Method and system for evaluating service life of constant stress timing tail-cutting accelerated life test
阅读说明：本技术 恒定应力定时截尾加速寿命试验寿命评估方法及系统 (Method and system for evaluating service life of constant stress timing tail-cutting accelerated life test ) 是由 李凌 陈达 张博轩 于 2021-09-16 设计创作，主要内容包括：恒定应力定时截尾加速寿命试验寿命评估方法及系统,评估方法包括：根据产品常见失效模式与机理,分析薄弱环节,获取激活能；利用激活能计算加速应力下产品相对于正常应力的加速因子,获得等效试验时间；建立无失效数据统计模型,根据无失效数据总体分布及其参数的先验信息,利用Bayes方法对定时截尾时间处的失效概率进行估计；将无失效数据在定时截尾时间处失效概率的估计值,利用最小二乘法进行曲线拟合,验证其寿命分布,并计算寿命参数；根据寿命参数获取不同可靠度的可靠寿命及平均寿命估计值；根据对数正态分布分布函数和可靠度函数特点,构建可靠寿命置信区间下限。本发明解决了加速寿命试验中的无失效数据寿命评估问题。(The constant stress timing tail-cutting accelerated life test life evaluation method and system comprises the following steps: analyzing weak links according to common failure modes and mechanisms of products to obtain activation energy; calculating an acceleration factor of the product relative to normal stress under the acceleration stress by using the activation energy to obtain equivalent test time; establishing a non-failure data statistical model, and estimating failure probability at the timing tail-cutting time by using a Bayes method according to the total distribution of the non-failure data and prior information of parameters of the non-failure data; carrying out curve fitting on the estimated value of the failure probability of the non-failure data at the timing tail-cutting time by using a least square method, verifying the service life distribution of the non-failure data, and calculating service life parameters; obtaining reliable service life and average service life estimated values with different reliability degrees according to the service life parameters; and constructing a reliable life confidence interval lower limit according to the characteristics of the log-normal distribution function and the reliability function. The invention solves the problem of service life evaluation of non-failure data in an accelerated service life test.)
1. A constant stress timing tail-cutting accelerated life test life evaluation method is characterized by comprising the following steps:
analyzing weak links according to common failure modes and mechanisms of products to obtain activation energy;
calculating an acceleration factor of the product relative to normal stress under the acceleration stress by using the activation energy to obtain equivalent test time;
establishing a non-failure data statistical model, and estimating failure probability at the timing tail-cutting time by using a Bayes method according to the total distribution of the non-failure data and prior information of parameters of the non-failure data;
carrying out curve fitting on the estimated value of the failure probability of the non-failure data at the timing tail-cutting time by using a least square method, verifying the service life distribution of the non-failure data, and calculating service life parameters;
obtaining reliable service life and average service life estimated values with different reliability degrees according to the service life parameters;
and constructing a reliable life confidence interval lower limit according to the characteristics of the log-normal distribution function and the reliability function, and finishing life evaluation according to the reliable life confidence interval lower limit.
2. The method for evaluating the life of the constant stress timing tail-cutting accelerated life test according to claim 1, wherein the activation energy corresponding to different failure mechanisms of the electronic component is shown in the following table:
And analyzing weak links according to the corresponding relation of the table and common failure modes and mechanisms of the products to obtain activation energy.
