A method and system for predicting the lifetime of light-emitting diodes based on statistical data.
By constructing a multidimensional performance degradation statistical function and combining it with the sequential Bayesian method to update parameters online, the problem of low model fitting accuracy in existing LED lifetime prediction is solved, and efficient and reliable LED lifetime prediction is achieved.
Patent Information
- Application Number
- CN202611236209.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-14
- Publication Date
- 2026-11-10
AI Technical Summary
Existing LED lifetime prediction methods fail to effectively combine multi-dimensional degradation indicators such as forward voltage drift and dominant wavelength shift, resulting in low model fitting accuracy and an inability to achieve high-precision lifetime prediction.
By constructing a multidimensional performance degradation statistical function and combining it with the sequential Bayesian method for online parameter updates, a comprehensive remaining lifetime prediction result is generated using the multidimensional degradation trajectory function of forward voltage, luminous flux, and dominant wavelength.
It improves the fitting accuracy of the degradation trajectory model, realizes efficient automation and reliability of LED lifetime prediction, reduces data bias, and outputs stable and consistent lifetime prediction results.
Smart Images

Figure CN123069879A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of electrical variable processing technology, and in particular to a method and system for predicting the lifetime of light-emitting diodes based on data statistics. Background Technology
[0002] Current LED lifetime prediction methods mostly employ a single photoelectric parameter to construct a degradation model for lifetime extrapolation, relying solely on luminous flux attenuation data to estimate remaining lifetime without simultaneously incorporating multidimensional joint analysis of two key degradation indicators: forward voltage drift and dominant wavelength shift. Offline aging data undergoes only simple averaging, failing to remove outliers and verify sample consistency. Initial baseline parameters are subject to noise interference, resulting in low accuracy in fitting degradation curves. The model cannot accurately match the initial photoelectric characteristics of a single LED under test, leading to significant deviations between the static degradation model and the actual degradation trajectory of the device.
[0003] Traditional LED lifetime prediction models have fixed parameters and cannot be dynamically updated based on real-time device monitoring data. They rely solely on offline aging tests to extrapolate failure times using a fixed degradation function, without incorporating sequential Bayesian statistical methods to correct model coefficients in real time. Each optoelectronic parameter independently outputs a lifetime prediction value, and the minimum value is directly selected as the final lifetime result. This approach fails to allocate differentiated weights based on the fitting residuals of each degradation trajectory to achieve multi-index fusion evaluation. Consequently, the prediction results exhibit poor robustness and large errors, and the overall lifetime prediction efficiency and accuracy are insufficient to meet the requirements of high-precision LED reliability testing. Summary of the Invention
[0004] This invention provides a method and system for predicting the lifetime of light-emitting diodes based on data statistics, in order to solve the problems mentioned in the background art.
[0005] To achieve the above objectives, the present invention provides a method for predicting the lifetime of light-emitting diodes based on statistical data, comprising: Step 1: Perform mean filtering on the initial photoelectric parameter set of the LED under constant current drive to obtain the standard initial parameter set of the LED under test; Step 2: Based on the historical offline aging test data of the LED under test, using the forward voltage, luminous flux and dominant wavelength in the standard initial parameter set as independent variables, and the cumulative lighting time of the LED under test as the independent variable, construct a multidimensional performance degradation statistical function for the LED under test. Step 3: During the operation of the LED under test, based on the real-time photoelectric parameter processing records of multiple historical monitoring periods and the fitting residuals of each degradation trajectory function in the multidimensional performance degradation statistical function, the current parameters of the multidimensional performance degradation statistical function are updated online using the sequential Bayesian method. Step 4: Extrapolate the updated degradation trajectory function to the failure threshold formed by the boundary values of the failure of the performance parameters of the LED under test, and generate the forward voltage lifetime prediction value, luminous flux lifetime prediction value and dominant wavelength lifetime prediction value corresponding to the independent variables in the degradation trajectory function respectively. Step 5: Weight and fuse the predicted forward voltage lifetime, the predicted luminous flux lifetime, and the predicted dominant wavelength lifetime to generate a comprehensive remaining lifetime prediction result for the LED under test.
[0006] In a preferred embodiment, the initial photoelectric parameter set of the LED under constant current drive is subjected to mean filtering to obtain a standard initial parameter set of the LED under test, including: At multiple consecutive sampling times, multiple sets of initial photoelectric parameter sampling values of the LED under test are obtained respectively. Each set of initial photoelectric parameter sampling values includes multiple readings that correspond one-to-one with forward voltage, luminous flux and dominant wavelength. The arithmetic mean of multiple readings belonging to the same parameter type in the initial photoelectric parameter sampling values is used as the periodic average forward voltage value, periodic average luminous flux value, and periodic average dominant wavelength value at the sampling time. Remove periodic average parameter groups whose differences exceed a threshold from the periodic average parameter group consisting of the periodic average forward voltage value, the periodic average luminous flux value, and the periodic average dominant wavelength value. The periodic average parameter set after elimination is arithmetically averaged again, and the average forward voltage value, average luminous flux value and average dominant wavelength value obtained after the second arithmetic average are jointly determined as the standard initial parameter set.
[0007] In a preferred embodiment, the historical offline aging test data based on the LED under test includes: Acquire aging photoelectric datasets of the same batch of LEDs under test at multiple sampling time points recorded in the offline aging test. The aging photoelectric datasets include forward voltage measurement values, luminous flux measurement values, dominant wavelength measurement values, and cumulative lighting duration records corresponding to each sampling time point. For each measurement value belonging to the same sampling time point in the aging photoelectric dataset, a consistency check is performed to remove the measurement value with the largest deviation from the original value. Each verified measurement value is associated with and stored with the corresponding cumulative lighting duration record value to form a standard offline sample set.
[0008] In a preferred embodiment, the construction of a multidimensional performance degradation statistical function for the LED under test, using the forward voltage, luminous flux, and dominant wavelength from the standard initial parameter set as independent variables and the cumulative lighting time of the LED under test as an independent variable, includes: Based on the cumulative lighting duration records and the corresponding changes in the forward voltage measurement, luminous flux measurement, and dominant wavelength measurement values in the standard offline sample set, the forward voltage change curve, luminous flux change curve, and dominant wavelength change curve are fitted respectively. The positive voltage change curve, the luminous flux change curve, and the dominant wavelength change curve are collectively determined as the initial degradation trajectory function set, wherein each of the change curves adopts a curve form with the goal of minimizing the overall fitting residual with the offline sample data.
[0009] In a preferred embodiment, constructing the multidimensional performance degradation statistical function of the LED under test includes: The forward voltage, luminous flux, and dominant wavelength from the standard initial parameter set are respectively used as the initial intercept values of the forward voltage change curve, the luminous flux change curve, and the dominant wavelength change curve at the moment when the cumulative lighting duration is zero. The position parameters of each curve in the initial degenerate trajectory function set are translated and calibrated using the initial intercept value. The calibrated positive voltage variation curve, the luminous flux variation curve, and the dominant wavelength variation curve are combined into the multidimensional performance degradation statistical function.
