Evaluation Method for the Accuracy of Predicting the Performance Degradation of Electronic Devices with Small Sample Sizes
By designing the test temperature and sample number under a small sample size, combining the thermal simulation model and performance degradation prediction results, accelerated degradation tests were carried out, and the problem of evaluating the accuracy of performance degradation prediction of electronic equipment under a small sample size was solved, and the effect of reducing test costs and improving evaluation rigor is achieved.
Patent Information
- Application Number
- CN202411829483.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-12
- Publication Date
- 2025-06-24
- Estimated Expiration
- 2044-12-12
AI Technical Summary
Under small sample sizes, it is difficult to effectively evaluate the expected accuracy of electronic equipment performance degradation, resulting in significant deviations from the actual value of the calculation results, increasing the cost and risk of testing.
By considering the maximum operating temperature of components in electronic equipment and the thermal simulation model design test temperature, combining the expected results of performance degradation and the number of degradation dispersion design test samples, thermal stress accelerated degradation test is carried out to calculate the expected error range of performance degradation.
Under the condition of limited sample size, the accuracy of performance degradation prediction can be effectively evaluated, the cost of testing is reduced, and the evaluation rigor can be improved, breaking through the limitations of traditional methods that require large samples to obtain accurate results.
Smart Images
Figure CN119692117B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to a method for evaluating the reliability of electronic devices, and more particularly to a method for evaluating the accuracy of predicting the performance degradation of electronic devices with a small sample size. Background Art
[0002] Predicting the performance degradation of electronic devices is crucial for ensuring system stability and reducing failure risks, and the accuracy of the prediction directly affects the effectiveness of maintenance strategies. However, with the increasing manufacturing cost of electronic devices and the cost of accelerated degradation tests, it is challenging to conduct accelerated degradation tests with a small sample size and obtain reliable prediction errors because it is difficult to determine the actual degradation mean with a small sample size, resulting in a significant deviation between the calculated prediction accuracy and the actual value. Summary of the Invention
[0003] To solve the problem of difficultly effectively evaluating the accuracy of predicting the performance degradation of electronic devices with a small sample size, the present invention provides a method for evaluating the accuracy of predicting the performance degradation of electronic devices with a small sample size. This method can effectively evaluate the accuracy of predicting performance degradation under the condition of limited sample size, reduce the test cost, and improve the rigor of the evaluation.
[0004] The object of the present invention is achieved by the following technical solutions:
[0005] A method for evaluating the accuracy of predicting the performance degradation of electronic devices with a small sample size, comprising the following steps:
[0006] Step S1: Considering the maximum operating temperature of the components in the electronic device and combining with the thermal simulation model of the electronic device to design the test temperature;
[0007] Step S2: Considering the measurement error to determine the shortest test duration;
[0008] Step S3: Combining the predicted results of performance degradation and considering the degradation dispersion to design the number of test samples;
[0009] Step S4: Conducting a thermal stress accelerated degradation test to obtain the performance degradation data of the electronic device;
[0010] Step S5: Calculating the predicted error interval of the performance degradation of the electronic device.
[0011] Compared with the prior art, the present invention has the following advantages:
[0012] 1. The present invention can effectively evaluate the accuracy of predicting performance degradation under the condition of limited sample size, reduce the test cost, and improve the rigor of the evaluation.
[0013] 2. In actual tests, 5 three-phase inverter electrical appliances were selected for accelerated degradation tests, and reliable performance degradation prediction accuracy results could still be obtained, breaking through the limitation that a large sample size must be used by traditional methods to obtain accurate results. Description of the Drawings
[0014] Figure 1 It is a schematic flow diagram of a method for evaluating the accuracy of performance degradation prediction of electronic devices with a small sample size.
[0015] Figure 2 It is a result diagram of the thermal simulation model in step S12 of the embodiment.