3. The method for evaluating the life of the constant stress timing truncated acceleration life test according to claim 1, wherein the specific step of calculating the acceleration factor of the product under the acceleration stress relative to the normal stress by using the activation energy to obtain the equivalent test time comprises the following steps:
hypothesis of product miningCarrying out accelerated life test by using constant temperature stress, wherein the constant stress level is k, and the timing tail-cutting time under the ith stress level is tiI is 1,2, … k, and the product does not fail before the tail cutting moment;
the test time at high temperature is equivalent to the test truncation time at the target temperature by using an acceleration factor;
the acceleration factor calculation formula is as follows:
where Ea is the activation energy, K is the Boltzmann constant, TiFor test temperature, T0Is the target temperature;
calculating test truncation time t 'at the ith stress level equivalent target temperature according to the acceleration factor'i＝ti·τi,i＝1,2,…k。
4. The method for assessing the service life of the constant stress timing truncation accelerated life test according to claim 3, wherein the specific steps of establishing a non-failure data statistical model, and estimating the failure probability at the timing truncation time by using a Bayes method according to the total distribution of the non-failure data and the prior information of the parameters of the non-failure data include:
will test the tail-cutting time t'iSorting according to the sequence from small to large to obtain an order statistic t'(1)＜t′(2)…＜t′(k)The number of test specimens at the corresponding stress level is ni；(t′(i),ni) I-1, 2, … k is a set of non-failure data, and S is assumedi＝ni+…+nkAnd i ═ 1,2, … k, and denotes time t'(i)Total number of previous non-stale data;
at tail-biting time t'(i)Obtaining an estimate of the failure probability, Pi＝P(T＜τi) Wherein T is product life; first, P is obtainedkThen successively find Pk-1,Pk-2,…,P2,P1An estimated value of (d);
hypothesis probability of failure PkThe prior distribution is uniform distribution on (0, 1), and if no failure occurs in the test process, the posterior distribution is Beta distribution B (1, n +1), and P is under the square loss principlekThe computational expression of Bayes estimation of (1) is:
if there is further a priori information, consider thatNot exceeding a constant lambdakAt this time, assume PkAt (0, lambda)k) If the oral administration is uniformly distributed, then the formula is used to obtain P according to the square loss principlekBayes estimation of (a):
to obtain PkAfter the Bayes estimation, there is Pk-1Falls in the interval (0, P)k) In (b) obtaining P by the following formulak-1Bayes estimation of (a):
repeating the above processes to calculate the failure probability P of each pointk-2,…,P2,P1。
5. The method for evaluating the life of the constant stress timing tail-cutting accelerated life test according to claim 1, wherein the specific steps of performing curve fitting on the estimated value of the failure probability of the non-failure data at the timing tail-cutting time by using a least square method, verifying the life distribution of the non-failure data, and calculating the life parameter comprise:
taking the commonly used lognormal distribution of electronic components as an example, the density function and the distribution function are as follows:
lnt derived from the above two expressions obeys normal distribution N (mu, sigma)2) The distribution function of the lognormal distribution is expressed as:
converting and linearizing to obtain:
lnt＝μ+σ·Φ-1(F(t))
pair (lnt ') by least squares estimation'(i),Φ-1(Pi) I 1,2, … k, and based on the curveThe parameter μ is obtained when Φ (1) is 0.84, and the σ estimate is as follows:
6. the method as claimed in claim 1, wherein the obtaining of the estimated values of the reliability life and the average life with different reliabilities according to the life parameters comprises:
reliable life estimation values of different reliabilities are obtained according to the following reliability functions:
the average life estimation values of different reliability degrees are obtained according to the following average life calculation expression:
7. the method for assessing the service life of the constant stress timing truncation accelerated life test according to claim 1, wherein the specific step of constructing the lower limit of the confidence interval of the reliable service life according to the characteristics of a lognormal distribution function and a reliability function comprises the following steps:
lnt0.5the lower limit of the confidence interval is calculated as follows:
log life lnt with reliability RRThe lower confidence interval limit is calculated by the following calculation expression:
α is the confidence and n is the total number of test samples.
8. A constant stress timing tail-biting accelerated life test life evaluation system is characterized by comprising:
the activation energy acquisition module is used for analyzing weak links according to common failure modes and mechanisms of products to acquire activation energy;
the equivalent test time acquisition module is used for calculating an acceleration factor of the product relative to normal stress under the acceleration stress by utilizing the activation energy to obtain equivalent test time;
the failure probability estimation module is used for establishing a non-failure data statistical model and estimating the failure probability at the timing truncation time by using a Bayes method according to the total distribution of the non-failure data and the prior information of the parameters of the non-failure data;
the service life parameter calculation module is used for carrying out curve fitting on the estimated value of the failure probability of the non-failure data at the timing tail-cutting time by using a least square method, verifying the service life distribution of the non-failure data and calculating the service life parameters;
the reliable service life and average service life estimation module is used for acquiring reliable service lives and average service life estimation values with different reliability degrees according to the service life parameters;
and the reliable service life confidence interval construction module is used for constructing a lower limit of the reliable service life confidence interval according to the characteristics of the log-normal distribution function and the reliability function and finishing service life evaluation according to the lower limit of the reliable service life confidence interval.