[0010] In a preferred embodiment, during the operation of the LED under test, the online statistical update of the current parameters of the multidimensional performance degradation statistical function is performed using a sequential Bayesian method based on the fitting residuals of real-time photoelectric parameter processing records from multiple historical monitoring periods and the degradation trajectory functions in the multidimensional performance degradation statistical function. This includes: Extract processed records from multiple consecutive historical monitoring periods prior to the current moment in chronological order. Each processed record contains the positive voltage statistics, luminous flux statistics, and dominant wavelength statistics corresponding to the cumulative lighting duration within the period. Substitute the cumulative lighting duration from each of the processed records into the degradation trajectory functions of the current multidimensional performance degradation statistics function, and calculate the function estimate of each degradation trajectory function at each of the cumulative lighting durations. The formula for calculating the function estimate is as follows: ; In the formula, For the first Each photoelectric parameter during duration The estimated value of the function at that point, For the first The zeroth-order coefficients of a degenerate trajectory function, For the first The first-order coefficients of the degenerate trajectory function For the first The second-order coefficients of a degenerate trajectory function The cumulative lighting duration; The deviation of each function estimate is compared with the statistical value of the same parameter type in the corresponding processed record to obtain the independent residual term of the degenerate trajectory function; For each of the degradation trajectory functions, the independent residual terms belonging to different historical monitoring periods are sorted by absolute value, and a relative influence factor is set for each of the independent residual terms based on the sorting results; Using the processed records and relative influence factors from all the historical monitoring periods, the current parameters of the degradation trajectory function are weighted and revised, and the revised parameters are used as the updated parameters of the degradation trajectory function.
[0011] In a preferred embodiment, the failure threshold, which is composed of the boundary values of the failure of the performance parameters of the LED under test, includes: Critical recorded values of various performance parameters corresponding to the failure state of LEDs in the same batch during offline aging tests are extracted. The critical recorded values include a first critical value corresponding to the forward voltage, a second critical value corresponding to the luminous flux, and a third critical value corresponding to the dominant wavelength. The first critical value, the second critical value, and the third critical value are respectively used as the boundary values for the failure of the forward voltage, the luminous flux, and the dominant wavelength, and together they are used as the failure threshold.
[0012] In a preferred embodiment, generating the predicted positive voltage lifetime, predicted luminous flux lifetime, and predicted dominant wavelength lifetime corresponding to the independent variables in the degradation trajectory function includes: The updated degradation trajectory functions are used as continuous curves of their respective parameters as the cumulative lighting duration. Starting from the current cumulative lighting duration position on each of the continuously changing curves, extend the curve in the direction of increasing duration until the function value of the curve reaches the corresponding failure boundary value, and read the duration value corresponding to the boundary value. The duration at which the degradation trajectory function corresponding to the positive voltage reaches the first critical value is recorded as the predicted positive voltage lifetime value. The duration during which the degradation trajectory function corresponding to the luminous flux reaches the second critical value is recorded as the luminous flux lifetime prediction value. The duration at which the degradation trajectory function corresponding to the dominant wavelength reaches the third critical value is recorded as the dominant wavelength lifetime prediction value.
[0013] In a preferred embodiment, the weighted fusion of the predicted forward voltage lifetime, the predicted luminous flux lifetime, and the predicted dominant wavelength lifetime to generate a comprehensive remaining lifetime prediction result for the LED under test includes: Based on the degradation trajectory function, the absolute values of each independent residual term obtained in each historical monitoring period are taken and accumulated to obtain the residual cumulative value of the degradation trajectory function. The residual cumulative value corresponds to the three residual cumulative values of the positive voltage, the light flux and the dominant wavelength, respectively. The reciprocals of the three cumulative residual values are used as the initial weight base values for each parameter. The three initial weight base values are simultaneously scaled by the same ratio so that the sum of the three scaled values is one. The scaled values are then used as the first weighting coefficient of the forward voltage lifetime prediction value, the second weighting coefficient of the luminous flux lifetime prediction value, and the third weighting coefficient of the dominant wavelength lifetime prediction value, respectively. Multiply the predicted forward voltage lifetime by the first weighting coefficient, the predicted luminous flux lifetime by the second weighting coefficient, and the predicted dominant wavelength lifetime by the third weighting coefficient; The three product results are summed, and the sum is used as the comprehensive remaining lifetime prediction result of the LED under test.
[0014] To address the above problems, the present invention also provides a data statistics-based light-emitting diode lifetime prediction system, the system comprising: The initial parameter processing module is used to perform mean filtering on the initial photoelectric parameter set of the LED under constant current drive to obtain the standard initial parameter set of the LED under test. The degradation statistical model module is used to construct a multidimensional performance degradation statistical function of the LED under test based on the historical offline aging test data of the LED under test, with the forward voltage, luminous flux and dominant wavelength in the standard initial parameter set as independent variables and the cumulative lighting time of the LED under test as independent variables. The parameter online update module is used to update the current parameters of the multidimensional performance degradation statistical function online by means of the sequential Bayesian method, based on the real-time photoelectric parameter processing records of multiple historical monitoring periods and the fitting residuals of each degradation trajectory function in the multidimensional performance degradation statistical function. The extrapolation lifetime prediction module is used to extrapolate the updated degradation trajectory function to the failure threshold formed by the boundary values of the failure of the performance parameters of the LED under test, and generate the forward voltage lifetime prediction value, luminous flux lifetime prediction value and dominant wavelength lifetime prediction value corresponding to the independent variables in the degradation trajectory function, respectively. The weighted fusion evaluation module is used to weight and fuse the predicted forward voltage lifetime, the predicted luminous flux lifetime, and the predicted dominant wavelength lifetime to generate a comprehensive remaining lifetime prediction result for the LED under test.
[0015] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention constructs a noise-free standard initial parameter set by filtering the mean of multi-time period sampling and removing abnormal parameter groups. Simultaneously, it performs consistency verification and outlier removal on the offline aging dataset. Based on the joint fitting of multi-dimensional optoelectronic parameters, it generates a degradation statistical function of the adapter device's initial characteristics, accurately depicting the degradation changes of LED forward voltage, luminous flux, and dominant wavelength with lighting time. This effectively improves the fitting accuracy of the degradation trajectory model and reduces data deviation in the basic modeling stage. Based on the sequential Bayesian method combined with real-time monitoring of residuals to dynamically update the degradation function coefficients, and by allocating influence factors according to the absolute value of the residuals to revising the model parameters, the degradation model achieves continuous adaptive calibration according to the device's operating state, extending the effective adaptation period of the model.
[0016] 2. This invention establishes a multi-dimensional failure threshold based on the aging failure critical value of the same batch, extrapolates the independent lifetime prediction values corresponding to three types of photoelectric parameters, and then dynamically allocates normalized weighting coefficients according to the cumulative residual values of each degradation trajectory to complete the multi-index fusion calculation and output a comprehensive remaining lifetime result. The entire process relies on data statistical methods to achieve full-process quantitative calculation, which can stably output LED remaining lifetime prediction results with high consistency and smaller errors. The overall prediction process has a high degree of automation, significantly improving the calculation efficiency and reliability of LED lifetime prediction results. Attached Figure Description
[0017] Figure 1 A flowchart illustrating a data statistics-based method for predicting the lifetime of light-emitting diodes (LEDs) according to an embodiment of the present invention is shown below. Figure 2 A functional block diagram of a data statistics-based LED lifetime prediction system provided in an embodiment of the present invention; The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation
[0018] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
[0019] This application provides a method for predicting the lifespan of light-emitting diodes (LEDs) based on statistical data. The execution entity of this method includes, but is not limited to, at least one of the following electronic devices that can be configured to execute the method provided in this application: a server, a terminal, etc. In other words, the method for predicting the lifespan of LEDs based on statistical data can be executed by software or hardware installed on a terminal device or a server device. The server includes, but is not limited to, a single server, a server cluster, a cloud server, or a cloud server cluster. The server can be an independent server or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms.