[0016] Figure 3 It is a diagram of the predicted error interval obtained with different sample sizes and significance levels in step S5 of the embodiment. Detailed Embodiment
[0017] The technical solution of the present invention will be further described below in conjunction with the drawings, but it is not limited thereto. Any modification or equivalent replacement of the technical solution of the present invention without departing from the spirit and scope of the technical solution of the present invention shall be covered by the protection scope of the present invention.
[0018] The present invention provides a method for evaluating the accuracy of performance degradation prediction of electronic devices with a small sample size, as Figure 1 shown, the method includes the following steps:
[0019] Step S1: Considering the highest operating temperature of the components in the electronic device, design the test temperature in combination with the thermal simulation model of the electronic device. The specific steps are as follows:
[0020] Step S11: Determine the highest operating temperature T max of all components in the electronic device.
[0021] Step S12: Establish a finite element thermal simulation model of the electronic device, and the thermal simulation error of the model is within ±3°C.
[0022] Step S13: Determine the temperature range [T a , T b , so that the heating temperatures of all components do not exceed their T a at the ambient temperature T max , and there is a situation where the heating temperature exceeds its T b at the ambient temperature T max .
[0023] Step S14: Calculate the value of (T a + T b ) / 2 and round down, denoted as the ambient temperature T c . Perform thermal simulation at T c to obtain the heating temperatures of each component.
[0024] Step S15: If there is no component whose heating temperature exceeds its T max condition, and the difference between T max and the heating temperature of the component is greater than 2°C, then T a = T c and repeat Step S14; if there is no component whose heating exceeds its T max condition, but there is a situation where the difference between T max and the heating temperature of the component is less than or equal to 2°C, then determine T c as the test temperature; if there is a component whose heating temperature exceeds T max condition, but the difference between the heating temperature of the component and T max is less than or equal to 2°C, then subtract 2°C from T c to determine the test temperature; if there is a component whose heating exceeds T max condition, and the difference between the heating temperature of the component and its T max temperature is greater than 2°C, then T b = T c and repeat Step S14.
[0025] Step S2: Considering the measurement error, determine the shortest test duration. The specific steps are as follows:
[0026] Step S21: Set the test temperature T test .
[0027] Step S22: Repeat the test on the parameters of the electronic device of concern at the test temperature T test and record the results. Among them: the number of electronic devices is not less than 1, the number of tests is not less than 30 times, and in each test, steps such as repeated circuit connection and test equipment connection are repeated.
[0028] Step S23: Calculate the mean value of the repeated parameter tests, and subtract the mean value from each result to obtain the measurement error value.
[0029] Step S24: Calculate the standard deviation σ of the measurement error.
[0030] Step S25: Combining the predicted results of the performance degradation of the electronic device, determine the time corresponding to the change in the performance degradation of the parameter of concern (|degraded parameter - initial value of the parameter|) being 3σ, and this time is used as the shortest test duration.
[0031] Step S3: Combining the predicted results of the performance degradation, consider the degradation dispersion to design the number of test samples. The specific steps are as follows:
[0032] Step S31: Set the test duration t.
[0033] Step S32: Combining the predicted results of the performance degradation of the electronic device, determine the predicted mean value, variance, and distribution type of the parameter of concern at time t.
[0034] Step S33: Denote the mean value of Step S32 as D, randomly generate n samples m times based on the result of Step S32, where 2 ≤ n < 20 (small sample size) and m ≥ 10000.
[0035] Step S34: For each randomly generated sample in Step S32, draw n samples with replacement to calculate the mean value, and repeat the drawing k times, where k ≥ 10000, to obtain k mean values and their distribution. Set the significance level α, and calculate the mean value interval according to the following formula:
[0036]
[0037] where θ is the electronic device parameter of interest, represents the estimate of θ, represents the j-th quantile (lower limit), represents the k-th quantile (upper limit). The quantiles j and k are calculated according to the following formula:
[0038]
[0039] where Φ is the standard normal cumulative distribution function, α is the significance level, z (α / 2) is the 100×(α / 2) quantile of the standard normal distribution, n is the number of samples, is the bias correction coefficient of the bias correction and accelerated Bootstrap method, is the acceleration coefficient of the bias correction and accelerated Bootstrap method.