Technical Field
The invention belongs to the field of accelerated life tests of electronic components, and particularly relates to a constant stress timing tail-cutting accelerated life test life evaluation method and system.
Background
In the service life test of electronic components, an accelerated service life test mode is often adopted to improve the accelerated effect and shorten the test time, wherein a constant stress timing tail-ending test is a common accelerated service life test method. With the technological progress, the reliability of products is higher and higher, and the accelerated life test usually faces the condition of no failure and no degradation, namely, the product performance has no failure and no degradation. Research results for analyzing the service life of failure or degradation data are becoming mature, but how to evaluate the reliability index of a product based on the failure-free data under the condition of no failure and no degradation becomes a new research subject. The analysis of the non-failure data has considerable difficulty in theory, the obtained reliability information is quite limited due to the absence of the failure data, but the reliability information does contain the information, and the problem of service life and reliability index evaluation of the non-failure data has important theoretical and practical value.
At present, an evaluation method based on non-failure data is mainly divided into a classic method and a Bayes method, wherein the classic method comprises a sample space ordering method, a likelihood function method and the like, but existing accumulated data is not considered in the methods, and evaluation results are always conservative. The Bayes method can utilize various kinds of prior information, and determine the prior distribution of the product based on the failure probability, so that the accuracy of the evaluation result is improved, and the Bayes method becomes a common evaluation method for non-failure data.
In the Bayes method evaluation process without failure data, prior information of products is utilized, and how to reduce interference of subjective factors is the key point of research when processing and utilizing expert experience or information.
In some existing research documents, such as reliability analysis of product non-failure data (operational research and management volume 12, 10 months 2003) a multi-layer prior method is adopted to analyze non-failure data of a certain type of engine, and multi-layer Bayes estimation of engine reliability is given. The method reduces human factors caused by prior distribution hyper-parameter determination, but the method is complex in calculation and inconvenient for practical application. Bayes reliability analysis of Weibull distribution non-failure data (System engineering theory and practice No. 11) determines prior distribution of failure probability by using a concave-convex method, and Bayes estimation of product reliability indexes is obtained. The method has better robustness, but does not consider the accelerated life test situation and does not relate to the common distribution-lognormal distribution of electronic components. A weighted E-Bayes reliability evaluation method based on non-failure data (system engineering and electronic technology, volume 37, period 1, 2015 1 month) constructs prior distribution of product failure probability according to engineering experience, and provides a weighted least square method combined with the E-Bayes reliability evaluation method and a model to provide E-Bayes estimation of the product failure probability by taking Beta distribution as conjugate prior distribution. The method fully utilizes various prior information of engineering experience, reduces the interference of subjective factors, is convenient for practical application, but requires engineering technicians to know abundant experience of product reliability, and the information is accurate and objective. However, it is often difficult to obtain sufficient empirical information for highly reliable products, with limited knowledge.
Disclosure of Invention
The invention aims to provide a method and a system for evaluating the service life of a constant stress timing tail-cutting accelerated life test, aiming at solving the problem of service life evaluation of non-failure data in the accelerated life test.
In order to achieve the purpose, the invention has the following technical scheme:
in a first aspect, an embodiment of the present invention provides a method for evaluating a life of a constant stress timing truncation acceleration life test, including:
analyzing weak links according to common failure modes and mechanisms of products to obtain activation energy;
calculating an acceleration factor of the product relative to normal stress under the acceleration stress by using the activation energy to obtain equivalent test time;
establishing a non-failure data statistical model, and estimating failure probability at the timing tail-cutting time by using a Bayes method according to the total distribution of the non-failure data and prior information of parameters of the non-failure data;
carrying out curve fitting on the estimated value of the failure probability of the non-failure data at the timing tail-cutting time by using a least square method, verifying the service life distribution of the non-failure data, and calculating service life parameters;
obtaining reliable service life and average service life estimated values with different reliability degrees according to the service life parameters;
and constructing a reliable life confidence interval lower limit according to the characteristics of the log-normal distribution function and the reliability function, and finishing life evaluation according to the reliable life confidence interval lower limit.