[0020] Reference Figure 1 The diagram shown is a flowchart illustrating a data-based statistical method for predicting the lifetime of a light-emitting diode (LED) according to an embodiment of the present invention. In this embodiment, the data-based statistical method for predicting the lifetime of an LED includes: Step 1: Perform mean filtering on the initial photoelectric parameter set of the LED under constant current drive to obtain the standard initial parameter set of the LED under test; In this embodiment of the invention, the initial photoelectric parameter set of the LED under constant current drive is subjected to mean filtering to obtain the standard initial parameter set of the LED under test, including: At multiple consecutive sampling times, multiple sets of initial photoelectric parameter sampling values of the LED under test are obtained respectively. Each set of initial photoelectric parameter sampling values includes multiple readings that correspond one-to-one with forward voltage, luminous flux and dominant wavelength. The arithmetic mean of multiple readings belonging to the same parameter type in the initial photoelectric parameter sampling values is used as the periodic average forward voltage value, periodic average luminous flux value, and periodic average dominant wavelength value at the sampling time. Remove periodic average parameter groups whose differences exceed a threshold from the periodic average parameter group consisting of the periodic average forward voltage value, the periodic average luminous flux value, and the periodic average dominant wavelength value. The periodic average parameter set after elimination is arithmetically averaged again, and the average forward voltage value, average luminous flux value and average dominant wavelength value obtained after the second arithmetic average are jointly determined as the standard initial parameter set.
[0021] According to the preset continuous acquisition time nodes, the photoelectric data acquisition operation of the LED under test is completed. Each acquisition time node corresponds to the generation of a complete set of initial photoelectric parameter sampling values. Each set of initial photoelectric parameter sampling values synchronously records three types of measured data that are matched one-to-one: forward voltage reading, luminous flux reading, and dominant wavelength reading. After all acquisitions are completed, the multiple sets of initial photoelectric parameter sampling values generated in all time periods are completely retained.
[0022] For all forward voltage readings generated at the same acquisition time node, a unified numerical summation operation is performed. The total value obtained by the summation operation is divided by the total number of forward voltage reading records for that time period. After the operation, the periodic average forward voltage value corresponding to that acquisition time node is obtained. The same numerical summation operation is used to process all luminous flux readings for that time period. After the operation, the periodic average luminous flux value corresponding to that acquisition time node is obtained. The same numerical operation logic is used to process all dominant wavelength readings for that time period. After the operation, the periodic average dominant wavelength value corresponding to that acquisition time node is obtained. The periodic average forward voltage value, periodic average luminous flux value, and periodic average dominant wavelength value generated at the same acquisition time node are integrated into a complete periodic average parameter group. All periodic average parameter groups generated at all acquisition time nodes are completely preserved.
[0023] The parameter difference judgment thresholds, which are pre-set according to the factory testing standards of LEDs of the same specifications, are retrieved. The periodic average forward voltage value, periodic average luminous flux value, and periodic average dominant wavelength value of each complete periodic average parameter group are extracted one by one. The difference between the three values in a single group and the overall benchmark value of the corresponding parameters in the entire periodic average parameter group is calculated. If the calculated result of the difference of any parameter is greater than the preset parameter difference judgment threshold, the complete periodic average parameter group is directly removed. Only the periodic average parameter group in which the calculated result of the difference of all three parameters is less than or equal to the preset parameter difference judgment threshold is retained.
[0024] After the rejection process, extract all periodic average forward voltage values from the remaining periodic average parameter groups. Sum all extracted periodic average forward voltage values to obtain the total forward voltage value. Divide the total forward voltage value by the total number of remaining periodic average parameter groups to calculate the average forward voltage value. Extract all periodic average luminous flux values from the remaining periodic average parameter groups. Repeat the process of summing all values and dividing by the total number of remaining periodic average parameter groups to calculate the average luminous flux value. Extract all periodic average dominant wavelength values from the remaining periodic average parameter groups. Repeat the process of summing all values and dividing by the total number of remaining periodic average parameter groups to calculate the average dominant wavelength value. Integrate the calculated average forward voltage value, average luminous flux value, and average dominant wavelength value into a set. This set of data is directly defined as the standard initial parameter set corresponding to the LED under test.
[0025] The beneficial effect is that by collecting three types of photoelectric readings completely over multiple time periods, performing stratified mean calculations and filtering out abnormal parameter combinations based on preset difference thresholds, the data is purified step by step to obtain a standardized initial parameter set. This can eliminate data disturbances caused by single sampling and ensure that the benchmark photoelectric data used to construct the degradation model has stable and unified benchmark attributes.
[0026] Step 2: Based on the historical offline aging test data of the LED under test, using the forward voltage, luminous flux and dominant wavelength in the standard initial parameter set as independent variables, and the cumulative lighting time of the LED under test as the independent variable, construct a multidimensional performance degradation statistical function for the LED under test. In this embodiment of the invention, the historical offline aging test data based on the LED under test includes: Acquire aging photoelectric datasets of the same batch of LEDs under test at multiple sampling time points recorded in the offline aging test. The aging photoelectric datasets include forward voltage measurement values, luminous flux measurement values, dominant wavelength measurement values, and cumulative lighting duration records corresponding to each sampling time point. For each measurement value belonging to the same sampling time point in the aging photoelectric dataset, a consistency check is performed to remove the measurement value with the largest deviation from the original value. Each verified measurement value is associated with and stored with the corresponding cumulative lighting duration record value to form a standard offline sample set.
[0027] The method involves constructing a multidimensional performance degradation statistical function for the LED under test, using the forward voltage, luminous flux, and dominant wavelength from the standard initial parameter set as independent variables, and the cumulative illumination time of the LED under test as the independent variable. This function includes: Based on the cumulative lighting duration records and the corresponding changes in the forward voltage measurement, luminous flux measurement, and dominant wavelength measurement values in the standard offline sample set, the forward voltage change curve, luminous flux change curve, and dominant wavelength change curve are fitted respectively. The positive voltage change curve, the luminous flux change curve, and the dominant wavelength change curve are collectively determined as the initial degradation trajectory function set, wherein each of the change curves adopts a curve form with the goal of minimizing the overall fitting residual with the offline sample data.
[0028] The construction of the multidimensional performance degradation statistical function of the LED under test includes: The forward voltage, luminous flux, and dominant wavelength from the standard initial parameter set are respectively used as the initial intercept values of the forward voltage change curve, the luminous flux change curve, and the dominant wavelength change curve at the moment when the cumulative lighting duration is zero. The position parameters of each curve in the initial degenerate trajectory function set are translated and calibrated using the initial intercept value. The calibrated positive voltage variation curve, the luminous flux variation curve, and the dominant wavelength variation curve are combined into the multidimensional performance degradation statistical function.