[0040] Step S35: Record the number of times m c that the expected mean value D is within the mean value interval of Step S34 in m times, and calculate the coverage rate:
[0041]
[0042] Step S36: Select the sample number with a high coverage rate as the test sample number.
[0043] Step S4: Conduct a thermal stress accelerated degradation test to obtain the performance degradation data of the electronic device. The specific steps are as follows:
[0044] Conduct an accelerated degradation test on the electronic device at the test temperature determined in Step S1. Each time during the test, wait for the electronic device to reach a stable temperature at the test temperature T test and then conduct the test. Select the test duration to be greater than the shortest test duration determined in Step S2, and conduct the thermal stress accelerated degradation test with the test sample number determined in Step S3 to obtain the performance degradation data of the electronic device.
[0045] Step S5: Calculate the predicted error range of the electronic device performance degradation.
[0046] In this step, the calculation formula for the predicted error Error of the electronic device performance degradation is:
[0047]
[0048] where D t is the predicted mean value of the performance degradation at time t, θ0 is the initial measured value of the electronic device parameters, and θ e is the actual mean value of the performance degradation in the experiment. Furthermore, the calculation formula for the predicted error range of the performance degradation is:
[0049] Error range = [min(Error), max(Error)] (6);
[0050] where min(Error) is the minimum value of the predicted error of the electronic device performance degradation, and max(Error) is the maximum value of the predicted error of the electronic device performance degradation.
[0051] Example:
[0052] In this example, a three-phase inverter is used as the object to evaluate the accuracy of the predicted performance degradation of the electronic device with a small sample size.
[0053] Step S1: Consider the maximum operating temperature of the components inside the electronic device, and design the test temperature in combination with the thermal simulation model of the electronic device. The specific steps are as follows:
[0054] Step S11: Consider the maximum operating temperatures of the power MOSFET, DC-DC module, current sensor chip, operational amplifier chip, and capacitor, and determine the maximum operating temperature T max to be 150°C, 125°C, 125°C, 85°C, and 105°C respectively.
[0055] Step S12: Use commercial finite element thermal simulation software to establish a three-phase inverter model, and conduct actual measurement verification at 25°C. The results are as Figure 2 shown, with an error less than ±3°C.
[0056] Step S13: Set [T a , T b = [25°C, 90°C]. At an ambient temperature of 25°C, the heating temperatures of all components do not exceed their maximum operating temperatures. At an ambient temperature of 90°C, the operational amplifier chip exceeds its operating temperature.
[0057] Step S14: Calculate T c = 57°C, and conduct thermal simulation at an ambient temperature of 57°C.
[0058] Step S15: When Tc When it is 57°C, there is no component whose heating temperature exceeds its T max situation, and the difference between T max and the component heating temperature is greater than 2°C, and T a = T c And repeat step S14. Finally, when T c = 81°C, the heating temperature of the operational amplifier is about 84°C, and the difference from T max is less than or equal to 2°C, and 81°C is determined as the test temperature. At this temperature, the fastest performance degradation rate can be obtained on the premise of ensuring that the three-phase inverter does not have overstress failure, so as to reduce the test cost.
[0059] Step S2: Considering the measurement error, determine the shortest test duration. The specific steps are as follows:
[0060] Step S21: Set the test temperature T test = 25°C.
[0061] Step S22: Repeat the test 30 times for the effective value of the output current of phase A of 5 three-phase inverters.
[0062] Step S23: Calculate the means of 30 times for 5 single units, which are 3.046A, 2.952A, 3.006A, 3.029A, and 2.989A respectively. Subtract the mean from each group of repeated measurement values to obtain the measurement error value.
[0063] Step S24: Calculate the standard deviation of the measurement error σ = 0.016.