As a preferred embodiment of the method for evaluating the lifetime of the present invention, the activation energies corresponding to different failure mechanisms of the electronic component are shown in the following table:
failure mechanism
Activation energy
Gate oxide defect
0.3-0.5
Bulk silicon defect
0.3-0.5
Silicon junction defects
0.6-0.8
Metallization defect
0.5
Growth of intermetallic compounds between gold and aluminum
1.05
Electromigration
0.6-0.9
Corrosion of metal
0.45-0.7
Assembly defect
0.5-0.7
Bonding correlation
1.0
Wafer fabrication, chemical contamination
0.8-1.1
Wafer fabrication, silicon/crystal contamination
0.5-0.6
Dielectric breakdown, area greater than 0.04um thick
0.3
Dielectric breakdown, area less than or equal to 0.04um thick
0.7
Adhesive tack: adhesive-non-adhesive
0.65-1.0
And analyzing weak links according to the corresponding relation of the table and common failure modes and mechanisms of the products to obtain activation energy.
As a preferable scheme of the life evaluation method of the present invention, the specific steps of calculating an acceleration factor of the product under accelerated stress relative to normal stress by using the activation energy to obtain the equivalent test time include:
assuming that the product adopts constant temperature stress to carry out an accelerated life test, the constant stress level is k, and the timing tail-cutting time under the ith stress level is tiI is 1,2, … k, and the product does not fail before the tail cutting moment;
the test time at high temperature is equivalent to the test truncation time at the target temperature by using an acceleration factor;
the acceleration factor calculation formula is as follows:
where Ea is the activation energy, K is the Boltzmann constant, TiFor test temperature, T0Is the target temperature;
calculating test truncation time t 'at the ith stress level equivalent target temperature according to the acceleration factor'i＝ti·τi，i＝1，2，…k。
As a preferred scheme of the life evaluation method of the present invention, the specific steps of establishing a statistical model of non-failure data, and estimating the failure probability at the timing truncation time by using a Bayes method according to the total distribution of the non-failure data and the prior information of the parameters thereof include:
will test the tail-cutting time t'iSorting according to the sequence from small to large to obtain an order statistic t'(1)＜t′(2)…＜t′(k)The number of test specimens at the corresponding stress level is ni；(t′(i)，ni) I-1, 2, … k is a set of non-failure data, and S is assumedi＝ni+…+nkAnd i ═ 1,2, … k, and denotes time t'(i)Total number of previous non-stale data;
at tail-biting time t'(i)Obtaining an estimate of the failure probability, Pi＝P(T＜τi) Wherein T is product life; first, P is obtainedkThen successively find Pk-1,Pk-2，…，P2,P1An estimated value of (d);
hypothesis probability of failure PkThe prior distribution is uniform distribution on (0, 1), and if no failure occurs in the test process, the posterior distribution is Beta distribution B (1, n +1), and P is under the square loss principlekThe computational expression of Bayes estimation of (1) is:
if there is further a priori information, consider thatNot exceeding a constant lambdakAt this time, assume PkAt (0, lambda)k) If the oral administration is uniformly distributed, then the formula is used to obtain P according to the square loss principlekBayes estimation of (a):
to obtain PkAfter the Bayes estimation, there is Pk-1Falls in the interval (0, P)k) In (b) obtaining P by the following formulak-1Bayes estimation of (a):
repeating the above processes to calculate the failure probability P of each pointk-2，…，P2，P1。
As a preferred scheme of the life evaluation method of the present invention, the specific steps of performing curve fitting on the estimated value of the failure probability of the non-failure data at the timing tail-end time by using a least square method, verifying the life distribution of the non-failure data, and calculating the life parameters include:
taking the commonly used lognormal distribution of electronic components as an example, the density function and the distribution function are as follows:
lnt derived from the above two expressions obeys normal distribution N (mu, sigma)2) The distribution function of the lognormal distribution is expressed as:
converting and linearizing to obtain:
lnt＝μ+σ·Φ-1(F(t))
pair (lnt ') by least squares estimation'(i)，Φ-1(Pi) I 1,2, … k, and based on the curveThe parameter μ is obtained when Φ (1) is 0.84, and the σ estimate is as follows:
as a preferred embodiment of the method for estimating lifetime of the present invention, the obtaining the estimated values of the reliable lifetime and the average lifetime with different reliabilities according to the lifetime parameters includes:
reliable life estimation values of different reliabilities are obtained according to the following reliability functions:
the average life estimation values of different reliability degrees are obtained according to the following average life calculation expression:
as a preferred embodiment of the life evaluation method of the present invention, the specific step of constructing the lower limit of the confidence interval of the reliable life according to the characteristics of the lognormal distribution function and the reliability function includes:
lnt0.5the lower limit of the confidence interval is calculated as follows:
log life lnt with reliability RRThe lower confidence interval limit is calculated by the following calculation expression:
α is the confidence and n is the total number of test samples.
