Method for optimizing aging test time of electronic component

Abstract

The invention provides a method for optimizing the aging test time of an electronic component, and the method comprises the steps: 1, carrying out the collection and preprocessing of test data, and obtaining failure data; step 2, establishing a two-parameter Weibull distribution model based on the failure data collected in the step 1; step 3, performing failure rate curve fitting on the failure rate data obtained in the step 1 by using the Weibull distribution model established in the step 2, and performing aging time optimization; and step 4, analyzing the effectiveness of the optimized aging test scheme according to aging test data of previous batches.

Description

Technical Field

[0001] This invention belongs to the field of electronic component reliability testing technology, and specifically relates to a method for evaluating the optimal aging time of electronic components based on aging test data. Background Technology

[0002] The failure rate of electronic components typically follows a bathtub curve distribution, with the failure rate in the early stages of failure potentially increasing significantly due to design flaws, insufficient material properties, or manufacturing process fluctuations. If components are integrated directly into a complete system without prior reliability screening, it will lead to a substantial increase in the system-level early failure rate, thereby reducing the overall reliability of the product. This will trigger a series of problems related to troubleshooting, elimination, and re-tuning, while also increasing unnecessary early failure repair costs.

[0003] To balance the dual requirements of reliability and cost in commercial aerospace, components should be screened and eliminated for defects or potential weaknesses before leaving the factory or being installed. In engineering, aging tests are commonly used, which apply stress to the tested electronic components for a specified time and at temperatures exceeding their rated operating conditions to induce early defects, thereby effectively eliminating components with defects that could lead to premature failure.

[0004] To standardize testing procedures and ensure screening effectiveness, relevant domestic and international organizations have established systematic aging and screening standards, providing technical support for high-reliability application scenarios. Internationally, the integrated circuit aging test method for high-reliability engineering fields is MIL-STD-883 Method 1015. With the application of commercial off-the-shelf products in aerospace, weaponry, and other fields, relevant international organizations have proposed specific requirements, such as NASA's EEE-INST-002 "EEE Device Selection, Screening, and Quality Assurance Guidelines" and PEM-INST-001 "Plastic Packaged Device Selection, Screening, and Quality Assurance Guidelines." Domestically, relevant standards and specifications have been formulated for aging tests of discrete devices and integrated circuits. For discrete device power aging, GJB 128A Methods 1038, 1039, and 1042 are used as a reference, while integrated circuit aging is performed according to GJB 548B Method 1051.1.

[0005] For components that require aging, the most important thing is to determine the corresponding aging time. Currently, the aging test time for components is mainly set according to industry standards and specifications. For components of the same type, a uniform aging time is usually adopted. The table below shows the correspondence between aging time and temperature in the GJB 548B method.

[0006] Table 1. Correspondence between aging time and temperature in GJB 548B method

[0007]

[0008] In the context of increasingly urgent low-cost requirements in the commercial aerospace field, this standardized processing method does not fully consider the performance parameter differences of individual components, resulting in two prominent problems in the aging process: when the aging time is too short, the aging screening cannot maximize the elimination of early-failure products; when the aging time is too long, the continuous over-stress applied during the test may cause potential damage to normal devices that meet quality standards, and long-term aging tests will significantly increase the production and application costs of the devices. Therefore, how to achieve the following three goals by optimizing the aging test scheme while ensuring the screening effect: (1) effectively shorten the aging time; (2) significantly reduce the test cost; (3) maximize the elimination rate of early-failure products has become a key technical problem that urgently needs to be solved in the field of component reliability engineering.

