermal change characteristics sensitive to the lifetime of the LED chip, and constructing the photoelectric thermal pulse lifetime fingerprint of the LED chip.

[0076] According to the photoelectric thermal pulse response data set, the photoelectric thermal feature change amount and the change rate per unit time of each LED chip relative to the previous time before and after each current pulse are calculated to obtain a photoelectric thermal change trajectory set organized according to the LED chip number and the current pulse sequence.

[0077] Further, according to the photoelectric thermal pulse response data set, the photoelectric thermal feature data of each LED chip in the calibration sample set of the LED chip at the start and end time of each current pulse are read in pairs, the photoelectric thermal feature change amount at the end time of each current pulse relative to the previous time is calculated, and the change rate per unit time is obtained by dividing the photoelectric thermal feature change amount by the corresponding time interval, the photoelectric thermal feature change amount and the change rate per unit time of each LED chip under all current pulses are arranged in order according to the LED chip number and the current pulse sequence, and a photoelectric thermal change trajectory set for continuously recording the photoelectric thermal evolution process under each current pulse is generated, which provides input for subsequent analysis of the distinguishing ability of different photoelectric thermal feature change amounts and change rates to the degradation degree according to the photoelectric thermal change trajectory set.

[0078] According to the photoelectric thermal change trajectory set and the corresponding relationship between the photoelectric thermal change and the lifetime performance in the historical aging data, the distinguishing ability of different photoelectric thermal feature change amounts and change rates to the degradation degree is analyzed, the photoelectric thermal change amounts and change rates capable of distinguishing different lifetime levels are screened, and a photoelectric thermal change characteristic set is formed.

[0079] Further, according to the photoelectric-thermal change trajectory set and the corresponding relationship between the photoelectric-thermal change and the life performance in the historical aging data, the correlation index and the grouping statistical quantity of each type of photoelectric-thermal characteristic change amount and change rate in the photoelectric-thermal change trajectory set and the life performance are calculated respectively. By comparing the numerical distribution difference of the photoelectric-thermal characteristic change amount and change rate of the LED chips at different life levels under the same current pulse number, the stability and sensitivity of the photoelectric-thermal characteristic change amount and change rate in distinguishing the degradation degree are evaluated. The photoelectric-thermal characteristic change amount and change rate with stronger distinguishing ability are sorted and merged into the photoelectric-thermal change characteristic set sensitive to the life of the LED chip, and input is provided for subsequent construction of the photoelectric-thermal pulse life fingerprint of the LED chip.

[0080] The photoelectric-thermal characteristic change amount and change rate with stronger distinguishing ability refer to that when the LED chips are grouped according to the life length, if the mean difference of a certain photoelectric-thermal characteristic change amount or change rate between the long-life group and the short-life group is much larger than the internal fluctuation of each group and the correlation coefficient with the life performance is close to positive one or negative one, it is considered that the photoelectric-thermal characteristic change amount and change rate have stronger distinguishing ability.

[0081] It should be noted that the corresponding relationship between the photoelectric-thermal change and the life performance in the historical aging data specifically refers to the corresponding rule between the photoelectric-thermal change such as luminous flux change, color coordinate offset, junction temperature change and the time required for the LED chip to reach the failure criterion within a certain period of time in the historical aging data. For example, the LED chip with larger luminous flux decline and continuous junction temperature rise tends to reach the failure criterion earlier, while the LED chip with smaller luminous flux change and stable junction temperature has relatively longer life.

[0082] The photoelectric-thermal change amount and change rate capable of distinguishing different life levels specifically refer to the photoelectric-thermal characteristics whose numerical values present stable differences between the long-life LED chips and the short-life LED chips under the same test time or the same current pulse number. For example, the LED chip with smaller luminous flux decline and slower reverse leakage current rise in the first few current pulses tends to correspond to longer life, while the LED chip with rapid luminous flux decline and faster reverse leakage current growth corresponds to shorter life.

[0083] According to the photoelectric-thermal change characteristic set, the photoelectric-thermal change amount and change rate of each LED chip under each current pulse are spliced according to the preset characteristic order and fixed length, and the photoelectric-thermal pulse life fingerprint of the LED chip for subsequent life estimation is constructed.