[0029] All test records of the entire offline aging test process of all LEDs in the same batch were retrieved. Data were assigned according to the fixed acquisition time set during the test. A complete aging optoelectronic dataset was generated for each acquisition time. The forward voltage measurement value, luminous flux measurement value, and dominant wavelength measurement value obtained at that time were stored synchronously within each aging optoelectronic dataset. The cumulative lighting time of the device from the start of the test to the current acquisition time was also recorded synchronously. All aging optoelectronic datasets generated at all acquisition times were fully included.
[0030] Extract all forward voltage measurements belonging to the same acquisition time individually, calculate the intermediate reference value of all forward voltage measurements in the group, calculate the difference between each forward voltage measurement and the intermediate reference value, compare the absolute values of all differences, and remove the forward voltage measurement corresponding to the maximum absolute value. Perform the same reference value acquisition, difference calculation, and maximum difference data removal operation on all luminous flux measurements belonging to the same acquisition time. Repeat the entire set of verification and elimination operations on all main wavelength measurements belonging to the same acquisition time. The data remaining after completing the three types of measurement value elimination operations are the measurement values that have passed the consistency verification.
[0031] Extract the forward voltage measurement value, luminous flux measurement value, and dominant wavelength measurement value retained after the consistency verification operation. Establish a one-to-one binding relationship between the three types of verified measurement values and the cumulative lighting duration record value recorded at the corresponding acquisition time of the data set. Collect and integrate all the data that have been bound and associated. After integration, the complete set of data is defined as a standard offline sample set.
[0032] Extract all cumulative illumination duration records from the complete standard offline sample set. Simultaneously extract the forward voltage measurement value associated with each cumulative illumination duration record. Compare the changes in the forward voltage measurement value when the cumulative illumination duration records change within the same sample set. Construct a continuous change trajectory based on the one-to-one correspondence between the two sets of data to complete the generation of the forward voltage change curve. Extract all cumulative illumination duration records from the complete standard offline sample set. Simultaneously extract the luminous flux measurement value associated with each cumulative illumination duration record. Compare the changes in the forward voltage measurement value with each cumulative illumination duration record within the same sample set. The changes in the recorded illumination duration are synchronized with the changes in the measured luminous flux. A continuous trajectory is constructed based on the one-to-one correspondence between two sets of data, thus completing the generation of the luminous flux change curve. All cumulative illumination duration records within the complete standard offline sample set are extracted, and the dominant wavelength measurement value associated with each cumulative illumination duration record is extracted simultaneously. The changes in the dominant wavelength measurement value synchronized with the changes in the cumulative illumination duration records within the same sample set are compared sequentially. A continuous trajectory is constructed based on the one-to-one correspondence between two sets of data, thus completing the generation of the dominant wavelength change curve.
[0033] The generated curves for positive voltage variation, luminous flux variation, and dominant wavelength variation are integrated into a single dataset. This integrated set of curves is named the Initial Degradation Trajectory Function Group. Various curve shapes are selected and matched against all data in the standard offline sample set. The total deviation of each curve shape from all sample data is calculated. The curve shape with the lowest total deviation is selected as the fixed shape for the positive voltage variation curve. The same comparative statistical method is used to select the fixed curve shape corresponding to the luminous flux variation curve. The entire comparative statistical logic is reused to determine the fixed curve shape corresponding to the dominant wavelength variation curve, ensuring that each curve in the Initial Degradation Trajectory Function Group matches a curve shape that meets the minimum standard for the overall fitting residual value.
[0034] Extract the average forward voltage value retained within the complete standard initial parameter set, and fix this value as the initial intercept value corresponding to the moment when the cumulative lighting time reaches zero, as shown on the forward voltage change curve. Extract the average luminous flux value retained within the complete standard initial parameter set, and fix this value as the initial intercept value corresponding to the moment when the cumulative lighting time reaches zero, as shown on the luminous flux change curve. Extract the average dominant wavelength value retained within the complete standard initial parameter set, and fix this value as the initial intercept value corresponding to the moment when the cumulative lighting time reaches zero, as shown on the dominant wavelength change curve. The initial intercept values of the three types of curves, after matching, are retained for subsequent calibration operations.
[0035] Retrieve the positive voltage change curve contained within the initial degradation trajectory function group. Adjust the overall coordinate position of the curve according to the initial intercept value set for that curve, so that the corresponding position of the curve when the cumulative lighting time reaches zero completely coincides with the initial intercept value. Retrieve the luminous flux change curve contained within the initial degradation trajectory function group. Adjust the overall coordinate position of the curve according to the initial intercept value set for that curve, so that the corresponding position of the curve when the cumulative lighting time reaches zero completely coincides with the initial intercept value. Retrieve the dominant wavelength change curve contained within the initial degradation trajectory function group. Adjust the overall coordinate position of the curve according to the initial intercept value set for that curve, so that the corresponding position of the curve when the cumulative lighting time reaches zero completely coincides with the initial intercept value. After completing the adjustment of all curve positions, three independent change curves after translation calibration are obtained.
[0036] The system collects all the forward voltage variation curves, luminous flux variation curves, and dominant wavelength variation curves that have undergone translation calibration. It integrates the three calibrated curves into a unified dataset, and the integrated set of curves is uniformly named the Multidimensional Performance Degradation Statistical Function.
[0037] The beneficial effects are as follows: a standardized aging optoelectronic dataset is constructed based on complete offline aging test time-series data; the measurement data with the highest deviation in each group is eliminated by benchmark difference comparison; the duration is bound with the calibrated measurement data to form a standard offline sample set; three types of optoelectronic parameter change curves are generated based on the inherent correspondence of the samples; and the curve shape with the lowest overall data deviation is selected to form an initial degradation trajectory function group. The initial intercept of each curve is set to zero duration using the standard initial parameters of the device; the translation calibration of all curves is completed based on the intercept; and the calibrated curves are integrated to form a multi-dimensional performance degradation statistical function. This can eliminate the interference of abnormal data in offline tests, unify the initial benchmark of the curves, completely restore the multi-dimensional synchronous aging degradation law of the device, and improve the matching degree between the degradation model and the initial state of a single device under test.
