Method and device for predicting service life of accelerated test product and electronic equipment

Abstract

The invention discloses a life prediction method and device for an accelerated test product and electronic equipment, and the method carries out typical failure mechanism analysis based on fault data, so as to determine a key device and sensitive stress of a target product. And carrying out clustering analysis on the fault data related to the key device, and determining the failure contribution rate of each sensitive stress to the target product. And according to the failure contribution rate corresponding to each sensitive stress, carrying out a multi-stress accelerated life test of the target product. And training the neural network model by adopting test data of the multi-stress accelerated life test. The accelerated test life of a target product under the target stress condition is predicted by using the trained neural network model, life prediction of the accelerated test product is realized, the neural network model is trained by using a multi-stress accelerated life test based on fault physics, the influence of the coupling effect between environmental stresses on the product is emphasized, and the service life prediction of the accelerated test product is realized. Therefore, the model better conforms to the actual application process of the product, and the life prediction precision is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of equipment fault prediction, and particularly to a method, device and electronic device for predicting the life of an accelerated test product. Background Art

[0002] With the development of technology and the progress of society, people's requirements for the high reliability and long life of industrial products are getting higher and higher.

[0003] Traditional reliability acceleration tests need to determine the mathematical relationship between product life and stress, that is, the acceleration life model, and then use the model to predict the life of the product.

[0004] At present, researchers have developed various physical acceleration models based on the physical process of product failure, but these models often ignore the coupling effect between stresses, resulting in defects in the objectivity and correctness of the models, thus leading to low accuracy of life prediction results. Summary of the Invention

[0005] The present invention provides a method, device and electronic device for predicting the life of an accelerated test product to improve the accuracy of life prediction.

[0006] According to an aspect of the present invention, there is provided a method for predicting the life of an accelerated test product, including:

[0007] Based on the fault data of the target product, conduct a typical failure mechanism analysis to determine the key components and sensitive stresses of the target product;

[0008] Perform cluster analysis on the fault data related to the key components to determine the failure contribution rate of each sensitive stress to the target product;

[0009] According to the failure contribution rate corresponding to each sensitive stress, conduct a multi-stress accelerated life test on the target product;

[0010] Use the test data of the multi-stress accelerated life test to train a neural network model;

[0011] Use the trained neural network model to predict the accelerated test life of the target product under target stress conditions.

[0012] Optionally, the fault data of the target product includes environmental stress conditions, product operation time and fault conditions;

[0013] The conducting a typical failure mechanism analysis based on the fault data of the target product to determine the key components and sensitive stresses of the target product includes:

[0014] Based on the failure situation, determine the key components of the target product;

[0015] Based on the product operation time and the environmental stress conditions related to the key components, and in combination with the failure time calculation formula of typical failure mechanisms, determine the relevant failure mechanisms;

[0016] Based on the relevant stresses of the relevant failure mechanisms, determine the sensitive stresses for the target product.

[0017] Optionally, performing a clustering analysis on the failure data related to the key components to determine the failure contribution rate of each sensitive stress to the target product, including:

[0018] Using a clustering analysis algorithm, perform a clustering analysis on the failure data related to the key components according to the sensitive stresses;

[0019] Based on the results of the clustering analysis, determine the failure contribution rate of each sensitive stress to the target product.

[0020] Optionally, the clustering analysis algorithm includes a quick clustering analysis algorithm.

[0021] Optionally, based on the failure contribution rate corresponding to each sensitive stress, conduct a multi-stress accelerated life test on the target product, including:

[0022] Based on the number of sensitive stresses, determine the number of groups of the multi-stress accelerated life test;

[0023] Based on the experimental data fitting distribution model selected for the multi-stress accelerated life test, determine the number of samples in each group of the multi-stress accelerated life test;

[0024] Use a preset high value to conduct an accelerated life test on each group of samples to obtain the life data of the target product under different stress conditions, where the accelerated life test includes a failure test and a degradation test.

[0025] Optionally, the experimental data fitting distribution model includes a Weibull distribution model, and the number of samples in each group corresponding to the Weibull distribution model in the multi-stress accelerated life test is greater than 3.

[0026] Optionally, the determining the number of groups of the multi-stress accelerated life test based on the number of sensitive stresses includes:

[0027] Solve the number combination value of the sensitive stresses;

[0028] Determine the number of groups of the multi-stress accelerated life test to be greater than or equal to the number combination value.

[0029] Optionally, the neural network model includes a backpropagation neural network model.

[0030] According to another aspect of the present invention, there is provided a device for predicting the life of an accelerated test product, characterized by comprising: a failure mechanism analysis module, a clustering analysis module, an accelerated life test module, and a model training module;

[0031] The failure mechanism analysis module is used to conduct a typical failure mechanism analysis based on the failure data of the target product to determine the key components and sensitive stresses of the target product;

[0032] The clustering analysis module is used to perform clustering analysis on the failure data related to the key components to determine the failure contribution rate of each sensitive stress to the target product;

[0033] The accelerated life test module is used to conduct a multi-stress accelerated life test of the target product according to the failure contribution rate corresponding to each sensitive stress;

[0034] The model training module is used to train a neural network model by using the test data of the multi-stress accelerated life test;

[0035] The prediction module is used to predict the accelerated test life of the target product under target stress conditions by using the trained neural network model.