[0084] Further, according to the photoelectric thermal change feature set, the photoelectric thermal change amount and rate of each LED chip under each current pulse are sequentially arranged according to the photoelectric thermal order, first sequentially arranged according to the photoelectric thermal change feature amount and rate order recorded in the photoelectric thermal change feature set within a single current pulse, and then sequentially expanded in the current pulse sequence dimension according to the current pulse order to form a one-dimensional feature sequence; the one-dimensional feature sequence of each LED chip is intercepted according to a fixed length, so that the one-dimensional feature sequences of all LED chips are consistent in length, and the one-dimensional feature sequence with uniform length is bound with the corresponding LED chip number to construct the photoelectric thermal pulse life fingerprint of the LED chip for subsequent life estimation.

[0085] It should be noted that the photoelectric thermal pulse life fingerprint of the LED chip condenses the complex photoelectric thermal changes of the LED under pulse stress into a feature set that can sensitively reflect the degradation degree, so that the life rapid estimation model can more accurately judge the chip life and remaining life by relying on a small amount of short-time test data.

[0086] S4, constructing a life rapid estimation model, and predicting the life of the LED chip according to the photoelectric thermal pulse life fingerprint and the working condition information to generate the predicted life of the LED chip.

[0087] The life rapid estimation model is constructed based on an input fusion layer, a physical index extraction layer, a feature representation layer and a life regression layer.

[0088] The input fusion layer splices the photoelectric thermal pulse life fingerprint of the LED chip and the working condition information in a fixed order, and performs standardization processing to generate a unified input feature vector as the starting point for subsequent calculation.

[0089] Further, the input fusion layer reads the photoelectric thermal pulse life fingerprint of each LED chip and the corresponding working condition information, sequentially arranges and splices each dimension photoelectric thermal feature in the photoelectric thermal pulse life fingerprint with the working condition information according to the order, and obtains a complete dimension spliced feature sequence; based on the spliced feature sequence, the mean and standard deviation of the spliced feature sequence of all LED chips in the training phase are counted according to the feature dimension to generate a standardization parameter set for subsequent scale adjustment; the spliced feature sequence of each LED chip is standardized dimension by dimension using the standardization parameter set, so that the photoelectric thermal feature data and the working condition information of different dimensions are in a unified numerical scale, and the spliced feature sequence after standardization is output as a unified input feature vector, providing a unified input for the subsequent physical index extraction layer and feature representation layer.

[0090] It should be noted that the fixed order is the arrangement order of "photoelectric features first, electrical features second, thermal features third, and working condition information last".

[0091] The working condition information refers to the environmental temperature, working current, and driving voltage and other use condition parameters of the LED chip during current pulse testing or actual working, and is obtained by reading the driving power setting value and synchronously recording the data of the environmental monitoring sensor.

[0092] The unified input feature vector reflects the photoelectric-thermal change mode and working condition influence of the light characteristics, electrical characteristics, thermal characteristics, and working condition information of the LED chip evolving with the pulse order under current pulse stress, and the degradation state embodied thereby.

[0093] The physical index extraction layer extracts the photoelectric-thermal response rate information (the light flux change rate per unit time, the reverse leakage current increment, and the junction temperature rise rate) according to the degradation rate rule, and converts it into the degradation rate index and the thermal stress index that can directly reflect the early degradation speed and the thermal stress intensity through the principal component analysis method.

[0094] Further, the physical index extraction layer locates the feature dimensions of the light flux change rate per unit time, the reverse leakage current increment, and the junction temperature rise rate in the photoelectric-thermal response rate information in the unified input feature vector, arranges the photoelectric-thermal response rate information corresponding to each LED chip in each current pulse interval into a photoelectric-thermal response rate matrix in the order of the current pulse; after obtaining the photoelectric-thermal response rate matrix, the photoelectric-thermal response rate information is subjected to covariance analysis and feature vector decomposition through the principal component analysis method, the principal component that can comprehensively represent the light flux change rate and the reverse leakage current increment is extracted as the degradation rate index, and the principal component that can comprehensively represent the junction temperature rise rate is extracted as the thermal stress index, and the degradation rate index and the thermal stress index are output together for subsequent fusion and compression reorganization by the feature representation layer.

[0095] It should be noted that the degradation rate rule is obtained according to the corresponding relationship between different photoelectric-thermal response rates and the length of the life in the historical aging data, and the light flux change rate, the reverse leakage current increment, and the junction temperature rise rate and their statistical calculation methods that can stably distinguish the life level are selected in combination with the thermoelectric physical mechanism analysis, which is used to determine the feature dimensions of the photoelectric-thermal response rate information in the unified input feature vector and to specify the statistical window and the average or extreme value calculation method in each current pulse interval, thereby providing a unified calculation specification for extracting the degradation rate index and the thermal stress index.