[0038] Step 3: During the operation of the LED under test, based on the real-time photoelectric parameter processing records of multiple historical monitoring periods and the fitting residuals of each degradation trajectory function in the multidimensional performance degradation statistical function, the current parameters of the multidimensional performance degradation statistical function are updated online using the sequential Bayesian method. In this embodiment of the invention, the step of updating the current parameters of the multidimensional performance degradation statistical function online using a sequential Bayesian method, based on the fitting residuals of real-time photoelectric parameter processing records from multiple historical monitoring periods and the degradation trajectory functions in the multidimensional performance degradation statistical function, includes: Extract processed records from multiple consecutive historical monitoring periods prior to the current moment in chronological order. Each processed record contains the positive voltage statistics, luminous flux statistics, and dominant wavelength statistics corresponding to the cumulative lighting duration within the period. Substitute the cumulative lighting duration from each of the processed records into the degradation trajectory functions of the current multidimensional performance degradation statistics function, and calculate the function estimate of each degradation trajectory function at each of the cumulative lighting durations. The formula for calculating the function estimate is as follows: ; In the formula, For the first Each photoelectric parameter during duration The estimated value of the function at that point, For the first The zeroth-order coefficients of a degenerate trajectory function, For the first The first-order coefficients of the degenerate trajectory function For the first The second-order coefficients of a degenerate trajectory function The cumulative lighting duration; The deviation of each function estimate is compared with the statistical value of the same parameter type in the corresponding processed record to obtain the independent residual term of the degenerate trajectory function; For each of the degradation trajectory functions, the independent residual terms belonging to different historical monitoring periods are sorted by absolute value, and a relative influence factor is set for each of the independent residual terms based on the sorting results; Using the processed records and relative influence factors from all the historical monitoring periods, the current parameters of the degradation trajectory function are weighted and revised, and the revised parameters are used as the updated parameters of the degradation trajectory function.
[0039] The zero-order coefficient is taken from the basic reference value of the photoelectric parameter degradation trajectory function inside the multidimensional performance degradation statistical function after the translation calibration operation is completed. This reference value is locked after the curve translation calibration is completed based on the photoelectric parameter values inside the standard initial parameter set. The calibration operation uses the initial intercept value corresponding to the cumulative lighting time returning to zero as the unified adjustment standard. The basic reference value is determined after fitting and screening the curve shape of all offline aging standard sample sets.
[0040] The first-order coefficients are taken from the fixed values of the parameter variation amplitude corresponding to a single change in duration within the photoelectric parameter degradation trajectory function after the initial degradation trajectory function group is screened and finalized. These values are determined by fully matching the changing trends of multiple sets of cumulative lighting duration records and corresponding photoelectric measurement values in the standard offline sample set. The lowest total overall fitting residual is used as the screening criterion throughout the process.
[0041] The second-order coefficients are taken from the fixed values of the parameter variation amplitude caused by the second change in the duration of the degradation trajectory function corresponding to the initial degradation trajectory function group after it is finalized. These values are generated by fitting the time series data of the complete offline aging process of the same batch of devices. All abnormal measurement data have been removed by consistency verification. The values are locked only by the standard offline sample set after verification.
[0042] The cumulative lighting duration is taken from the duration record content synchronously stored in the processed records of each historical monitoring period during the operation of the device under test. The record value is continuously accumulated from the moment the device is first powered on and lit. Each monitoring period acquisition operation will synchronously update the duration record and store it in the corresponding processed record.
[0043] The output value at the corresponding position of the photoelectric parameter is the theoretical output value generated after substituting the degradation trajectory function of the corresponding photoelectric parameter into the cumulative lighting time. This value is used to compare the deviation with the statistical value of the photoelectric parameter obtained by real-time monitoring, and to generate the independent residual term corresponding to a single time period.
[0044] According to the chronological order of the monitoring records generated by the device operation, all continuous and uninterrupted historical monitoring periods before the current acquisition time are retrieved and processed records are retrieved. Each processed record retains the positive voltage statistics, luminous flux statistics, and dominant wavelength statistics generated for the cumulative lighting time of the device during the corresponding monitoring period. All retrieved processed records are retained for subsequent comparison and calculation.
[0045] Retrieve each degradation trajectory function contained within the multidimensional performance degradation statistics function currently in use. Match the cumulative illumination duration of the records within a single processed record to the degradation trajectory function corresponding to the positive voltage to obtain the theoretical output value corresponding to that duration. Match the cumulative illumination duration of the records within the same processed record to the degradation trajectory function corresponding to the luminous flux to obtain the theoretical output value corresponding to that duration. Match the cumulative illumination duration of the records within the same processed record to the degradation trajectory function corresponding to the dominant wavelength to obtain the theoretical output value corresponding to that duration. After all matching and value retrieval operations are completed, generate the estimated values of the three types of functions for each processed record.
[0046] Extract the function estimate value of a single degradation trajectory function corresponding to the same historical monitoring period, extract the statistical value of the same photoelectric parameter in the processed records of that period, and calculate the difference between the two sets of values as the independent residual term generated for this degradation trajectory function in that period. Repeat the entire difference calculation operation for the degradation trajectory functions corresponding to positive voltage, luminous flux, and dominant wavelength respectively, and fully collect all the independent residual terms generated in all periods.
[0047] Extract all independent residuals generated during all historical monitoring periods from a single degradation trajectory function. Calculate the absolute value (without positive or negative distinction) for each independent residual. Arrange all independent residuals in a fixed order from low to high absolute values. Assign a unique relative influence factor to each independent residual based on their order of arrangement. Perform the same sorting and factor assignment operation on the degradation trajectory functions corresponding to positive voltage, luminous flux, and dominant wavelength respectively. Completely retain the relative influence factors associated with all independent residuals.
[0048] The system integrates all processed records corresponding to all historical monitoring periods and the relative influence factors matched for each independent residual term. Using the relative influence factors as the allocation benchmark, it adjusts the correction intensity of the original content of the degradation trajectory function for each period. Based on the allocated correction intensity, it completes the overall revision operation of the original content of the degradation trajectory function. After the revision operation is completed, the adjusted function content is obtained. The adjusted function content is uniformly identified as the updated parameter content of the degradation trajectory function. The entire set of revision and update operations is completed sequentially for all degradation trajectory functions within the multidimensional performance degradation statistical function.
[0049] The beneficial effects include providing a real-time computational foundation for retrieving complete historical monitoring records in chronological order, estimating theoretical output values for each time period based on multidimensional performance degradation statistical functions, obtaining residual terms by calculating the difference between measured statistical values and theoretical values, allocating differentiated influencing factors based on the absolute value of the residuals, and completing weighted revision and updates of the degradation trajectory function by combining records and corresponding influencing factors. This enables continuous dynamic correction of the degradation model content based on real-time device operating data, reducing the deviation between theoretically calculated values and the actual photoelectric state of the device, and continuously maintaining the degradation model's ability to match the real-time aging state of the device.
[0050] Step 4: Extrapolate the updated degradation trajectory function to the failure threshold formed by the boundary values of the failure of the performance parameters of the LED under test, and generate the forward voltage lifetime prediction value, luminous flux lifetime prediction value and dominant wavelength lifetime prediction value corresponding to the independent variables in the degradation trajectory function respectively. In this embodiment of the invention, the failure threshold, which is composed of the boundary values of the failure of the performance parameters of the light-emitting diode under test, includes: Critical recorded values of various performance parameters corresponding to the failure state of LEDs in the same batch during offline aging tests are extracted. The critical recorded values include a first critical value corresponding to the forward voltage, a second critical value corresponding to the luminous flux, and a third critical value corresponding to the dominant wavelength. The first critical value, the second critical value, and the third critical value are respectively used as the boundary values for the failure of the forward voltage, the luminous flux, and the dominant wavelength, and together they are used as the failure threshold.
