A method for optimizing the design of accelerated tests based on virtual simulation

Abstract

The present invention discloses an accelerated test optimization design method based on virtual simulation, which belongs to the field of accelerated test simulation and includes the following steps: S1, establishing an energy damage model of a product under random factors and multiple stresses; S2, conducting a virtual accelerated test of the model under step-down stress load based on the energy damage model; S3, evaluating the acceleration performance of the product using existing actual accelerated test data. The present invention adopts the above-mentioned accelerated test optimization design method based on virtual simulation, and conducts modeling simulation analysis on key components and main failure mechanisms of aerospace electronic products, and establishes corresponding models for analysis through special simulation software, failure mechanism theory and data-driven methods, which can quickly evaluate the life of the product in a storage environment and under long-term hot standby, realize the comprehensive analysis capability of multi-source data for electronic products, and can greatly save the time and cost of accelerated testing.

Description

Technical Field

[0001] The invention relates to the technical field of accelerated test simulation optimization, and in particular to an accelerated test optimization design method based on virtual simulation. Background Art

[0002] Spacecraft, also known as space vehicle or space vehicle, includes various types of aircraft that operate in space according to the laws of celestial mechanics and perform specific tasks such as exploring, developing, and utilizing space and celestial bodies.

[0003] During long-term storage, spacecraft are constantly subjected to storage environment stresses such as temperature, vibration, and temperature changes; and during use, they are frequently affected by mechanical stress, transportation vibration stress, electrical stress, impact force and other working conditions. Various storage environments and working conditions have a long-term and continuous impact on the performance parameters of electronic products on them through failure mechanisms such as "penetration" and "fatigue", which ultimately manifests as regular changes in one or more performance parameters, and even storage failures. Therefore, the importance and urgency of research on spacecraft storage reliability and life are highlighted.

[0004] However, the existing technology still has the following defects: on the one hand, it is difficult to solve major problems existing in the development stage of new spacecraft, such as the lack of aging diagnosis methods, unclear performance degradation laws, and difficulties in storage reliability and life assessment, and it is difficult to provide a methodological basis for scientific life determination in the development stage; on the other hand, there is a lack of technical basic data for storage life research.

[0005] In response to the above problems, the existing technology generally accelerates the degradation of product performance through accelerated life testing without changing the product degradation trajectory and failure mechanism. However, the application objects of accelerated testing at this stage are mainly components or material-level products, which are of low level and provide limited support for the storage life assessment of the entire equipment. Summary of the invention

[0006] In view of the problem that it is difficult to quickly evaluate the life of electronic products in a storage environment and under long-term hot standby, the present invention provides an accelerated test optimization design method based on virtual simulation. A virtual accelerated test model is adopted to implement weak link fault injection technology based on a constitutive model and analyze product performance characterization, and explore the law of energy damage accumulation. At the same time, in view of the complex test environment of aerospace electronic products, an energy damage model of the product considering random factors and multiple stresses is established, the performance evolution process under multi-dimensional influencing factors is demonstrated, and compared with real experiments, a comprehensive evaluation of accelerated performance and life is achieved.

[0007] To achieve the above object, the present invention provides an accelerated test optimization design method based on virtual simulation, comprising the following steps:

[0008] S1. Establish an energy damage model for products under random factors and multiple stresses;

[0009] S2. Conduct virtual acceleration test of the model under step-down stress load based on energy damage model;

[0010] S3. Use existing actual acceleration test data to evaluate the acceleration performance of the product.

[0011] Preferably, the energy damage model described in step S1 is established based on the Anand constitutive model, and the parameters of the Anand constitutive model are set as follows:

[0012] Initial value of deformation impedance s o =11.1, unit is MPa;

[0013] Activation energy / gas constant Q / R = 8178, in 1 / k;

[0014] Pre-exponential coefficient factor A = 28690, unit is 1 / sec;

[0015] Material constant ξ=4;

[0016] Strain hardening constant m = 0.22;

[0017] The strain rate sensitivity exponent h of stress o =13334, unit is MPa;

[0018] Deformation impedance saturation coefficient The unit is MPa;

[0019] The strain rate sensitivity index of the deformation impedance saturation value is n = 0.011;

[0020] Strain hardening exponent α=1.2.

[0021] Preferably, the multiple stresses described in step S1 include temperature stress and electrical stress.

[0022] Preferably, step S2 specifically includes the following steps:

[0023] S21, setting basic parameters of the temperature cycle benchmark load spectrum, performing an accelerated test, and obtaining accelerated simulation degradation data;

[0024] S22. Use the truncation method to obtain the virtual acceleration test time under the influence of step-down stress: expand the number of samples through the Monte Carlo method, and obtain multiple sample models that conform to the Gaussian distribution in advance for stress loading tests.

[0025] Preferably, in step S21, the position with the shortest failure time is taken as the measured point, and the resistance strain is taken as the degradation amount, and the degradation data of the measured point is tested regularly to obtain the degradation curve of the measured point under the action of temperature stress.