In a second aspect, an embodiment of the present invention provides a constant stress timing tail-biting accelerated life test life evaluation system, including:
the activation energy acquisition module is used for analyzing weak links according to common failure modes and mechanisms of products to acquire activation energy;
the equivalent test time acquisition module is used for calculating an acceleration factor of the product relative to normal stress under the acceleration stress by utilizing the activation energy to obtain equivalent test time;
the failure probability estimation module is used for establishing a non-failure data statistical model and estimating the failure probability at the timing truncation time by using a Bayes method according to the total distribution of the non-failure data and the prior information of the parameters of the non-failure data;
the service life parameter calculation module is used for carrying out curve fitting on the estimated value of the failure probability of the non-failure data at the timing tail-cutting time by using a least square method, verifying the service life distribution of the non-failure data and calculating the service life parameters;
the reliable service life and average service life estimation module is used for acquiring reliable service lives and average service life estimation values with different reliability degrees according to the service life parameters;
and the reliable service life confidence interval construction module is used for constructing a lower limit of the reliable service life confidence interval according to the characteristics of the log-normal distribution function and the reliability function and finishing service life evaluation according to the lower limit of the reliable service life confidence interval.
Compared with the prior art, the invention has the following beneficial effects:
aiming at the situation of no failure data in a timing truncation life test, the life evaluation method of the invention fully utilizes the structure and process characteristics of the product, combines the potential failure mechanism and engineering experience of the product, determines various kinds of prior information, establishes a model and a statistical inference method combining the classical mathematical statistics theory and the Bayesian theory, converts the no failure data of the accelerated life test product into failure probability models at different moments under the target temperature, and provides point estimation and interval estimation of the product reliability index. By implementing case analysis, the effectiveness and the applicability of the constant stress timing tail-cutting accelerated life test life evaluation method are verified, the method is convenient for practical engineering application, and engineering technicians are facilitated to solve the problem of no failure data in the accelerated life test process.
Drawings
FIG. 1 is a schematic diagram of a product life data distribution fitting curve according to an embodiment of the present invention.
Detailed Description
The present invention will be described in further detail with reference to the accompanying drawings and examples.
The invention provides a constant stress timing tail-cutting accelerated life test life evaluation method, establishes a high-reliability long-life product reliability evaluation model and a statistical inference method, and solves the problem of non-failure data life evaluation in an accelerated life test.
And (3) with reference to the activation energy or acceleration factor given by relevant standards, documents or existing test results, equating the storage test time at high temperature to the storage test time at normal temperature, and establishing the non-failure data life evaluation method of the constant stress timing truncation acceleration life test by combining the test time and the number of test samples and utilizing a mode of combining the classical mathematical statistics theory and the Bayesian theory.
Specifically, the method for evaluating the service life of the constant stress timing tail-cutting accelerated life test comprises the following steps:
1) and analyzing weak links according to common failure modes and mechanisms of the products to obtain activation energy of the weak links.
When the test piece is not invalid and the parameters are not degraded, the information such as the acceleration factor, the activation energy, the service life distribution and the like of the test piece can not be obtained only by using the test data. The method commonly used in engineering is to give activation energy by referring to relevant standards, literatures or engineering test results in combination with failure mechanisms. The activation energies for different failure mechanisms of electronic components are shown in table 1.
TABLE 1
2) And calculating the acceleration factor of the product relative to the normal stress under the acceleration stress by using the activation energy to obtain the equivalent test time.