[0051] The step of generating the predicted positive voltage lifetime, predicted luminous flux lifetime, and predicted dominant wavelength lifetime corresponding to the independent variables in the degradation trajectory function includes: The updated degradation trajectory functions are used as continuous curves of their respective parameters as the cumulative lighting duration. Starting from the current cumulative lighting duration position on each of the continuously changing curves, extend the curve in the direction of increasing duration until the function value of the curve reaches the corresponding failure boundary value, and read the duration value corresponding to the boundary value. The duration at which the degradation trajectory function corresponding to the positive voltage reaches the first critical value is recorded as the predicted positive voltage lifetime value. The duration during which the degradation trajectory function corresponding to the luminous flux reaches the second critical value is recorded as the luminous flux lifetime prediction value. The duration at which the degradation trajectory function corresponding to the dominant wavelength reaches the third critical value is recorded as the dominant wavelength lifetime prediction value.
[0052] Retrieve complete test archive records of all LEDs in the same batch that have completed the complete offline aging test process, screen all photoelectric parameter records retained at the acquisition time corresponding to the time when the device was determined to have reached the failure state during the aging process, separate the measured value corresponding to the forward voltage recorded at the failure determination time and define it as the first critical value, separate the measured value corresponding to the luminous flux recorded at the failure determination time and define it as the second critical value, separate the measured value corresponding to the dominant wavelength recorded at the failure determination time and define it as the third critical value, and retain the three types of critical values to construct the failure determination criteria.
[0053] The first critical value is fixed as the failure boundary value corresponding to the failure of the forward voltage determination device, the second critical value is fixed as the failure boundary value corresponding to the failure of the optical flux determination device, and the third critical value is fixed as the failure boundary value corresponding to the failure of the main wavelength determination device. The three types of failure boundary values that are matched with the corresponding photoelectric parameters are integrated and included into a set of judgment standards. The integrated set of judgment standards is uniformly named failure threshold.
[0054] The degradation trajectory functions corresponding to the forward voltage, luminous flux, and dominant wavelength, which have been updated by sequential Bayesian method after parameter weighting and revision, are retrieved. Each updated degradation trajectory function is then completely transformed into a continuous curve showing the corresponding photoelectric parameters changing synchronously with the cumulative lighting time. Three independent continuous curves are then completely retained for time extrapolation operations.
[0055] Locate the coordinate point corresponding to the current cumulative lighting time of the LED under test recorded inside each continuously changing curve. Use this point as the starting reference point for the extrapolation operation. Extend the entire continuously changing curve along the direction of continuous increase in cumulative lighting time. Continuously compare the output value of the photoelectric parameter corresponding to the curve with the failure boundary value matching the failure threshold during the extension process. Terminate the curve extension operation when the curve output value and the corresponding failure boundary value are completely matched, and record the cumulative lighting time value corresponding to the matching point.
[0056] Retrieve the cumulative lighting duration value of the continuous change curve corresponding to the positive voltage and the corresponding termination point record, and name this duration value as the positive voltage lifespan prediction value and save it separately for archiving.
[0057] Retrieve the cumulative illumination duration value of the continuous change curve corresponding to the luminous flux and the corresponding termination point record, and name this duration value as the luminous flux lifetime prediction value and save it separately for archiving.
[0058] Retrieve the cumulative lighting duration value of the continuous change curve corresponding to the main wavelength and the corresponding termination point. Name this duration value as the main wavelength lifetime prediction value and save it separately for archiving.
[0059] The beneficial effects are that the critical values of three types of photoelectric parameters are extracted from the original offline aging failure records of devices in the same batch, and a standardized failure threshold is formed as a unified failure judgment basis. Based on the real-time updated degradation trajectory function, a continuous change curve is generated. The curve is extrapolated from the current working time of the device until the corresponding failure boundary value is matched, and the independent lifetime prediction results of the three types of photoelectric parameters are obtained. It is possible to set a unified failure judgment standard based on the real failure data of batch devices, accurately calculate the working time corresponding to the failure state of each photoelectric parameter in different dimensions, and realize the independent quantitative output of multi-dimensional lifetime indicators.
[0060] Step 5: Weight and fuse the predicted forward voltage lifetime, the predicted luminous flux lifetime, and the predicted dominant wavelength lifetime to generate a comprehensive remaining lifetime prediction result for the LED under test.
[0061] In this embodiment of the invention, the weighted fusion of the predicted forward voltage lifetime, the predicted luminous flux lifetime, and the predicted dominant wavelength lifetime to generate a comprehensive remaining lifetime prediction result for the LED under test includes: Based on the degradation trajectory function, the absolute values of each independent residual term obtained in each historical monitoring period are taken and accumulated to obtain the residual cumulative value of the degradation trajectory function. The residual cumulative value corresponds to the three residual cumulative values of the positive voltage, the light flux and the dominant wavelength, respectively. The reciprocals of the three cumulative residual values are used as the initial weight base values for each parameter. The three initial weight base values are simultaneously scaled by the same ratio so that the sum of the three scaled values is one. The scaled values are then used as the first weighting coefficient of the forward voltage lifetime prediction value, the second weighting coefficient of the luminous flux lifetime prediction value, and the third weighting coefficient of the dominant wavelength lifetime prediction value, respectively. Multiply the predicted forward voltage lifetime by the first weighting coefficient, the predicted luminous flux lifetime by the second weighting coefficient, and the predicted dominant wavelength lifetime by the third weighting coefficient; The three product results are summed, and the sum is used as the comprehensive remaining lifetime prediction result of the LED under test.
[0062] The degradation trajectory function corresponding to the positive voltage is retrieved and matched against all independent residual terms generated during all historical monitoring periods. After removing the inherent positive and negative attributes of each independent residual term, all values are summed and aggregated to obtain the cumulative residual value corresponding to the positive voltage. The degradation trajectory function corresponding to the luminous flux is retrieved and matched against all independent residual terms generated during all historical monitoring periods. After removing the inherent positive and negative attributes of each independent residual term, all values are summed and aggregated to obtain the cumulative residual value corresponding to the luminous flux. The degradation trajectory function corresponding to the dominant wavelength is retrieved and matched against all independent residual terms generated during all historical monitoring periods. After removing the inherent positive and negative attributes of each independent residual term, all values are summed and aggregated to obtain the cumulative residual value corresponding to the dominant wavelength. The cumulative residual values generated by all three types are completely retained for weight calculation operations.
[0063] The reverse value corresponding to the cumulative residual value of the forward voltage is taken as the initial weight base value for matching the forward voltage lifetime prediction value. The reverse value corresponding to the cumulative residual value of the luminous flux is taken as the initial weight base value for matching the luminous flux lifetime prediction value. The reverse value corresponding to the cumulative residual value of the dominant wavelength is taken as the initial weight base value for matching the dominant wavelength lifetime prediction value. All three sets of initial weight base values are included for unified scaling processing.
[0064] The three initial weight base values are uniformly incorporated into the same scaling adjustment process, and the overall size of the three sets of values is continuously and synchronously adjusted until the sum of the three sets of scaled values is consistent with the fixed judgment benchmark value. After the scaling operation is completed, the value corresponding to the positive voltage after the scaling operation is defined as the first weighting coefficient, the value corresponding to the luminous flux after the scaling operation is defined as the second weighting coefficient, and the value corresponding to the main wavelength after the scaling operation is defined as the third weighting coefficient. The three sets of weighting coefficients are completely retained for fusion calculation.