[0009] According to the bathtub curve theory, the failure rate of electronic components exhibits three stages over time: early failure period (decreasing failure rate), random failure period (stable failure rate), and wear-out failure period (increasing failure rate). The optimal aging time should be set at the inflection point between the early failure period and the random failure period. The theoretical basis for this is that this inflection point is the critical point where early defects are concentratedly exposed. At this point, the marginal benefit of continuing to extend the aging time for improving the failure rate approaches zero. Specifically, this means that (1) mathematically, the first derivative of the failure rate function at this point is zero; (2) in engineering, it means that most of the early failures caused by process defects have been eliminated; (3) beyond this point, the random failure period will begin, and failures will be mainly caused by random factors. Continuing to age the components will actually accelerate the aging of normal devices. Therefore, accurately determining this inflection point is the core of optimizing the aging time.

[0010] There is currently limited research on determining the optimal aging time, and existing methods each have limitations, as detailed below:

[0011] 1. Huang Xiuping et al. constructed a hybrid competitive model of sudden failure and degradation failure, aiming to maximize the mean remaining life (MRL). They used the Weibull distribution to fit the failure rate curve and combined it with numerical calculation to determine the optimal aging time. However, this method is highly dependent on the pre-assumptions of the failure rate distribution, degradation process parameters (such as Wiener process drift and diffusion coefficients), and the proportion of defective products. These parameters are difficult to estimate or verify accurately in actual engineering. The deviation between the assumptions and the actual results will lead to inaccurate calculation results.

[0012] 2. Ning Yongcheng et al. proposed an optimization method for the aging time of silicon integrated circuits by fitting the relationship curve between aging temperature and time in relevant standards. They used the fitting function to calculate the aging time at non-specified temperatures and used activation energy to verify the rationality of the model. However, this method relies too much on the completeness and representativeness of the standard data. If the standard does not cover a specific temperature or device type, the accuracy of the extrapolation results is difficult to guarantee.

[0013] 3. Gao Huizhuang et al., combining the principles of integrated circuit aging tests, lifetime distribution models, and reaction rate theory, proposed an aging time optimization method based on activation energy and characteristic lifetime. Using steady-state lifetime test results as a reference, they calculated the acceleration factor under different temperature stresses using the Arrhenius model and set a target failure rate (e.g., 10). -6 The method determines the inflection point time of early failure period using the Weibull distribution (m = 0.5), but it does not consider the differences in the shape parameters of the Weibull failure distribution of different components, which reduces the reliability of the model in empirical studies.

[0014] Currently, only one relevant invention patent, CN103942605B, was found regarding the optimization of aging time for electronic components. This patent applies to a method for optimizing the aging temperature and time of automotive electronic components. The core steps of this method include: setting multiple aging temperatures and time points (from 4 hours to 72 hours); measuring key parameters (such as output voltage or current); evaluating the quality stability at each temperature using the PPK (Progressive Process Capability Index); selecting a temperature with a PPK level not lower than A as the optimization temperature; and then determining the optimal aging time by analyzing the time it takes for parameters to reach a stable state and the changes at adjacent time points. However, this method uses parameters after 72 hours of aging as a stability benchmark. If this time point itself is not truly stable, it may lead to incorrect judgments at subsequent time points, affecting the reliability of the optimization. Furthermore, this method does not incorporate accelerated aging models or failure physics analysis, has a narrow temperature selection range, and lacks theoretical support for the criteria for determining the stable state, which may lead to deviations in the optimization results in practical applications. Summary of the Invention

[0015] The purpose of this invention is to provide an optimization method for the aging test time of electronic components to reduce the cost of component aging screening. This method is based on the time-dependent differences in the failure characteristics of aerospace components; that is, the early failure period (defect-dominated failure mode) and the random failure period (random failure mode) have different failure mechanisms. The inflection point between the two stages is determined as the optimal aging time. By conducting aging optimization tests, the number of failures of samples within a fixed time interval is statistically analyzed, the failure rate at each time point is calculated, and the curve of the failure rate changing with time is fitted, thereby accurately estimating the time corresponding to the inflection point. This method can significantly shorten the aging test time, thereby improving the cost-effectiveness of reliability assurance, and has reference value for the optimization of aging tests for similar products.