[0026] Preferably, step S22 specifically includes the following steps:

[0027] S221. The loaded stress is graded from high to low: S1, S2, S3, ...S n , and corresponding to the graded stress, the loading time of the stress is graded according to the number of specimen failures;

[0028] S222. Stress loading is performed on a plurality of sample models in sequence from high to low according to the classification of the loaded stress, and step-down stress accelerated degradation trajectories of the plurality of sample models are obtained respectively.

[0029] Preferably, step S3 specifically includes the following steps: for the existing actual accelerated test data, based on the Weibull distribution and the Arrhenius life-stress model, the data between different levels of stress are converted, and the life in the lower level stress loading process is converted to the life in the higher level stress loading process, so as to realize the conversion of data between different levels of stress, thereby expanding the test data.

[0030] Preferably, step S3 specifically includes the following steps:

[0031] S311. Use the Arrhenius formula considering temperature stress as an accelerated model to describe the relationship between stress and life;

[0032] S312. Assume that the life distribution conforms to the distribution of the Weibull model and verify it using the Van-Montfort method. If the verification passes, use the Weibull model to describe the life distribution of the sample model:

[0033] S313, fitting the distribution function of the Weibull model using the actual life data and the simulated life data to obtain the distribution functions under various levels of stress, thereby obtaining the characteristic life based on the actual experiment and the characteristic life based on the simulation;

[0034] S314, by comparing the characteristic life based on the actual experiment with the characteristic life based on the simulation under different levels of stress, it is determined whether the error between the two is within the set error threshold, and if so, it is determined that the accuracy of the virtual simulation test meets the requirements;

[0035] S315, fitting the distribution function of the Weibull model using the life data of the lower stress level and the life data of the higher stress level to obtain the distribution function under each level of stress, thereby obtaining the characteristic life of the lower stress level and the characteristic life of the higher stress level;

[0036] Assume that the characteristic life of the model under stress S1 is T1, the acceleration factor is AF1, the characteristic life of the model under S2 is T2, and T2>T1, the acceleration factor is AF2, and the characteristic life under S2 stress loading is converted to S1 using the following formula:

[0037] T′2=T1+(T2-T1)·AF1 / AF2

[0038] Where, T′2 is the characteristic life after S2 is converted to S1;

[0039] S316. Parameters of the Arrhenius model are estimated using the characteristic life of the lower stress level and the characteristic life of the higher stress level, and the acceleration coefficients between stresses of different levels are calculated based on the Arrhenius model.

[0040] Preferably, step S312 specifically includes the following steps:

[0041] S3121, set the life distribution function F(t f ) is expressed as:

[0042] F(t f )=1-exp(-(t f / η)m)

[0043] Where, t f is the life data, η is the scale parameter or characteristic life, and m is the shape parameter;

[0044] S3122. Use the Van-Montfort method to test whether the life distribution is a Weibull distribution. The test process is as follows:

[0045] Assume that there are n samples under each stress combination for timed truncation life test, and r failure data t1, t2…t r , and t1≤t2≤…≤t r ≤τ, where τ is the sample life under normal stress, assuming that the product life distribution is F(t), and the following test hypothesis is established for this purpose:

[0046] HO: F(t) follows Weibull distribution

[0047] Let X = ln(t), then the Weibull distribution function is transformed into an extreme value distribution, and we get:

[0048]

[0049] Where G is the extreme value distribution function; μ is the mean of sample life distribution; σ is the variance of sample life distribution;

[0050] Then the HO test is transformed into testing whether the distribution of X follows the extreme value distribution;

[0051] make

[0052]

[0053] In the formula, l i is the parameter to be tested; x i+1 =ln(t i+1 ); x i =ln(t i );Z i is the i-th order statistic of the standard extreme value distribution G(x); Z i+1 is the i+1th order statistic of the standard extreme value distribution G(x); Among them μ=lnη, σ=1 / m; E(Z i ) is Z i The mathematical expectation, E(Z i+1 ) is Z i+1 The mathematical expectation, E(Z i ) and E(Z i+1 ) are obtained by looking up the table;

[0054] Van Montfort proved that under the assumption that the conditions are met, l i Asymptotically independent and asymptotically subject to standard exponential distribution, the following statistics are constructed:

[0055]

[0056] In the formula, W is the statistic to be tested; r is the number of failed samples;

[0057] S3123, respectively perform hypothesis testing and life distribution fitting on the actual test data and the virtual test data, and make the following judgments:

[0058] Under the condition that the null hypothesis is true, the statistic W follows an F distribution with degrees of freedom of 2[(r-1) / 2] and 2[r / 2]. For a given significance level α, if W ≤ F 1-α [2[(r-1) / 2], 2[r / 2]] or W ≥ F 1-α [2[(r-1) / 2], 2[r / 2]], then the null hypothesis is not true; otherwise, the null hypothesis is true, and the measured point is determined to be in accordance with the distribution of the Weibull model.