[0065] The pre-reserved and archived forward voltage lifetime prediction value is extracted and multiplied by the first weighting coefficient. The pre-reserved and archived luminous flux lifetime prediction value is extracted and multiplied by the second weighting coefficient. The pre-reserved and archived dominant wavelength lifetime prediction value is extracted and multiplied by the third weighting coefficient. The product results of the three sets of multiplication operations are completely retained for subsequent summary operations.
[0066] The product results of the three independent calculations are all incorporated into a unified numerical summary process. The three product values are then added together to obtain a single summary value. This summary value is then uniformly named the comprehensive remaining lifetime prediction result of the LED under test and the final retention output is completed.
[0067] The beneficial effects are as follows: by summarizing all residual values of each photoelectric parameter without positive or negative attributes, the cumulative residual value is obtained; based on the cumulative residual value, the initial weight base value is generated in reverse and uniformly scaled to obtain the standardized weighting coefficient; the weighting coefficient is used to match the lifetime prediction value of each dimension to complete the numerical calculation; and finally, all calculation results are summarized to obtain a unified comprehensive remaining lifetime prediction result. It can allocate differentiated weights according to the degree of fit between each degradation trajectory function and the actual operating data of the device, weaken the interference of parameters with higher data deviation on the final prediction result, and improve the matching degree between the comprehensive lifetime prediction result and the actual aging state of the device.
[0068] like Figure 2 The diagram shown is a functional block diagram of a light-emitting diode lifetime prediction system based on data statistics provided in an embodiment of the present invention.
[0069] The LED lifetime prediction system M00 based on statistical data described in this invention can be installed in an electronic device. Depending on the functions implemented, the LED lifetime prediction system M00 may include an initial parameter processing module M01, a degradation statistical model module M02, an online parameter update module M03, an extrapolation lifetime prediction module M04, and a weighted fusion evaluation module M05. The modules described in this invention can also be referred to as units, which are a series of computer program segments that can be executed by the processor of an electronic device and perform a fixed function, stored in the memory of the electronic device.
[0070] In this embodiment, the functions of each module / unit are as follows: The initial parameter processing module M01 is used to perform mean filtering on the initial photoelectric parameter set of the LED under constant current drive to obtain the standard initial parameter set of the LED under test. The degradation statistical model module M02 is used to construct a multidimensional performance degradation statistical function of the LED under test based on the historical offline aging test data of the LED under test, with the forward voltage, luminous flux and dominant wavelength in the standard initial parameter set as independent variables and the cumulative lighting time of the LED under test as independent variables. The parameter online update module M03 is used to update the current parameters of the multidimensional performance degradation statistical function online by means of the sequential Bayesian method, based on the real-time photoelectric parameter processing records of multiple historical monitoring periods and the fitting residuals of each degradation trajectory function in the multidimensional performance degradation statistical function. The extrapolation lifetime prediction module M04 is used to extrapolate the updated degradation trajectory function to the failure threshold composed of the boundary values of the failure of the performance parameters of the LED under test, and generate the forward voltage lifetime prediction value, luminous flux lifetime prediction value and dominant wavelength lifetime prediction value corresponding to the independent variables in the degradation trajectory function, respectively. The weighted fusion evaluation module M05 is used to perform weighted fusion of the predicted forward voltage lifetime, the predicted luminous flux lifetime, and the predicted dominant wavelength lifetime to generate a comprehensive remaining lifetime prediction result for the LED under test.
[0071] In the several embodiments provided by this invention, it should be understood that the disclosed methods and systems can be implemented in other ways. For example, the system embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and other division methods may be used in actual implementation.
[0072] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment, depending on actual needs.
[0073] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in the form of hardware plus software functional modules.
[0074] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.
[0075] This application embodiment can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence is the theory, method, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.
[0076] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.
Claims
1. A method for predicting the lifetime of light-emitting diodes based on statistical data, characterized in that, The method includes: Step 1: Perform mean filtering on the initial photoelectric parameter set of the LED under constant current drive to obtain the standard initial parameter set of the LED under test; Step 2: Based on the historical offline aging test data of the LED under test, using the forward voltage, luminous flux and dominant wavelength in the standard initial parameter set as independent variables, and the cumulative lighting time of the LED under test as the independent variable, construct a multidimensional performance degradation statistical function for the LED under test. Step 3: During the operation of the LED under test, based on the real-time photoelectric parameter processing records of multiple historical monitoring periods and the fitting residuals of each degradation trajectory function in the multidimensional performance degradation statistical function, the current parameters of the multidimensional performance degradation statistical function are updated online using the sequential Bayesian method. Step 4: Extrapolate the updated degradation trajectory function to the failure threshold formed by the boundary values of the failure of the performance parameters of the LED under test, and generate the forward voltage lifetime prediction value, luminous flux lifetime prediction value and dominant wavelength lifetime prediction value corresponding to the independent variables in the degradation trajectory function respectively. Step 5: Weight and fuse the predicted forward voltage lifetime, the predicted luminous flux lifetime, and the predicted dominant wavelength lifetime to generate a comprehensive remaining lifetime prediction result for the LED under test.
2. The method for predicting the lifetime of light-emitting diodes based on data statistics as described in claim 1, characterized in that, The initial photoelectric parameter set of the LED under constant current drive is mean filtered to obtain the standard initial parameter set of the LED under test, including: At multiple consecutive sampling times, multiple sets of initial photoelectric parameter sampling values of the LED under test are obtained respectively. Each set of initial photoelectric parameter sampling values includes multiple readings that correspond one-to-one with forward voltage, luminous flux and dominant wavelength. The arithmetic mean of multiple readings belonging to the same parameter type in the initial photoelectric parameter sampling values is used as the periodic average forward voltage value, periodic average luminous flux value, and periodic average dominant wavelength value at the sampling time. Remove periodic average parameter groups whose differences exceed a threshold from the periodic average parameter group consisting of the periodic average forward voltage value, the periodic average luminous flux value, and the periodic average dominant wavelength value. The periodic average parameter set after elimination is arithmetically averaged again, and the average forward voltage value, average luminous flux value and average dominant wavelength value obtained after the second arithmetic average are jointly determined as the standard initial parameter set.
3. The method for predicting the lifetime of light-emitting diodes based on data statistics as described in claim 1, characterized in that, The historical offline aging test data based on the LED under test includes: Acquire aging photoelectric datasets of the same batch of LEDs under test at multiple sampling time points recorded in the offline aging test. The aging photoelectric datasets include forward voltage measurement values, luminous flux measurement values, dominant wavelength measurement values, and cumulative lighting duration records corresponding to each sampling time point. For each measurement value belonging to the same sampling time point in the aging photoelectric dataset, a consistency check is performed to remove the measurement value with the largest deviation from the original value. Each verified measurement value is associated with and stored with the corresponding cumulative lighting duration record value to form a standard offline sample set.