[0064] Step S25: Calculate 3σ ≈ 0.05. This means that due to the existence of measurement error, the accelerated degradation test should at least make the effective value of the phase A current change by more than 0.05A, and the evaluation of the predicted accuracy of performance degradation is meaningful. In the predicted result of performance degradation, the duration corresponding to the change of 0.05A in the effective value of the phase A current is 3232 hours, so the accelerated degradation test should be carried out for at least 3232 hours.
[0065] Step S3: Combining the predicted results of performance degradation, consider the degradation dispersion to design the test sample size. The specific steps are as follows:
[0066] Step S31: Set the test duration t = 4000 hours, which is greater than the minimum duration of 3232 hours.
[0067] Step S32: Predict that in 4000 hours, the mean of the effective value of the phase A current is 3.056, the standard deviation is 0.056, and it follows a normal distribution.
[0068] Step S33: Denote D = 3.056, set n = 2 to 10, m = 10000, and generate samples.
[0069] Step S34: Set k = 10000, significance levels α = 0.01, 0.02, 0.03, 0.04, 0.05, and calculate the mean intervals.
[0070] Step S35: The calculated coverage rate results are shown in Table 1.
[0071] Table 1 Mean Interval Coverage Rate
[0072]
[0073] Step S36: When selecting 5 samples, the mean interval coverage rate calculated at a significance level α = 0.01 is 89.34%. This means that due to the dispersion of the effective value of the A-phase current, after randomly selecting 5 three-phase inverters for testing, the probability that the mean interval of the effective value of the A-phase current calculated at a significance level α = 0.01 contains the true mean value of the effective value of the A-phase current is 89.34%. A higher sample size can provide a higher coverage rate, and the coverage rate index can be used to trade off with the test cost to determine an appropriate sample size.
[0074] Step S4: Conduct an accelerated degradation test on the electronic device at 81°C. Each time during the test, wait for the electronic device to reach thermal stability at the test temperature T test = 25°C and then conduct the test. The test duration is selected as 4000 hours, and the sample size is chosen as 10.
[0075] Step S5: The predicted error intervals of the effective value of the A-phase current calculated using different sample sizes and significance levels are as Figure 3 shown. The predicted error of the effective value of the A-phase current of a three-phase inverter with a sample size of 10 and a significance level α = 0.05 for 4000 hours is [20.4%, 107.2%]. Directly calculating the mean value of the effective value of the A-phase current of 10 three-phase inverters in the 4000-hour test is 3.1 A, and the corresponding predicted error is 56%. Comparing with the results obtained by this method, it can be seen that it is much lower than the maximum error of 107.2%. The error results obtained by this method are more rigorous and can provide important support for ensuring the safe and reliable operation of the equipment.
Claims
1. A method for evaluating the accuracy of electronic device performance degradation prediction under small sample size, characterized in that The method comprises the following steps: Step S1, considering the maximum operating temperature of components in the electronic device, and designing the test temperature in combination with the electronic device thermal simulation model; Step S2, considering the measurement error, determining the shortest test duration; Step S3: combining the performance degradation prediction results and considering the degradation dispersion to design the number of test samples. The specific steps are as follows: Step S31: Setting the test duration t ; Step S32: Determine based on the predicted results of electronic equipment performance degradation t The expected mean, variance, and distribution type of the parameter of interest at all times; Step S33: record the mean value of step S32 as D , randomly generated based on the result of step S32 n Samples m times, where 2≤ n <20, m ≥10000; Step S34: For each randomly generated sample in step S33, extract with replacement n Calculate the mean of samples and draw repeatedly k times, among which k ≥10000, get k means and obtain its distribution; Setting the significance level α , calculate the mean interval as follows: ； in, θ is the electronic equipment parameter of interest, Express θ The estimate, Indicates j Quantiles, Indicates k quantile, quantile j and k Calculate as follows: ； ； where Φ is the standard normal cumulative distribution function, α is the significance level, z (α / 2) is the 100×(α / 2)th quantile of the standard normal distribution, is the bias correction coefficient for the bias correction and accelerated Bootstrap method, is the acceleration factor for bias correction and accelerated Bootstrap method; Step S35: Record m The estimated mean D The number of times in the mean interval of step S34 m c , calculate the coverage: ； Step S36, selecting the number of samples with high coverage as the number of test samples; Step S4, conducting a thermal stress accelerated degradation test to obtain performance degradation data of the electronic equipment; Step S5: Calculate the expected error range of the electronic device performance degradation.