[0059] Preferably, step S316 specifically includes the following steps:

[0060] The Arrhenius model evaluated by using the characteristic life of the lower stress level and the higher stress level is used to obtain the activation energy and constant of the Arrhenius model coefficient, thereby obtaining the corresponding acceleration factor;

[0061] Among them, the Arrhenius model The activation energy E a The material correlation coefficient a is obtained by fitting the characteristic life η of multiple samples, k is the Boltzmann constant, and k = 8.617 × 10 -5 eV; T is the test temperature corresponding to the sample;

[0062] Then the acceleration coefficient T is obtained by the following formula: AF :

[0063]

[0064] Where, L nomal is the life under normal stress; L stress is the lifespan at high temperature; T nomal is the absolute temperature of room temperature; T stress is the absolute temperature at high temperature, E a is the sample activation energy, and k is the Boltzmann constant.

[0065] The present invention has the following beneficial effects:

[0066] By modeling and simulating the key components and main failure mechanisms of aerospace electronic products, and establishing corresponding models for analysis through dedicated simulation software, failure mechanism theory and data-driven methods, we can quickly evaluate the life of products in storage environments and long-term hot standby conditions, and achieve comprehensive analysis capabilities of multi-source data for electronic products, which can greatly save time and costs for accelerated testing.

[0067] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Figure 1 It is a flow chart of an accelerated test optimization design method based on virtual simulation according to the present invention;

[0069] Figure 2 This is a step-down temperature accelerated stress spectrum of a verification example of the present invention;

[0070] Figure 3 A life bar chart of a product subjected to a virtual accelerated test of a step-down stress in a verification example of the present invention;

[0071] Figure 4 It is a stress level diagram of an actual accelerated test of a verification example of the present invention;

[0072] Figure 5 It is a diagram of the resistance strain rate and degradation trajectory fitting model under normal stress (25°C) of the verification example of the present invention;

[0073] Figure 6 This is a diagram of the resistance strain rate and degradation trajectory fitting model at 50°C of a verification example of the present invention;

[0074] Figure 7 This is a diagram of the resistance strain rate and degradation trajectory fitting model at 75°C of a verification example of the present invention;

[0075] Figure 8 This is a diagram of the resistance strain rate and degradation trajectory fitting model at 100°C of a verification example of the present invention;

[0076] Fig. 9 A probability density function distribution diagram of the distribution of the Weibull model adopted in the verification example of the present invention;

[0077] Fig.10 It is a probability density function distribution diagram of 50° C. of the verification example of the present invention;

[0078] Fig.11 It is a probability density function distribution diagram of 75° C. of the verification example of the present invention;

[0079] Fig.12 It is a probability density function distribution diagram of 100° C. of the verification example of the present invention;

[0080] Fig.13 The failure life bar chart of the verification example of the present invention converted under the stress state of 100°C;

[0081] Fig.14 A histogram of the failure life of a PCB board under normal stress in a verification example of the present invention;

[0082] Fig.15 A life distribution model diagram under a normal stress state of a verification example of the present invention;

[0083] Fig.16 It is a probability density function distribution diagram under a normal stress state of a verification example of the present invention;

[0084] Fig.17 This is a life distribution model diagram under 100°C stress of a verification example of the present invention;

[0085] Fig.18 It is a probability density function distribution diagram of the verification example of the present invention under 100°C stress. DETAILED DESCRIPTION

[0086] In order to make the purpose, technical scheme and advantages disclosed in the embodiments of the present invention clearer, the embodiments of the present invention are further described in detail in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the embodiments of the present invention and are not intended to limit the embodiments of the present invention. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application. Examples of the embodiments are shown in the accompanying drawings, where the same or similar reference numerals throughout represent the same or similar elements or elements with the same or similar functions.

[0087] It should be noted that the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units that are not explicitly listed or inherent to these processes, methods, products or devices.

[0088] Like reference numerals and letters denote similar items in the following drawings, and thus, once an item is defined in one drawing, further definition and explanation thereof is not required in subsequent drawings.

[0089] Since understanding the performance degradation law of aerospace electronic products during the development stage is the basis for accurately predicting the remaining life and failure time of products, it is also the key link to ensure the storage reliability level during the life cycle of spacecraft. Figure 1 As shown, an accelerated test optimization design method based on virtual simulation includes the following steps:

[0090] S1. Establish an energy damage model for products under random factors and multiple stresses;

[0091] Preferably, the energy damage model described in step S1 is established based on the Anand constitutive model, and the parameters of the Anand constitutive model are set as follows:

[0092] Initial value of deformation impedance s o =11.1, unit is MPa;

[0093] Activation energy / gas constant Q / R = 8178, in 1 / k;

[0094] Pre-exponential coefficient factor A = 28690, unit is 1 / sec;

[0095] Material constant ξ=4;

[0096] Strain hardening constant m = 0.22;

[0097] The strain rate sensitivity exponent h of stresso =13334, unit is MPa;

[0098] Deformation impedance saturation coefficient The unit is MPa;

[0099] The strain rate sensitivity index of the deformation impedance saturation value is n = 0.011;

[0100] Strain hardening exponent α=1.2.

[0101] Preferably, the multiple stresses described in step S1 include temperature stress and electrical stress.