4. The method for predicting the lifetime of light-emitting diodes based on data statistics as described in claim 3, characterized in that, The method involves constructing a multidimensional performance degradation statistical function for the LED under test, using the forward voltage, luminous flux, and dominant wavelength from the standard initial parameter set as independent variables, and the cumulative illumination time of the LED under test as the independent variable. This function includes: Based on the cumulative lighting duration records and the corresponding changes in the forward voltage measurement, luminous flux measurement, and dominant wavelength measurement values in the standard offline sample set, the forward voltage change curve, luminous flux change curve, and dominant wavelength change curve are fitted respectively. The positive voltage change curve, the luminous flux change curve, and the dominant wavelength change curve are collectively determined as the initial degradation trajectory function set, wherein each of the change curves adopts a curve form with the goal of minimizing the overall fitting residual with the offline sample data.
5. The method for predicting the lifetime of light-emitting diodes based on data statistics as described in claim 4, characterized in that, The construction of the multidimensional performance degradation statistical function of the LED under test includes: The forward voltage, luminous flux, and dominant wavelength from the standard initial parameter set are respectively used as the initial intercept values of the forward voltage change curve, the luminous flux change curve, and the dominant wavelength change curve at the moment when the cumulative lighting duration is zero. The position parameters of each curve in the initial degenerate trajectory function set are translated and calibrated using the initial intercept value. The calibrated positive voltage variation curve, the luminous flux variation curve, and the dominant wavelength variation curve are combined into the multidimensional performance degradation statistical function.
6. The method for predicting the lifetime of light-emitting diodes based on data statistics as described in claim 1, characterized in that, During the operation of the LED under test, based on the fitting residuals of real-time photoelectric parameter processing records from multiple historical monitoring periods and the degradation trajectory functions in the multidimensional performance degradation statistical function, the current parameters of the multidimensional performance degradation statistical function are updated online using a sequential Bayesian method, including: Extract processed records from multiple consecutive historical monitoring periods prior to the current moment in chronological order. Each processed record contains the positive voltage statistics, luminous flux statistics, and dominant wavelength statistics corresponding to the cumulative lighting duration within the period. Substitute the cumulative lighting duration from each of the processed records into the degradation trajectory functions of the current multidimensional performance degradation statistics function, and calculate the function estimate of each degradation trajectory function at each of the cumulative lighting durations. The formula for calculating the function estimate is as follows: ; In the formula, For the first Each photoelectric parameter during duration The estimated value of the function at that point, For the first The zeroth-order coefficients of a degenerate trajectory function, For the first The first-order coefficients of the degenerate trajectory function For the first The second-order coefficients of a degenerate trajectory function The cumulative lighting duration; The deviation of each function estimate is compared with the statistical value of the same parameter type in the corresponding processed record to obtain the independent residual term of the degenerate trajectory function; For each of the degradation trajectory functions, the independent residual terms belonging to different historical monitoring periods are sorted by absolute value, and a relative influence factor is set for each of the independent residual terms based on the sorting results; Using the processed records and relative influence factors from all the historical monitoring periods, the current parameters of the degradation trajectory function are weighted and revised, and the revised parameters are used as the updated parameters of the degradation trajectory function.
7. The method for predicting the lifetime of light-emitting diodes based on data statistics as described in claim 1, characterized in that, The failure threshold, comprised of the boundary values of the performance parameters of the LED under test, includes: Critical recorded values of various performance parameters corresponding to the failure state of LEDs in the same batch during offline aging tests are extracted. The critical recorded values include a first critical value corresponding to the forward voltage, a second critical value corresponding to the luminous flux, and a third critical value corresponding to the dominant wavelength. The first critical value, the second critical value, and the third critical value are respectively used as the boundary values for the failure of the forward voltage, the luminous flux, and the dominant wavelength, and together they are used as the failure threshold.
8. The method for predicting the lifetime of light-emitting diodes based on data statistics as described in claim 7, characterized in that, The step of generating the predicted positive voltage lifetime, predicted luminous flux lifetime, and predicted dominant wavelength lifetime corresponding to the independent variables in the degradation trajectory function includes: The updated degradation trajectory functions are used as continuous curves of their respective parameters as the cumulative lighting duration. Starting from the current cumulative lighting duration position on each of the continuously changing curves, extend the curve in the direction of increasing duration until the function value of the curve reaches the corresponding failure boundary value, and read the duration value corresponding to the boundary value. The duration at which the degradation trajectory function corresponding to the positive voltage reaches the first critical value is recorded as the predicted positive voltage lifetime value. The duration at which the degradation trajectory function corresponding to the luminous flux reaches the second critical value is recorded as the luminous flux lifetime prediction value. The duration at which the degradation trajectory function corresponding to the dominant wavelength reaches the third critical value is recorded as the dominant wavelength lifetime prediction value.
9. The method for predicting the lifetime of light-emitting diodes based on data statistics as described in claim 1, characterized in that, The step of weightedly fusing the predicted forward voltage lifetime, the predicted luminous flux lifetime, and the predicted dominant wavelength lifetime to generate a comprehensive remaining lifetime prediction result for the LED under test includes: Based on the degradation trajectory function, the absolute values of each independent residual term obtained in each historical monitoring period are taken and accumulated to obtain the residual cumulative value of the degradation trajectory function. The residual cumulative value corresponds to the three residual cumulative values of the positive voltage, the light flux and the dominant wavelength, respectively. The reciprocals of the three cumulative residual values are used as the initial weight base values for each parameter. The three initial weight base values are simultaneously scaled by the same ratio so that the sum of the three scaled values is one. The scaled values are then used as the first weighting coefficient of the forward voltage lifetime prediction value, the second weighting coefficient of the luminous flux lifetime prediction value, and the third weighting coefficient of the dominant wavelength lifetime prediction value, respectively. Multiply the predicted forward voltage lifetime by the first weighting coefficient, the predicted luminous flux lifetime by the second weighting coefficient, and the predicted dominant wavelength lifetime by the third weighting coefficient; The three product results are summed, and the sum is used as the comprehensive remaining lifetime prediction result of the LED under test.
10. A data-based statistical system for predicting the lifetime of light-emitting diodes, characterized in that, The system for implementing the data statistics-based LED lifetime prediction method of claim 1 includes: The initial parameter processing module is used to perform mean filtering on the initial photoelectric parameter set of the LED under constant current drive to obtain the standard initial parameter set of the LED under test. The degradation statistical model module is used to construct a multidimensional performance degradation statistical function of the LED under test based on the historical offline aging test data of the LED under test, with the forward voltage, luminous flux and dominant wavelength in the standard initial parameter set as independent variables and the cumulative lighting time of the LED under test as independent variables. The parameter online update module is used to update the current parameters of the multidimensional performance degradation statistical function online by means of the sequential Bayesian method, based on the real-time photoelectric parameter processing records of multiple historical monitoring periods and the fitting residuals of each degradation trajectory function in the multidimensional performance degradation statistical function. The extrapolation lifetime prediction module is used to extrapolate the updated degradation trajectory function to the failure threshold formed by the boundary values of the failure of the performance parameters of the LED under test, and generate the forward voltage lifetime prediction value, luminous flux lifetime prediction value and dominant wavelength lifetime prediction value corresponding to the independent variables in the degradation trajectory function, respectively. The weighted fusion evaluation module is used to weight and fuse the predicted forward voltage lifetime, the predicted luminous flux lifetime, and the predicted dominant wavelength lifetime to generate a comprehensive remaining lifetime prediction result for the LED under test.