2. The method for evaluating the accuracy of electronic device performance degradation prediction under small sample size according to claim 1 is characterized in that The specific steps of step S1 are as follows: Step S11: Determine the maximum operating temperature of all components in the electronic device T max ; Step S12, establishing a finite element thermal simulation model of the electronic equipment, wherein the thermal simulation error of the model is within ±3°C; Step S13, determine the temperature range [ T a , T b ], ambient temperature T a The heating temperature of all components below shall not exceed T max , ambient temperature T b There is a fever temperature exceeding its T max Condition; Step S14, calculate ( T a + T b ) / 2 and round down to the next integer, which is recorded as the ambient temperature T c ,exist T c Perform thermal simulation to obtain the heating temperature of each component; Step S15: If there is no component with a heating temperature exceeding its T max situation, and T max If the temperature difference with the heating temperature of the components is greater than 2°C, T a = T c Repeat step S14; if no component heats up beyond its T max situation, but exists T max If the temperature difference between the heating temperature of the components is less than or equal to 2°C, T c Determine the test temperature; if the temperature of the component exceeds T max However, the heating temperature of the components is T max If the difference is less than or equal to 2℃, T c Subtract 2°C to determine the test temperature; if there is a component heating exceeding T max situation, and the heating temperature of the components is T max If the temperature difference is greater than 2°C, T b = T c And repeat step S14.
3. The method for evaluating the accuracy of electronic device performance degradation prediction under small sample size according to claim 1 is characterized in that The specific steps of step S2 are as follows: Step S21: Setting the test temperature T test ; Step S22: at the test temperature T test Repeat the test of the electronic equipment parameters of concern and record the results; Step S23, calculating the mean of repeated parameter tests, and subtracting the mean from each result to obtain a measurement error value; Step S24: Calculate the standard deviation of the measurement error σ ; Step S25: Combined with the predicted results of electronic equipment performance degradation, determine that the performance degradation change of the parameter of interest is 3 σ The time corresponding to the time is taken as the minimum duration of the test.
4. The method for evaluating the accuracy of electronic device performance degradation prediction under small sample size according to claim 3 is characterized in that In step S22, the number of electronic devices is not less than 1, and the number of tests is not less than 30.
5. The method for evaluating the accuracy of electronic device performance degradation prediction under small sample size according to claim 1 is characterized in that The specific steps of step S4 are as follows: The electronic device is subjected to an accelerated degradation test at the test temperature determined in step S1. During each test, the electronic device is subjected to an accelerated degradation test at the test temperature. T test Next, the test is performed after the heating of the electronic device stabilizes. The test duration is selected to be greater than the shortest test duration determined in step S2. The thermal stress accelerated degradation test is performed with the number of test samples determined in step S3 to obtain the performance degradation data of the electronic device.
6. The method for evaluating the accuracy of electronic device performance degradation prediction under small sample size according to claim 1 is characterized in that In step S5, the calculation formula for the expected error range of electronic device performance degradation is: ； in, is the estimated error of electronic equipment performance degradation, is the minimum value of the expected error of electronic equipment performance degradation, It is the maximum value of the expected error of electronic equipment performance degradation.
7. The method for evaluating the accuracy of electronic device performance degradation prediction under small sample size according to claim 6 is characterized in that The calculation formula of the predicted error of the electronic equipment performance degradation is: ； in, D t for t The expected mean performance degradation at each moment, θ 0 is the measured initial value of the electronic equipment parameter, θ e is the mean value of actual performance degradation in the test.