[0096] The feature representation layer adopts feature splicing operation to fuse the unified input feature vector, the degradation rate index, and the thermal stress index, and performs compression reorganization through nonlinear transformation, to obtain a life representation vector that comprehensively represents the degradation degree and the use condition influence, and the expression is:

[0097] ;

[0098] wherein, is a life characterization vector, is a ReLU function, is a trainable weight matrix, is a unified input feature vector, is a degradation rate indicator, is a thermal stress indicator, is a bias vector;

[0099] It should be noted that the trainable weight matrix refers to the parameter matrix in the input fusion layer and the feature characterization layer for mapping the unified input feature vector to the life characterization vector, which is obtained by supervised learning using the measured life of part of the calibration samples and gradient iterative optimization.

[0100] The life regression layer estimates the predicted life of the LED chip based on the life characterization vector through gradient boosting decision tree (GBDT), and in the training stage, the error between the predicted life and the measured life is used to optimize the feature characterization layer and the input fusion layer in reverse, forming a complete life fast estimation model.

[0101] Further, the life regression layer takes the life characterization vector output by the feature characterization layer as input, and uses gradient boosting regression tree (GBDT) to perform multi-round iterative fitting on the life characterization vector. In each round, a new regression subtree is generated according to the current residual error and is added to the existing regression structure to obtain the predicted life of the LED chip corresponding to the life characterization vector; in the training stage, the measured life of part of the calibration samples is selected as training data, the predicted life of the LED chip output by the life regression layer is compared with the measured life of part of the calibration samples to calculate the error, and the tree structure parameters of the gradient boosting regression tree (GBDT) are updated according to the error, while the weights and biases in the feature characterization layer and the input fusion layer are adjusted through error back propagation, so that the life characterization vector is more consistent with the measured life response, and finally a life fast estimation model is formed, which can stably output the predicted life of the LED chip.

[0102] The photoelectric thermal pulse life fingerprint and working condition information are input into the life fast estimation model to obtain the predicted life of the LED chip, and the expression is:

[0103] ;

[0104] Wherein, is the predicted life of the LED chip (dimensional hours), is the total number of trees in the gradient boosting regression tree, is the learning rate of the th tree, is the number of leaf nodes of the th tree, is the th leaf node of the The predicted value of each leaf node. It is the first Tree No. The feature space region corresponding to each leaf node. It is an indicator function used to determine the lifetime characterization vector. Whether it falls into the first Tree No. The feature space region corresponding to each leaf node If the value falls into the range, output 1; otherwise, output 0.

[0105] It should be noted that the learning rate is a coefficient used in gradient boosting regression trees to scale the output of the m-th tree to control the overall model update step size, and is determined through parameter tuning during training via cross-validation. The feature space region refers to the set of value ranges in the multidimensional space composed of the features of the lifetime representation vector, defined by the splitting conditions of the gradient boosting regression tree, that cause the lifetime representation vector to be grouped into the same leaf node.

[0106] S5. Based on the photoelectric thermal pulse lifetime fingerprint, perform cluster analysis on LED chips and extract representative samples of each type for accelerated aging verification, obtain the deviation between predicted lifetime and measured lifetime, and optimize the lifetime rapid estimation model.

[0107] Based on the photoelectric and thermal pulse lifetime fingerprint of LED chips, cluster analysis is performed on the LED chips. The photoelectric and thermal pulse lifetime fingerprints are divided into several fingerprint cluster categories according to the degree of similarity, and lifetime fingerprint clustering results containing the category information of each LED chip are generated.

[0108] Furthermore, the photoelectric and thermal pulse lifetime fingerprints of all LED chips are normalized based on a unified dimension and length. Euclidean distance or cosine distance is used as the similarity measure. Unsupervised clustering is performed on the photoelectric and thermal pulse lifetime fingerprints of LED chips using clustering analysis. LED chips with high similarity in photoelectric and thermal change patterns are grouped into the same fingerprint cluster category, resulting in an initial clustering result containing the central features and a list of members for each fingerprint cluster category. Based on the initial clustering result, the number of clusters and clustering stability are checked according to the number of LED chips and the dispersion of photoelectric and thermal changes within each fingerprint cluster category. If necessary, the number of clusters is adjusted and the clustering analysis is repeated. Finally, a lifetime fingerprint clustering result is formed, recording the fingerprint cluster category to which each LED chip belongs. This provides a basis for extracting accelerated aging verification samples from the lifetime fingerprint clustering result.