[0102] S2. Conduct virtual acceleration test of the model under step-down stress load based on energy damage model;

[0103] Preferably, step S2 specifically includes the following steps:

[0104] S21, setting basic parameters of the temperature cycle benchmark load spectrum, performing an accelerated test, and obtaining accelerated simulation degradation data;

[0105] Preferably, in step S21, the position with the shortest failure time is taken as the measured point, and the resistance strain is taken as the degradation amount, and the degradation data of the measured point is tested regularly to obtain the degradation curve of the measured point under the action of temperature stress.

[0106] In this embodiment, according to the stress change of the measured points (weak points, solder joints) during the simulation process, the failure time of the PCB board under various stress states is obtained based on the creep failure theory.

[0107] S22. Use the truncation method to obtain the virtual acceleration test time under the influence of step-down stress: expand the number of samples through the Monte Carlo method, and obtain multiple sample models that conform to the Gaussian distribution in advance for stress loading tests.

[0108] Preferably, step S22 specifically includes the following steps:

[0109] S221. The loaded stress is graded from high to low: S1, S2, S3, ...S n , and corresponding to the graded stress, the loading time of the stress is graded according to the number of specimen failures;

[0110] In this embodiment, 15 sample models are used to carry out stress loading test, and the highest level of stress loading is carried out on the 15 sample models, and the failure threshold of the model acceleration test is set. When five failed samples appear, the acceleration time of the fifth failed sample is determined as the loading time of the first level stress; then the second level stress loading is carried out on the remaining 10 groups of samples, and the acceleration time of the fifth failed sample in this round is also determined as the loading time of the second level stress. On this basis, the last level of stress loading is carried out, and finally the step-down stress accelerated degradation trajectories of the five samples are obtained.

[0111] S222. Stress loading is performed on a plurality of sample models in sequence from high to low according to the classification of the loaded stress, and step-down stress accelerated degradation trajectories of the plurality of sample models are obtained respectively.

[0112] S3. Use existing actual acceleration test data to evaluate the acceleration performance of the product.

[0113] Preferably, step S3 specifically includes the following steps: for the existing actual accelerated test data, based on the Weibull distribution and the Arrhenius life-stress model, the data between different levels of stress are converted, and the life in the lower level stress loading process is converted to the life in the higher level stress loading process, so as to realize the conversion of data between different levels of stress, thereby expanding the test data.

[0114] In this embodiment, according to the step-down acceleration test design, 5 samples failed at each stress level. Except for the 5 samples in the first stage that failed at 100°C, the other samples experienced the combined action of 2 or 3 stresses. Therefore, in order to effectively use the step-down acceleration test data to evaluate the acceleration potential of aerospace electronic products, these data cannot be used directly to evaluate the acceleration potential, but the data needs to be converted to obtain more available test data.

[0115] And because the conversion of accelerated data depends on the life distribution function of the product, which means that the cumulative failure probability of working time under different levels of stress levels is equal. The accelerated stress of the step-down accelerated test is mainly temperature stress, so the Arrhenius model considering temperature stress is used as the accelerated model to describe the relationship between stress and life. In addition, from the data of the accelerated test, it is known that the strain rate is large in the early stage and becomes smaller and smaller later, so the Weibull model is mainly used to describe its life distribution.

[0116] At the same time, since the conversion of accelerated data depends on the principle that the cumulative failure probability is equal under different stress levels, the data between different stress levels are converted based on the Weibull distribution and the ArrheniusArrhenius life stress model. Data conversion depends on the acceleration coefficient between each stress level. In order to calculate the acceleration coefficient, it is necessary to obtain the characteristic life assessment acceleration model of two stress levels. Because all 15 samples have experienced the effect of 100℃ stress, the life in the low stress process is converted to high stress during data conversion to achieve data conversion between different stress levels.

[0117] Therefore, step S3 specifically includes the following steps:

[0118] S311. Use the Arrhenius formula considering temperature stress as an accelerated model to describe the relationship between stress and life;

[0119] S312. Assume that the life distribution conforms to the distribution of the Weibull model and verify it using the Van-Montfort method. If the verification passes, use the Weibull model to describe the life distribution of the sample model:

[0120] Preferably, step S312 specifically includes the following steps:

[0121] S3121, set the life distribution function F(t f ) is expressed as:

[0122] F(t f )=1-exp(-(t f / η)m)

[0123] Where, t f is the life data, η is the scale parameter or characteristic life, and m is the shape parameter;

[0124] S3122. Use the Van-Montfort method to test whether the life distribution is WeibullFENBU. The test process is as follows:

[0125] Assume that there are n samples under each stress combination for timed truncation life test, and r failure data t1, t2…t r , and t1≤t2≤…≤t r ≤τ, where τ is the sample life under normal stress, assuming that the product life distribution is F(t), and the following test hypothesis is established for this purpose:

[0126] HO: F(t) follows Weibull distribution

[0127] Let X = ln(t), then the Weibull distribution function is transformed into an extreme value distribution, and we get:

[0128]

[0129] Where G is the extreme value distribution function; μ is the mean of sample life distribution; σ is the variance of sample life distribution;