[0016] To address the aforementioned technical problems, this invention provides a method for optimizing the aging test time of electronic components, comprising the following steps:

[0017] Step 1: Collect and preprocess the test data to obtain failure data;

[0018] Step 2: Based on the failure data collected in Step 1, establish a two-parameter Weibull distribution model;

[0019] Step 3: Using the Weibull distribution model established in Step 2, fit the failure rate curve to the failure rate data obtained in Step 1, and optimize the aging time accordingly.

[0020] Step 4: Analyze the effectiveness of the optimized aging test scheme based on past batch aging test data.

[0021] Preferably, step 1 includes:

[0022] A batch of electronic components were tested at equal time intervals Δt during the aging process. The failure time of each component was recorded and statistically analyzed in groups.

[0023] The failure time interval is divided into several groups, and the frequency, relative frequency and cumulative frequency of each group are counted. Based on this, the early failure rate λ1(t) of the aging test at time t is calculated, and the expression is shown in the following formula (1):

[0024]

[0025] Where N is the total number of devices tested in the aging test, n(t) is the number of devices that have failed at this moment, and Δn is the time interval from the next moment to this moment, in hours.

[0026] Preferably, step 2 includes:

[0027] The early failure components to be excluded in the burn-in test usually follow a two-parameter Weibull distribution with a shape parameter less than 1 at the initial stage of their life. The failure density function f(t) and the failure rate function λ(t) are respectively shown in the following formulas (2) and (3):

[0028]

[0029]

[0030] where m is the shape parameter, which is related to the failure mechanism of the component; η is the scale parameter, reflecting the characteristic life.

[0031] Preferably, the step 3 includes:

[0032] Fitting the failure rate λ1(t) corresponding to the component at time t obtained in step 1 to obtain the failure rate λ(t) curve of the early failure of the device (0 < m < 1), and calculating the shape parameter m and the scale parameter η to judge the change trend of the component failure rate over time;

[0033] Taking the first derivative of the failure rate function with respect to time can obtain the change rate of the failure rate over time, that is, the slope change of the descending section of the bathtub curve;

[0034] When there exists a time t1, and the following formula (4) holds for a given discrimination level ε, it is considered that the early failure stage has ended, that is, the burn-in test should end;

[0035] At this time, the descending section of the bathtub curve has passed, and the component enters the accidental failure period at the bottom of the bathtub curve;

[0036]

[0037] where t1 is the optimized burn-in test time, with the unit of hour; t0 is the original burn-in test plan, and for the current spacecraft monolithic integrated circuit burn-in test time, 240 hours is taken.

[0038] Preferably, the step 4 includes:

[0039] For components with stable quality status and controlled processes, the defect rejection rate in the batch burn-in test should be at a stable level; by statistically analyzing the passing rates of the previous batches of burn-in tests of this type of component, the expected value of the passing rate distribution can be obtained.

[0040] Preferably, the step 4 includes:

[0041] If the aging test is carried out according to the optimized aging test plan, and the aging test reaches the same effect as the aging test before optimization at time t1, then the optimized aging test plan can be judged based on the following formula (5) to determine whether the optimized aging test plan has achieved the expected defect elimination effect and whether the quality status of the components has fluctuated significantly.

[0042]

[0043] in, To optimize the aging pass rate in component aging tests, This is the expected estimate of the aging pass rate obtained based on previous batches of unoptimized aging test data. This is the variance estimate of the pass rate of previous batches of aging.

[0044] The beneficial effects of this invention include:

[0045] 1. Parameter estimation is based on actual experimental data and does not require prior assumptions: This method calculates the optimal aging time by fitting the shape parameter (m) and scale parameter (η) of the Weibull distribution to the failure data during the aging process in real time. This avoids the bias caused by relying on empirical assumptions in existing methods and improves the objectivity and applicability of the aging optimization scheme.