[0109] Based on the lifetime fingerprint clustering results, several LED chips are extracted from the LED chips corresponding to each type of photoelectric thermal pulse lifetime fingerprint as accelerated aging verification objects, generating an accelerated aging verification sample set.

[0110] Further, according to the life fingerprint clustering result, the LED chips corresponding to each fingerprint clustering category are traversed, a representative score is calculated according to the distance between the opto-thermal pulse life fingerprint and the center of the corresponding fingerprint clustering category and the predicted life distribution of the LED chip, and the LED chips with a shorter distance to the clustering center and within the predicted life range of the category and covering different life levels are marked as candidate representative samples; after obtaining the candidate representative samples of each fingerprint clustering category, according to the preset extraction number of each fingerprint clustering category, the LED chips with a higher representative score and a different distribution in the opto-thermal pulse life fingerprint feature are selected from the candidate representative samples, an accelerated aging verification object list with a fingerprint clustering category label is formed, and the accelerated aging verification object lists of the fingerprint clustering categories are combined to generate an accelerated aging verification sample set containing the LED chip number, the corresponding fingerprint clustering category and the reserved accelerated aging condition configuration information.

[0111] The LED chips in the accelerated aging verification sample set are continuously operated under the accelerated aging condition until the failure criterion is met, the measured life of each LED chip is recorded, and the measured life is compared with the predicted life of the corresponding LED chip to generate a life deviation data set;

[0112] Further, according to the accelerated aging verification sample set, each LED chip is continuously powered and operated under the pre-set accelerated aging condition and the opto-thermal characteristic data is periodically monitored, and when the opto-thermal characteristic data meets the failure criterion, the corresponding operating time is recorded as the measured life of the LED chip; the predicted life of each LED chip in the accelerated aging verification sample set is calculated in advance by using the life rapid estimation model, the measured life and the predicted life of each LED chip are paired according to the LED chip number, the life difference or relative error of each pair of measured life and predicted life is calculated, and the LED chip number and the corresponding accelerated aging condition are arranged and filed together to finally generate a life deviation data set for subsequent optimization of the life rapid estimation model.

[0113] According to the life deviation data set, the parameters of the life regression layer in the life rapid estimation model and the opto-thermal change features in the feature representation layer are optimized and updated.

[0114] Further, according to the life deviation data set, the life deviation data set is associated with the corresponding life characterization vector and the photoelectric thermal pulse life fingerprint of the LED chip according to the LED chip number, a joint sample set containing the life characterization vector, the photoelectric thermal pulse life fingerprint of the LED chip and the life deviation is generated, on the basis of the joint sample set, the life deviation distribution on different life characterization vector intervals and different photoelectric thermal change characteristics is counted, the life deviation data set is grouped according to the life regression layer leaf node number and the photoelectric thermal change characteristic value range respectively, the average absolute life deviation or variance of each group is counted, and the leaf node and the photoelectric thermal change characteristic with larger average absolute life deviation or variance are selected as the objects that contribute more to the life deviation; the tree structure parameters in the life regression layer and the leaf node output value are updated according to the life deviation distribution, and the weight coefficient of the photoelectric thermal change characteristic in the feature characterization layer is adjusted, so as to reduce the life prediction error, and the updated life regression layer and the feature characterization layer jointly constitute the life fast estimation model corrected by the life deviation data set.

[0115] The embodiment also provides a computer device suitable for the case of the LED chip life fast estimation method, including a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute the computer executable instructions to realize the LED chip life fast estimation method proposed in the above embodiment.

[0116] The computer device can be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with external terminals in a wired or wireless manner. The wireless manner can be realized through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer overlaid on the display screen, or a key, trackball or touchpad arranged on the shell of the computer device. In addition, the input device can be an external keyboard, touchpad or mouse, etc.

[0117] The embodiment also provides a storage medium on which a computer program is stored, the program being executed by a processor to implement the method for rapidly estimating the life of an LED chip as proposed in the above embodiment; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as a static random access memory (SRAM), an electrically erasable programmable read-only memory (EEPROM), an erasable programmable read-only memory (EPROM), a programmable read-only memory (PROM), a read-only memory (ROM), a magnetic memory, a flash memory, a magnetic disk or an optical disk.