[0130] Then the HO test is transformed into testing whether the distribution of X follows the extreme value distribution;

[0131] make

[0132]

[0133] In the formula, l i is the parameter to be tested; x i+1 =ln(t i+1 ); x i =ln(t i );Z i is the i-th order statistic of the standard extreme value distribution G(x); Z i+1 is the i+1th order statistic of the standard extreme value distribution G(x); Among them μ=lnη, σ=1 / m; E(Z i ) is Z i The mathematical expectation, E(Z i+1 ) is Z i+1 The mathematical expectation, E(Z i ) and E(Z i+1 ) are obtained by looking up the table;

[0134] Van Montfort proved that under the assumption that the conditions are met, l i Asymptotically independent and asymptotically subject to standard exponential distribution, the following statistics are constructed:

[0135]

[0136] In the formula, W is the statistic to be tested; r is the number of failed samples;

[0137] S3123, respectively perform hypothesis testing and life distribution fitting on the actual test data and the virtual test data, and make the following judgments:

[0138] Under the condition that the null hypothesis is true, the statistic W follows an F distribution with degrees of freedom of 2[(r-1) / 2] and 2[r / 2]. For a given significance level α, if W ≤ F 1-α [2[(r-1) / 2], 2[r / 2]] or W ≥ F 1-α [2[(r-1) / 2], 2[r / 2]], then the null hypothesis is not true; otherwise, the null hypothesis is true, and the measured point is determined to be in accordance with the distribution of the Weibull model.

[0139] S313, fitting the distribution function of the Weibull model using the actual life data and the simulated life data to obtain the distribution functions under various levels of stress, thereby obtaining the characteristic life based on the actual experiment and the characteristic life based on the simulation;

[0140] S314, by comparing the characteristic life based on the actual experiment with the characteristic life based on the simulation under different levels of stress, it is determined whether the error between the two is within the set error threshold, and if so, it is determined that the accuracy of the virtual simulation test meets the requirements;

[0141] S315, fitting the distribution function of the Weibull model using the life data of the lower stress level and the life data of the higher stress level to obtain the distribution function under each level of stress, thereby obtaining the characteristic life of the lower stress level and the characteristic life of the higher stress level;

[0142] Assume that the characteristic life of the model under stress S1 is T1, the acceleration factor is AF1, the characteristic life of the model under S2 is T2, and T2>T1, the acceleration factor is AF2, and the characteristic life under S2 stress loading is converted to S1 using the following formula:

[0143] T′2=T1+(T2-T1)·AF1 / AF2

[0144] Where, T′2 is the characteristic life after S2 is converted to S1;

[0145] S316. Parameters of the Arrhenius model are estimated using the characteristic life of the lower stress level and the characteristic life of the higher stress level, and the acceleration coefficients between stresses of different levels are calculated based on the Arrhenius model.

[0146] In this embodiment, the data under normal stress and 100° C. high stress level can be used to estimate the parameters of the acceleration model Arrhenius model, so that the acceleration coefficients between different stress levels can be calculated according to the Arrhenius model.

[0147] Preferably, step S316 specifically includes the following steps:

[0148] The Arrhenius model evaluated by using the characteristic life of the lower stress level and the higher stress level is used to obtain the activation energy and constant of the Arrhenius model coefficient, thereby obtaining the corresponding acceleration factor;

[0149] Among them, the Arrhenius model The activation energy E a The material correlation coefficient a is obtained by fitting the characteristic life η of multiple samples, k is the Boltzmann constant, and k = 8.617 × 10-5 eV; T is the test temperature corresponding to the sample;

[0150] In this embodiment, the activation energy E in the Arrhenius model is fitted. a and constant coefficient a, first logarithmize the formula:

[0151] ln(η)=ln(a)-E a / kT

[0152] After logarithmization, for a given characteristic life data η at different temperatures i , we can take the logarithm and perform a linear fit, and the slope of the fit is -E a / kT, the intercept of the fit is ln(a). Where k and T are both known quantities, the activation energy E can be inferred from the slope and intercept. a and a constant coefficient a.

[0153] Then the acceleration coefficient T is obtained by the following formula: AF :

[0154]

[0155] Where, L nomal is the life under normal stress; L stress is the lifespan at high temperature; T nomal is the absolute temperature of room temperature; T stress is the absolute temperature at high temperature, E a is the sample activation energy, and k is the Boltzmann constant.

[0156] Verification Example

[0157] like Figure 2-Figure 4 As shown in the figure, the circuit board substrate of the acceleration model is made of FR-4 composite material, the electronic components are made of copper alloy material, and the solder joints are made of tin-lead alloy material. And because of temperature stress, electrical stress and vibration stress, the weak links of the circuit board are mainly concentrated on the solder joints between the electronic components and the circuit board. In order to realize the fatigue analysis of the solder joints under temperature cycle load, the Anand constitutive model is introduced (the Anand constitutive model describes the inelastic flow process of the material through the deformation impedance s related to factors such as the grain size of the material). The Anand constitutive model is the most commonly used model when studying the relationship between the stress and strain generated by the material. The time-related creep strain and the time-independent plastic strain generated by the material can be described by this model.