[0046] 2. Validating the effect based on historical data to ensure the accuracy of optimization: This method introduces an optimization scheme validity verification mechanism based on statistical hypothesis testing. By comparing and analyzing the aging pass rate of the optimized batch with the pass rate distribution of historical batches, it can objectively judge within the confidence interval whether the optimized scheme can still maintain the original defect elimination ability, thereby scientifically ensuring the risk controllability and result reliability of the aging optimization process. Attached Figure Description

[0047] Figure 1 This is a schematic diagram of the bathtub curves that the failure rates of components follow, provided in an embodiment of the present invention.

[0048] Figure 2 A flowchart illustrating the method for optimizing the aging test time of electronic components provided in this embodiment of the invention;

[0049] Figure 3 This invention provides a representation of the relationship between the number of failures of a certain type of commercial aerospace integrated circuit and the aging optimization test time.

[0050] Figure 4 The failure rate fitting curve of a certain type of commercial aerospace integrated circuit aging optimization test is provided for an embodiment of the present invention. Detailed Implementation

[0051] The technical solutions in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0052] Figure 1 This is a schematic diagram of the bathtub curve for the failure rate of electronic components provided in this embodiment of the invention. According to the bathtub curve theory, the failure rate of electronic components exhibits three stages over time: early failure period (decreasing failure rate), random failure period (stable failure rate), and wear-out failure period (increasing failure rate). The optimal aging time should be set at the inflection point between the early failure period and the random failure period. The theoretical basis for this is that this inflection point is the critical point where early defects are concentratedly exposed. At this point, the marginal benefit of continuing to extend the aging time to improve the failure rate approaches zero. Specifically, this means that (1) mathematically, the first derivative of the failure rate function at this point is zero; (2) in engineering, it means that most of the early failures caused by process defects have been eliminated; (3) beyond this point, the random failure period will begin, and the failure will be mainly caused by random factors. Continuing to age the components will accelerate the aging of normal devices. Therefore, accurately determining this inflection point is the core of optimizing the aging time.

[0053] Figure 2 This is a flowchart illustrating the method for optimizing the aging test time of electronic components according to an embodiment of the present invention. To determine the optimal aging time, the method for optimizing the aging test time of electronic components provided by the present invention includes the following steps:

[0054] 1. Data Acquisition and Preprocessing: A batch of electronic components were tested at equal time intervals Δt during the aging test. The failure time of each component was recorded and grouped for statistical analysis. The failure time interval was divided into several groups, and the frequency, relative frequency, and cumulative frequency of each group were counted. Based on this, the early failure rate λ1(t) at time t of the aging test was calculated, as shown in the following formula (1):

[0055]

[0056] Where N is the total number of devices tested in the aging test, n(t) is the number of devices that have failed at this moment, Δn is the number of devices that will fail at the next moment, and Δt is the time interval between the next moment and this moment, in hours.

[0057] 2. Establishment of Weibull distribution model: Early-failure components that need to be removed during aging tests usually follow a two-parameter Weibull distribution with a shape parameter less than 1 in the early stages of their lifespan. Their failure density function f(t) and failure rate function λ(t) are shown in equations (2) and (3) respectively:

[0058]

[0059] Among them, m is the shape parameter, which is related to the failure mechanism of the component; η is the scale parameter, reflecting the characteristic life.

[0060] 3. Failure rate curve fitting and burn-in time optimization: Fit the failure rate λ1(t) corresponding to the component at time t obtained in step 1 to obtain the failure rate λ(t) curve of the early failure of the device (0 < m < 1), and calculate the shape parameter m and the scale parameter η to judge the change trend of the component failure rate over time. Taking the first derivative of the failure rate function with respect to time can obtain the change rate of the failure rate over time, that is, the slope change of the descending section of the bathtub curve. When there exists a time t1, and the following formula (4) holds for a given discrimination level ε, it is considered that the early failure stage has ended, that is, the burn-in test should end. At this time, the descending section of the bathtub curve has been passed, and the component enters the accidental failure period at the bottom of the bathtub curve.