Specifically, assuming that the product is subjected to an accelerated life test by using constant temperature stress, the constant stress level is k, and the timing tail-off time is t at the ith (i ═ 1,2, … k) stress leveliAnd the product does not fail before the tail-end cutting moment. And (4) utilizing an acceleration factor to enable the test time at the high temperature to be equivalent to the test truncation time at the target temperature.
The acceleration factor calculation formula is as follows:
where Ea is the activation energy, K is the Boltzmann constant, TiFor test temperature (K), T0Is the target temperature.
Calculating test truncation time t 'at the ith stress level equivalent target temperature according to the acceleration factor'i＝ti·τi,i＝1,2,…k。
3) And establishing a non-failure data statistical model, and estimating the failure probability at the timing truncation time by using a Bayes method according to the overall distribution and the prior information of the parameters thereof.
Concretely, t 'is prepared'iSorting according to the sequence from small to large to obtain an order statistic t'(1)＜t′(2)…＜t′(k)The number of test specimens at the corresponding stress level is ni。(t′(i),ni) Where i is 1,2, … k is a set of non-failure data, and S is seti＝ni+…+nkAnd i ═ 1,2, … k, and denotes time t'(i)Total number of previously non-stale data.
At tail-biting time t'(i)Obtaining an estimate of the failure probability, Pi＝P(T＜τi) Where T is product life. First, calculate PkThen successively find Pk-1,Pk-2,…,P2,P1An estimate of (d).
Hypothesis probability of failure PkThe prior distribution is uniform distribution on (0, 1), and if no failure occurs in the test process, the posterior distribution is Beta distribution B (1, n + 1). Under the principle of square loss, PkBayes of (a) was estimated as:
if there is further a priori information, consider thatNot exceeding a constant lambdakAt this time, P can be assumedkAt (0, lambda)k) The oral administration is uniformly distributed, and then under the principle of square loss, P can be obtainedkBayes of (a) was estimated as:
to obtain PkAfter Bayes estimation, P should bek-1Falls in the interval (0, P)k) In the middle, P is obtained by imitationkMethod of estimating value of Pk-1Bayes of (a) was estimated as:
repeating the above process to calculate the failure probability P of each pointk-2,…,P2,P1。
4) Obtaining the failure probability P of the non-failure dataiAfter the estimation, the least square method is used for carrying out curve fitting, the service life distribution is verified, and the service life parameters are calculated.
Taking the commonly-used lognormal distribution of the electronic components as an example, the density function and the distribution function are as follows:
as can be seen from the expressions (5) and (6), lnt obeys a normal distribution N (. mu.,. sigma.)2) The distribution function of the lognormal distribution can be expressed as:
after conversion and linearization, the following can be obtained:
lnt＝μ+σ·Φ-1(F(t)) (8)
pair (lnt ') by least squares estimation'(i),Φ-1(Pi) I 1,2, … k, and based on the curveThe parameter μ, σ estimate, can be obtained when Φ (1) — 0.84:
5) further, the estimation values of the reliable life and the average life with different reliability can be obtained according to the following formulas:
the reliability function is:
the average life calculation expression is:
6) according to the characteristics of the log-normal distribution function and the reliability function, the lower limit of the reliable service life confidence interval can be constructed.
lnt0.5The lower confidence interval limit of (a) is:
log life lnt with reliability RRThe lower confidence interval limit of (a) is:
α is the confidence and n is the total number of test samples.
The invention also provides a constant stress timing tail-cutting accelerated life test life evaluation system, which comprises:
the activation energy acquisition module is used for analyzing weak links according to common failure modes and mechanisms of products to acquire activation energy;
the equivalent test time acquisition module is used for calculating an acceleration factor of the product relative to normal stress under the acceleration stress by utilizing the activation energy to obtain equivalent test time;
the failure probability estimation module is used for establishing a non-failure data statistical model and estimating the failure probability at the timing truncation time by using a Bayes method according to the total distribution of the non-failure data and the prior information of the parameters of the non-failure data;
the service life parameter calculation module is used for carrying out curve fitting on the estimated value of the failure probability of the non-failure data at the timing tail-cutting time by using a least square method, verifying the service life distribution of the non-failure data and calculating the service life parameters;
the reliable service life and average service life estimation module is used for acquiring reliable service lives and average service life estimation values with different reliability degrees according to the service life parameters;
and the reliable service life confidence interval construction module is used for constructing a lower limit of the reliable service life confidence interval according to the characteristics of the log-normal distribution function and the reliability function and finishing service life evaluation according to the lower limit of the reliable service life confidence interval.