[0158] And because the test cycle in the actual test process is relatively long, the resistance strain is selected as the degradation amount for the weak link solder joint, and its degradation data is tested regularly, so that the degradation curve of the product under temperature stress can be obtained, thereby predicting the life of the product.

[0159] In this experiment, a high and low temperature test chamber was used to carry out accelerated test data. During the test, temperature was used as the accelerating stress, that is, tests were carried out on three stress levels of 50℃, 75℃, and 100℃, and three groups of timed truncated constant stress accelerated life tests were carried out, each containing 8 samples. The test samples were placed in the high and low temperature test chamber at a spacing of 5 cm, and the samples were taken out regularly during the test for power-on measurement. The resistance value of the solder joint was obtained by testing the voltage and current at both ends of the solder joint, thereby obtaining the resistance strain of the solder joint. For the creep process, the resistance strain of the solder joint will first undergo a slow change and then an accelerated growth process, resulting in a continuous increase in the damage equivalent of the solder joint. According to fracture mechanics, the resistance strain of the solder joint increases with the increase of the crack and is affected by the crack propagation rate. According to previous studies and test results, when the damage equivalent of the solder joint is 0.2, the solder joint is considered to have failed. Therefore, the failure criterion can be obtained by combining the quantitative relationship between the resistance strain of the solder joint and the amount of solder joint damage caused by creep with the measured test data, thereby obtaining the life of the product.

[0160] At the same time, in order to verify the accelerated potential evaluation, a life test was carried out at room temperature (25°C). The resistance strain results and degradation trajectory results obtained during the test are shown in the following figure. Figure 5-Figure 8 shown.

[0161] Table 1 Actual accelerated test failure schedule (hours)

[0162] 25℃ 50℃ 75℃ 100℃ 1 3779167 435489 46229 15420 2 3998178 479644 65804 16730 3 3773978 387371 59825 12381 4 2881354 339270 88553 14630 5 4384238 429634 58881 16115 6 2484891 383452 58391 11564 7 2509096 346757 49418 14268 8 3392177 291435 63277 19983

[0163] Table 2 Distribution fitting test values ​​based on Van-Montfort of actual test data (extreme value distribution expectation)

[0164]

[0165]

[0166] Table 3 Distribution fitting test values ​​based on Van-Montfort for actual test data (distribution test intermediate calculation values)

[0167]

[0168]

[0169] Table 4 Distribution fit test values ​​based on Van-Montfort for virtual test data (extreme value distribution expectation)

[0170]

[0171] Table 5 Distribution fitting test values ​​based on Van-Montfort for virtual test data (distribution test intermediate calculation values)

[0172]

[0173]

[0174] In this embodiment, the significance level α is set to 0.1, and the F distribution table can be obtained. 0.95 (6, 8) = 0.24, F 0.05 (9, 10) = 3.5, the corresponding Van-Montfort statistics calculation results of each level of stress are shown in Table 6 below. It can be seen that the calculated values ​​of the three stress levels of the statistic W are all in F 0.95 and F 0.05 That is to say, the assumption HO is accepted, that is, the life of the PCB board obeys the Weibull distribution.

[0175] Table 6 Hypothesis test statistics calculation results

[0176]

[0177]

[0178] like Fig. 9 As shown in the figure, through the above test hypothesis, it can be known that the life of the PCB board obeys the Weibull distribution. Then the obtained life data is used to fit the distribution function to obtain the distribution function under each stress level. The distribution functions based on the actual test and the simulation are respectively as follows: Figure 10-12 As shown, the life distribution based on actual experiments is 412621.45, 66573.91, and 16221.7 hours, and the life corresponding to each stress level based on simulation is 430832.62, 71060.56, and 17505.19 hours respectively.

[0179] Table 7 Comparison of characteristic life between actual test and virtual test

[0180]

[0181] By comparing the characteristic life of actual tests and virtual tests at different stress levels, it can be found that the error between the two is within 10%, which also shows that the accuracy of the virtual simulation test can meet the requirements.

[0182] like Fig.13 and Fig.14As shown, in this embodiment, this simulation experiment is the same as the actual accelerated test, using simulated failure data to perform hypothesis testing on the Weibull distribution of the PCB board, and using the Van-Montfort method to perform hypothesis testing on the life distribution.

[0183] Table 8: l statistics based on Van-Montfort

[0184]

[0185]

[0186] At this time, take the significance level α = 0.1, and look up the F distribution table to get F 0.95 (13, 15) = 0.412, F 0.05 (13, 15) = 2.472, F 0.95 (6, 8) = 0.24, F 0.05 The calculation results of the Van-Montfort statistics of stress levels corresponding to (6, 8) = 3.58 are shown in Table 9 below. It can be seen that the calculated values ​​of the two stress levels of the statistic W are both in F 0.05 and F 0.95 That is to say, the assumption HO is accepted, that is, the life of the PCB board obeys the Weibull distribution.