[0061]

[0062] Among them, t1 is the optimized burn-in test time, in hours; t0 is the original burn-in test plan, and for the current spacecraft monolithic integrated circuit burn-in test time, 240 hours is taken.

[0063] 4. Analysis of the effectiveness of the optimized burn-in test plan: For components with stable quality status and controlled processes, the defect rejection rate in the batch burn-in test should be at a stable level. By statistically analyzing the pass rates of the previous batches of burn-in tests of this type of component, the expected value of the pass rate distribution can be obtained.

[0064] When the burn-in test is carried out according to the optimized burn-in test plan, when the burn-in test reaches time t1, the same effect as the non-optimized burn-in test before optimization can already be achieved. Then, it can be judged based on the following formula (5) whether the optimized burn-in test plan has achieved the expected defect rejection effect and whether there is an obvious fluctuation in the quality status of the components.

[0065]

[0066] Among them, is the burn-in pass rate in the component burn-in optimization test, is the expected estimated value of the burn-in pass rate obtained based on the data of the previous batches of non-optimized burn-in tests, is the variance estimated value of the burn-in pass rate of the previous batches.

[0067] An example of optimizing the burn-in test time of a certain type of domestic monolithic integrated circuit using the method provided by this invention is as follows:

[0068] 1. After the design of this monolithic integrated circuit was finalized, it was supplied in 10 consecutive batches, as shown in Table 2. The device meets all the criteria and requirements for aging optimization. According to the material support needs of commercial aerospace models, it is urgent to complete the aging test of the 11th batch (batch 1901).

[0069] Table 2. Statistics on the Aging Pass Rate of Previous Batches of a Certain Type of Commercial Aerospace Monolithic Integrated Circuit

[0070]

[0071]

[0072] 2. An aging optimization test was conducted on components from production batch 1901. This batch contained a total of 2119 components. The aging optimization test eliminated 103 components that failed early, resulting in a batch pass rate of 95.14%. The testing interval for the aging test components was 24 hours, and the test results are shown in Table 3.

[0073] Table 3. Statistical analysis of the pass rate of the aging and optimization test at equal time intervals for the 11th batch (1901st batch) of integrated circuits.

[0074]

[0075]

[0076] 3. Based on the data from the aging optimization experiment, a system can be established. Figure 3 The graph shows the relationship between the number of failures of this type of device and time during the aging test. It can be seen from the graph that most of the failures of this type of device occurred in the first three tests, while the number of early failures gradually decreased in the fourth and fifth tests.

[0077] Figure 4 To obtain the failure rate curve during the aging optimization test of the 11th batch of devices according to formula (3), through parameter fitting and goodness-of-fit test, it can be seen that the shape parameter m of the Weibull distribution shown in formula (2) is 0.01961, the scale parameter η is 0.8538, and the goodness-of-fit R is... 2 It is 0.8797.

[0078] According to formula (4), the discriminant level ε = 10 is taken. -6 h -2 Then, when the aging optimization test has been conducted up to t1 = 150 hours, the rate of change of the failure rate over time changes from 10... -6 h -2 Transform into 10 -7 h -2 The order of magnitude, meaning the bathtub curve formed by the early failure rate of the device has slowed down significantly, indicating that the bathtub curve has passed its descent phase.

[0079] 4. At this point, consider whether the 150-hour test plan is sufficient to eliminate defects. Under stable and controlled quality conditions, the pass rate of the aging test should not fluctuate significantly according to the existing process technology. As shown in Table 2, with a 95% confidence level, after the aging optimization test, the pass rate should be within the range shown in formula (11), i.e., 95.66% ± 0.62%. The pass rate of the aging optimization test for the 11th batch of products was 95.14%, so it can be considered that the quality of the component is stable, and the optimized aging test plan, i.e., 125℃ and 150 hours, can achieve sufficient aging of the component, thus achieving the goal of eliminating early defects. The optimized aging test plan can be used to guide the quality assurance work of this type of commercial aerospace monolithic integrated circuit product.