Examples
A constant stress accelerated life test is carried out by using 29 certain type of phase sensitive rectifiers, a timing tail-cutting mode is adopted in the test, and the tail-cutting time is 6168 hours. Four groups of temperature stress grouping tests are carried out on the samples at 130 ℃, 145 ℃, 160 ℃ and 175 ℃, and the number of test pieces at the four groups of temperatures is respectively 5, 5 and 14.
In the test process, samples are tested at intervals, test data obtained under 4 groups of temperature stress are analyzed, the result samples do not lose effectiveness, the average change amplitude of main parameters is small, and the parameters have no degeneration tendency.
The method of the invention is utilized to evaluate the service life of the test piece.
1) Activation energy selection
The packaging form of the phase sensitive rectifier is metal cover plate ceramic double-row packaging, a chip is bonded on a substrate through conductive glue, a bonding wire is a gold wire, an inner bonding point is gold-aluminum bonding, an outer bonding point is gold-gold bonding, and the chip process is a bipolar process.
According to the process structure characteristic analysis of the product, the common failure mechanism is bulk silicon defect and gold-aluminum bonding, and engineering experience shows that the activation energy evaluation value of the product with the same packaging, structure and material is 0.37 eV.
2) Calculating target temperature acceleration factor and equivalent tail-biting time by using activation energy
The target temperature 25 ℃ acceleration factor is shown in table 2, and the target temperature 25 ℃ equivalent tail-off time is shown in table 3.
TABLE 2
Test temperature
130℃
145℃
160℃
175℃
Acceleration factor
42.69
62.58
89.33
124.51
TABLE 3
Tail-off time at high temperature
6168h
6168h
6168h
6168h
Equivalent tail time at 25 DEG C
263365h
386019h
551004h
767982h
3) Sequencing the equivalent tail-cutting time from small to large to obtain order statistic t'(1)＜t′(2)…＜t′(k)The number of test specimens at the corresponding stress level is ni. Further obtaining S by the method of the inventioni＝ni+…+nkAnd i ═ 1,2, … k, and denotes time t'(i)Total number of previously non-stale data. The failure probability estimates are shown in table 4.
TABLE 4
25 ℃ equivalent tail-cutting time t'(i)
263365h
386019h
551004h
767982h
Number of samples ni
5 are
5 are
5 are
14 are
Si
29 are all
24 are
19 are only
14 are
Pi
0.005321
0.011256
0.025034
0.062500
4) To life data (lnt'(i),Φ-1(Pi) And i is 1,2, … k, and distribution fitting and testing are carried out, so that the product life follows lognormal distribution, the fitting degree is high, and the correlation coefficient is 0.982, as shown in fig. 1.
5) Further obtaining a life parameter estimation value:
the evaluation results of the reliability life and the lower limit of the confidence interval are shown in tables 5 and 6, respectively.
TABLE 5
Degree of reliability
0.5
0.8
0.9
0.99
Life at 25 ℃ (year)
471.11
194.12
122.13
40.63
TABLE 6
Confidence level
0.95
0.9
0.8
0.7
One-sided quantile of t-distribution
2.05
1.70
1.31
1.06
Allowable error
0.40
0.33
0.26
0.21
Reliability 0.9 Life confidence lower bound (year)
81.80
87.56
94.47
99.33
The reliability of the phase-sensitive rectifier is 0.9 and the service life t is evaluated0.9122.13 with a confidence of 0.95 t0.9The lower confidence interval limit was 87.56 years.
The above-mentioned embodiments are only preferred embodiments of the present invention, and are not intended to limit the technical solution of the present invention, and it should be understood by those skilled in the art that the technical solution can be modified and replaced by a plurality of simple modifications and replacements without departing from the spirit and principle of the present invention, and the modifications and replacements also fall into the protection scope covered by the claims.
- 上一篇：一种医用注射器针头装配设备
- 下一篇：一种支付机构卡面图片数字版权审核的方法和装置