[0187] Table 9 Hypothesis test statistics calculation results

[0188] Stress level(℃) Statistics W 25 0.27 100 1.41

[0189] The results of the above two hypothesis testing methods are consistent. It can be seen that the life of the PCB board obeys the Weibull distribution. The obtained life data is used to fit the distribution function, and the distribution functions under conventional stress levels are obtained as follows: Fig.15 and Fig.16 As shown, the corresponding characteristic life is 62811.93 hours.

[0190] like Fig.17 and Fig.18 As shown, the reduced life is also used to fit the life distribution model, and the characteristic life at the 100°C accelerated stress level is obtained to be 1083.65.

[0191] In this embodiment, the Arrhenius model for evaluating the characteristic life and stress level is used, and the corresponding model coefficient activation energy and constant are 0.52 and 0.00012, respectively, based on which the corresponding acceleration factor can be obtained.

[0192] Table 10 Acceleration coefficient of accelerated temperature stress relative to conventional temperature stress 25℃

[0193] Stress Level 100 75 50 Acceleration factor 57.96 18.19 4.77

[0194] In addition to the acceleration potential of a single stress, the acceleration potential of the entire step-down acceleration spectrum is evaluated from two dimensions: average time and maximum time. The corresponding acceleration factors are shown in the following table.

[0195] Table 11 Acceleration factors based on the entire step-down acceleration spectrum relative to conventional stress

[0196] Maximum time Average time Acceleration factor 4.26 11.52

[0197] It can be seen that with the increase of temperature, the change of the acceleration coefficient is not a linear increase. The higher the temperature, the more obvious the acceleration effect on the product. For example, 100℃ is twice the stress of 50℃, but the acceleration effect is close to 12 times the acceleration coefficient of 50℃. At the same time, for the cumulative effect of multiple stresses, the acceleration effect is located between the acceleration potential of the maximum stress and the minimum stress.

[0198] Therefore, the present invention adopts the above-mentioned accelerated test optimization design method based on virtual simulation, by modeling and simulating the key components and main failure mechanisms of aerospace electronic products, and establishing corresponding models for analysis through special simulation software, failure mechanism theory and data-driven methods, it can quickly evaluate the life of the product in a storage environment and under long-term hot standby, realize the comprehensive analysis capability of multi-source data for electronic products, and can greatly save the time and cost of accelerated testing.

[0199] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that they can still modify or replace the technical solution of the present invention with equivalents, and these modifications or equivalent replacements cannot cause the modified technical solution to deviate from the spirit and scope of the technical solution of the present invention.

Claims

1. An accelerated test optimization design method based on virtual simulation, characterized in that: The following steps are involved: S1. Establish an energy damage model for products under random factors and multiple stresses; S2. Conduct virtual acceleration test of the model under step-down stress load based on energy damage model; S3. Use existing actual accelerated test data to evaluate the accelerated performance of the product; Step S3 specifically includes the following steps: S311. Use the Arrhenius formula considering temperature stress as an accelerated model to describe the relationship between stress and life; S312. Assume that the life distribution conforms to the distribution of the Weibull model and verify it using the Van-Montfort method. If the verification passes, use the Weibull model to describe the life distribution of the sample model: S313, fitting the distribution function of the Weibull model using the actual life data and the simulated life data to obtain the distribution functions under various levels of stress, thereby obtaining the characteristic life based on the actual experiment and the characteristic life based on the simulation; S314, by comparing the characteristic life based on the actual experiment with the characteristic life based on the simulation under different levels of stress, it is determined whether the error between the two is within the set error threshold, and if so, it is determined that the accuracy of the virtual simulation test meets the requirements; S315, fitting the distribution function of the Weibull model using the life data of the lower stress level and the life data of the higher stress level to obtain the distribution function under each level of stress, thereby obtaining the characteristic life of the lower stress level and the characteristic life of the higher stress level; Assume that the characteristic life of the model under stress S1 is T1, the acceleration factor is AF1, the characteristic life of the model under S2 is T2, and T2>T1, the acceleration factor is AF2, and the characteristic life under S2 stress loading is converted to S1 using the following formula: T'2 = T1 + (T2 - T1) AF1 / AF2 Where, T'2 is the characteristic life after S2 is converted to S1; S316. Parameters of the Arrhenius model are estimated using the characteristic life of the lower stress level and the characteristic life of the higher stress level, and the acceleration coefficients between stresses of different levels are calculated based on the Arrhenius model.

2. The method for optimizing the accelerated test based on virtual simulation according to claim 1, characterized in that: The energy damage model described in step S1 is established based on the Anand constitutive model, and the parameters of the Anand constitutive model are set as follows: Initial value of deformation impedance s o =11.1, unit is MPa; Activation energy / gas constant Q / R = 8178, in 1 / k; Pre-exponential coefficient factor A = 28690, unit is 1 / sec; Material constant ξ=4; Strain hardening constant m1 = 0.22; The strain rate sensitivity exponent h of stress o =13334, unit is MPa; Deformation impedance saturation coefficient The unit is MPa; The strain rate sensitivity index of the deformation impedance saturation value is n1 = 0.011; Strain hardening exponent α1=1.