[0080]

[0081] Although the present invention has been described in detail through the preferred embodiments above, it should be understood that the above description should not be considered as a limitation of the present invention. Various modifications and substitutions to the present invention will be apparent to those skilled in the art after reading the above description. Therefore, the scope of protection of the present invention should be defined by the appended claims.

Claims

1. A method for optimizing the aging test time of electronic components, characterized in that, It includes the steps: Step 1: Collect and preprocess the test data to obtain failure data; Step 2: Based on the failure data collected in Step 1, establish a two-parameter Weibull distribution model; Step 3: Use the Weibull distribution model established in Step 2 to fit the failure rate curve of the failure rate data obtained in Step 1, and optimize the burn-in time based on this; Step 4: Analyze the effectiveness of the optimized burn-in test plan based on the burn-in test data of past batches.

2. The method according to claim 1, characterized in that, The said Step 1 includes: Test a batch of electronic components at equal time intervals Δt during the burn-in test, record the failure time of each device and conduct grouped statistics; Divide the failure time interval into several groups, count the frequency, frequency and cumulative frequency of each group, and calculate the early failure rate λ1(t) of the burn-in test at time t according to the following formula (1): where N is the total number of test devices in the burn-in test, n(t) is the number of devices that have failed at this moment, and Δn is the time interval from the next moment to this moment, with the unit of hour.

3. The method according to claim 2, characterized in that, The said Step 2 includes: The early failure components to be excluded in the burn-in test usually follow a two-parameter Weibull distribution with a shape parameter less than 1 at the initial stage of their life. Its failure density function f(t) and failure rate function λ(t) are shown in the following formulas (2) and (3) respectively: where m is the shape parameter, which is related to the failure mechanism of the component; η is the scale parameter, reflecting the characteristic life.

4. The method according to claim 3, characterized in that, The said Step 3 includes: Fit the failure rate λ1(t) corresponding to the component at time t obtained in Step 1 to obtain the failure rate λ(t) curve of the early failure of the device (0 < m < 1), and calculate the shape parameter m and the scale parameter η to judge the change trend of the component failure rate over time; Take the first derivative of the failure rate function with respect to time to obtain the change rate of the failure rate over time, that is, the slope change of the descending section of the bathtub curve; When there exists a time t1, and the following formula (4) holds for a given discrimination level ε, it is considered that the early failure stage has ended, that is, the burn-in test should end; At this time, the descending section of the bathtub curve has passed, and the component enters the accidental failure period at the bottom of the bathtub curve; where t1 is the optimized burn-in test time, with the unit of hour; t0 is the original burn-in test plan, and for the current spacecraft monolithic integrated circuit burn-in test time, it is taken as 240 hours.

5. The method according to claim 4, characterized in that, The said Step 4 includes: For components with stable quality status and controlled process, the defect rejection rate in the burn-in test between batches should be at a stable level; by statistically analyzing the qualified rate of the burn-in test of the components of this model in past batches, the expected value of the qualified rate distribution can be obtained.

6. The method according to claim 5, characterized in that, The said Step 4 includes: Execute the burn-in test according to the optimized burn-in test plan. When the burn-in test reaches time t1, the same effect as the burn-in test before optimization can already be achieved. Then, based on the following formula (5), it can be judged whether the optimized burn-in test plan has achieved the expected defect rejection effect and whether there is an obvious fluctuation in the quality status of the components; in, To optimize the aging pass rate in component aging tests, This is the expected estimate of the aging pass rate obtained based on previous batches of unoptimized aging test data. This is the variance estimate of the pass rate of previous batches of aging.

Images