[0118] To sum up, the present application realizes the processing link of directly outputting the life representation vector and the life result from the short-time photoelectric thermal test result under the given use condition by constructing the photoelectric thermal pulse life fingerprint of the LED chip, compressing the photoelectric thermal evolution process under the pulse stress into a unified life representation carrier, providing the directly callable degradation feature basis for the life modeling, constructing the life rapid estimation model, and generating the predicted life of the LED chip according to the photoelectric thermal pulse life fingerprint and the working condition information, which is convenient for rapidly obtaining the life information of the LED chip in the production screening and reliability evaluation.

[0119] It should be noted that the above embodiment is only used to illustrate the technical solutions of the present application but not limit the present application, although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or replaced equivalently without departing from the spirit and scope of the technical solutions of the present application, and all of them should be covered in the scope of the claims of the present application.

Claims

1. A method for rapid estimation of LED chip lifetime, characterized in that: The application relates to a method for rapid estimation of the service life of LED chips. The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps:

2. The method of claim 1, wherein: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps:

3. The method of claim 1, wherein: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The Under the constraints of safety margin and target degradation excitation strength, the current amplitude pulse width and pulse interval are selected to generate a current pulse stress sequence.

4. The method of claim 3, wherein: The verification completion and mapping storage are performed to generate a photoelectric thermal pulse response dataset, and the steps are as follows, Under a controlled environment temperature, a current pulse stress sequence is applied to each LED chip in sequence, photoelectric thermal response data are synchronously collected before and after each current pulse, and the data are collected according to the LED chip number and the current pulse order to form a photoelectric thermal response original data set; The photoelectric thermal response original data set is subjected to missing value completion, format unification and standardization processing to generate a photoelectric thermal pulse response dataset.

5. The method of claim 1, wherein: The life quick estimation model is constructed, and the life of the LED chip is predicted according to the photoelectric thermal pulse life fingerprint and the working condition information to generate the predicted life of the LED chip, and the steps are as follows, The life quick estimation model is constructed based on an input fusion layer, a physical index extraction layer, a feature characterization layer and a life regression layer; The input fusion layer splices the photoelectric thermal pulse life fingerprint of the LED chip and the working condition information in a fixed order and generates a unified input feature vector through standardization processing; The physical index extraction layer extracts photoelectric thermal response rate information according to a degradation rate rule and converts the information into a degradation rate index and a thermal stress index through principal component analysis; The feature characterization layer fuses the unified input feature vector with the degradation rate index and the thermal stress index, and obtains a life characterization vector through nonlinear transformation compression and reorganization; The life regression layer estimates the predicted life of the LED chip based on the life characterization vector, and reversely optimizes the feature characterization layer and the input fusion layer between the predicted life and the measured life in the training stage to form a complete life quick estimation model; The photoelectric thermal pulse life fingerprint and the working condition information are input into the life quick estimation model to obtain the predicted life of the LED chip.

6. The method of claim 1, wherein: The life quick estimation model is optimized according to the photoelectric thermal pulse life fingerprint, and the steps are as follows, The LED chips are subjected to cluster analysis according to the photoelectric thermal pulse life fingerprint, the photoelectric thermal pulse life fingerprint is divided into several fingerprint cluster categories according to the similarity, and a life fingerprint clustering result is generated; A plurality of LED chips are extracted from the LED chips corresponding to each type of photoelectric thermal pulse life fingerprint as accelerated aging verification objects to generate an accelerated aging verification sample set according to the life fingerprint clustering result; The LED chips in the accelerated aging verification sample set are continuously operated under accelerated aging conditions until the failure criterion is met, the measured life of each LED chip is recorded, and the measured life is compared with the predicted life of the corresponding LED chip to generate a life deviation data set; The parameters of the life regression layer and the photoelectric thermal change features in the feature characterization layer in the life quick estimation model are updated according to the life deviation data set.

7. The method of claim 1, wherein: The photoelectric thermal characteristic data includes luminous flux, color coordinates, peak wavelength, forward voltage, reverse leakage current, dynamic on-resistance, junction temperature estimation value, junction shell thermal resistance and temperature rise amplitude under pulse current.

8. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that: The processor implements the steps of the LED chip life quick estimation method of any one of claims 1-7 when executing the computer program.

9. A computer readable storage medium having stored thereon a computer program, characterized in that: The computer program is executed by the processor to implement the steps of the LED chip life quick estimation method of any one of claims 1-7.

Images