2.

3. The method for optimizing the accelerated test based on virtual simulation according to claim 1, characterized in that: The multiple stresses described in step S1 include temperature stress and electrical stress.

4. The method for optimizing the accelerated test based on virtual simulation according to claim 1, characterized in that: Step S2 specifically includes the following steps: S21, setting basic parameters of the temperature cycle benchmark load spectrum, performing an accelerated test, and obtaining accelerated simulation degradation data; S22. Use the truncation method to obtain the virtual acceleration test time under the influence of step-down stress: expand the number of samples through the Monte Carlo method, and obtain multiple sample models that conform to the Gaussian distribution in advance for stress loading tests.

5. The method for optimizing the accelerated test based on virtual simulation according to claim 4, characterized in that: In step S21, the position with the shortest failure time is taken as the measured point, and the resistance strain is taken as the degradation amount. The degradation data of the measured point is tested regularly, so as to obtain the degradation curve of the measured point under the action of temperature stress.

6. The method for optimizing the accelerated test based on virtual simulation according to claim 4, characterized in that: Step S22 specifically includes the following steps: S221. The loaded stress is graded from high to low: S1, S2, S3, ...S n , and corresponding to the graded stress, the loading time of the stress is graded according to the number of specimen failures; S222. Stress loading is performed on a plurality of sample models in sequence from high to low according to the classification of the loaded stress, and step-down stress accelerated degradation trajectories of the plurality of sample models are obtained respectively.

7. The method for optimizing the accelerated test based on virtual simulation according to claim 1, characterized in that: Step S3 specifically includes the following steps: for the existing actual accelerated test data, based on the Weibull distribution and the Arrhenius life-stress model, the data between different levels of stress are converted, and the life in the lower level stress loading process is converted to the life in the higher level stress loading process, so as to realize the conversion of data between different levels of stress, thereby expanding the test data.

8. The method for optimizing the accelerated test based on virtual simulation according to claim 1, characterized in that: Step S312 specifically includes the following steps: S3121. The life distribution function F(t) of the Weibull model is assumed to be: F(t f )=1-exp(-(t f / η)m) Where, t f is the life data, η is the scale parameter or characteristic life, and m is the shape parameter; S3122. Use the Van-Montfort method to test whether the life distribution is a Weibull distribution. The test process is as follows: Assume that there are n samples for each stress combination to conduct timed truncation life test, and obtain r failure samples t1, t2…t r , and t1≤t2≤…≤t r ≤τ, where τ is the sample life under normal stress, assuming that the product life distribution is F(t), and the following test hypothesis is established for this purpose: HO: F(t) follows Weibull distribution Let X = ln(t), then the Weibull distribution function is transformed into an extreme value distribution, and we get: Where G is the extreme value distribution function; μ is the mean of sample life distribution; σ is the variance of sample life distribution; Then the HO test is transformed into testing whether the distribution of X follows the extreme value distribution; make In the formula, l i is the parameter to be tested; x i+1 =ln(t i+1 ); x i =ln(t i );Z i is the i-th order statistic of the standard extreme value distribution G(x); Z i+1 is the i+1th order statistic of the standard extreme value distribution G(x); Among them μ=lnη, σ=1 / m; E(Z i ) is Z i The mathematical expectation, E(Z i+1 ) is Z i+1 The mathematical expectation, E(Z i ) and E(Z i+1 ) are obtained by looking up the table; Van Montfort proved that under the assumption that the conditions are met, l i Asymptotically independent and asymptotically subject to standard exponential distribution, the following statistics are constructed: In the formula, W is the statistic to be tested; r is the number of failed samples; S3123, respectively perform hypothesis testing and life distribution fitting on the actual test data and the virtual test data, and make the following judgments: Under the condition that the null hypothesis is true, the statistic W follows an F distribution with degrees of freedom of 2[(r-1) / 2] and 2[r / 2]. For a given significance level α, if W ≤ F 1-α [2[(r-1) / 2], 2[r / 2]] or W ≥ F 1-α [2[(r-1) / 2], 2[r / 2]], then the null hypothesis is not true; otherwise, the null hypothesis is true, and the measured point is determined to be in accordance with the distribution of the Weibull model.

9. The method for optimizing the accelerated test based on virtual simulation according to claim 1, characterized in that: Step S316 specifically includes the following steps: The Arrhenius model evaluated by using the characteristic life of the lower stress level and the higher stress level is used to obtain the activation energy and constant of the Arrhenius model coefficient, thereby obtaining the corresponding acceleration factor; Among them, the Arrhenius model The activation energy E a The material correlation coefficient a is obtained by fitting the characteristic life η of multiple samples, k is the Boltzmann constant, and k = 8.617 × 10 -5 eV; T is the test temperature corresponding to the sample; Then the acceleration coefficient T is obtained by the following formula: AF : Where, L nomal is the life under normal stress; L stress is the lifespan at high temperature; T nomal is the absolute temperature of room temperature; T stress is the absolute temperature at high temperature, E a is the sample activation energy, and k is the Boltzmann constant.

Images