[0036] According to another aspect of the present invention, there is provided an electronic device, the electronic device comprising:

[0037] At least one processor; and

[0038] A memory communicatively connected to the at least one processor; wherein,

[0039] The memory stores a computer program executable by the at least one processor, and when the computer program is executed by the at least one processor, the at least one processor is enabled to execute the method for predicting the life of an accelerated test product according to any embodiment of the present invention.

[0040] The method, device and electronic device for predicting the life of an accelerated test product provided by the present invention carry out typical failure mechanism analysis based on the failure data of the target product to determine the key components and sensitive stresses of the target product. Cluster analysis is performed on the failure data related to the key components to determine the failure contribution rate of each sensitive stress to the target product. According to the failure contribution rate corresponding to each sensitive stress, a multi-stress accelerated life test of the target product is carried out. The experimental data of the multi-stress accelerated life test are used to train a neural network model. The trained neural network model is used to predict the accelerated test life of the target product under the target stress conditions, realizing the prediction of the life of the accelerated test product. This method is based on physics of failure, uses the multi-stress accelerated life test to train the neural network model, and pays attention to the influence of the coupling effect between environmental stresses on the product, making the model more in line with the actual application process of the product and greatly improving the accuracy of life prediction.

[0041] It should be understood that the content described in this part is not intended to identify the key or important features of the embodiments of the present invention, nor is it used to limit the scope of the present invention. Other features of the present invention will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention, and those of ordinary skill in the art can also obtain other drawings based on these drawings without creative efforts.

[0043] Figure 1 It is a schematic flowchart of a method for predicting the life of an accelerated test product provided by an embodiment of the present invention;

[0044] Figure 2 It is a schematic flowchart of another method for predicting the life of an accelerated test product provided by an embodiment of the present invention;

[0045] Figure 3 It is a training framework diagram of a neural network model provided by an embodiment of the present invention;

[0046] Figure 4 It is a training flowchart of a GA-BP neural network model provided by an embodiment of the present invention;

[0047] Figure 5 It is a comparison diagram of the life prediction value obtained by a method for predicting the life of an accelerated test product provided by an embodiment of the present invention and the measured value of the life of the target product obtained by actual test;

[0048] Figure 6Schematic diagram of the composition of the life prediction device for the accelerated test product provided by the embodiment of the present invention;

[0049] Figure 7 Schematic diagram of the structure of an electronic device that can be used to implement the embodiments of the present invention is shown. Detailed implementation manners

[0050] In order to enable those skilled in the art to better understand the solution of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without making creative efforts shall fall within the protection scope of the present invention.

[0051] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and do not have to be used to describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device that includes a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0052] Based on the content mentioned in the background art, in the prior art, when carrying out reliability acceleration tests, it is often necessary to determine the mathematical relationship between product life and stress, that is, the acceleration life model, and then use the model to predict the life of the product. At present, researchers have developed various physical acceleration models based on the physical process of product failure, such as the Arrhenius temperature model; empirical acceleration models have been proposed based on long-term observation and summary of product performance, such as the well-known inverse power law model; based on the common action mechanism of multiple stresses on the product, multi-stress acceleration models such as the temperature and electrical stress acceleration model (Fallou) and the temperature and humidity stress life model (Hallberg-Peck) have been successively developed. After research by the inventor, it is found that these acceleration models consider that each stress is independent of each other and does not interact, ignoring the coupling effect between stresses, which makes the objectivity and correctness of the model flawed, thus affecting the accuracy of the life prediction result.

[0053] In view of the problems mentioned in these prior arts, the present invention will start from the analysis of the failure mechanism of typical electronic products to determine the key important components and sensitive stress types of the products. Then, cluster analysis is performed on the failure data of the products to obtain the contribution rate of each sensitive stress to the total failure. Based on the results of the cluster analysis, an accelerated life test is carried out on the target product to obtain the accelerated test life data of the product under multi-stress conditions. Furthermore, a neural network is used to learn the environmental stress and life data, and a multi-stress neural network acceleration model for electronic products is established. The multi-stress neural network acceleration model is used to predict the accelerated test life of the product, solving the problem of evaluating the reliability index of the product and improving the accuracy of the life prediction results. The following uses embodiments to introduce the solution of the present invention.

[0054] An embodiment of the present invention provides a method for predicting the life of an accelerated test product. Figure 1 It is a schematic flow chart of a method for predicting the life of an accelerated test product provided by an embodiment of the present invention. Referring to Figure 1 , the method for predicting the life of an accelerated test product includes:

[0055] S101. Based on the failure data of the target product, conduct a typical failure mechanism analysis to determine the key components and sensitive stresses of the target product.

[0056] Specifically, the target product is the research object of the accelerated life test and can be any industrial product as needed. Exemplarily, the target product can be electronic products such as integrated circuits, semiconductor devices, electronic components, or household appliances, or mechanical products such as mechanical parts or mechanical equipment, or new energy products such as solar cells or lithium batteries. The failure data refers to the state data related to product failures and product life during the life stages of the target product, such as the development process, qualification test, or field use. Exemplarily, the failure data can include environmental stress conditions, product operation time, and failure situations. The failure data can include multiple groups of data from multiple target products of the same model. Exemplarily, when the target product is an x-model mobile phone, the failure data can include the environmental stresses suffered by 1000 x-model mobile phone samples during the test or actual use, the obtained mobile phone life duration in the test, and the failed components and types.

[0057] Typical failure mechanism analysis is a method used to study the reasons and processes for products, systems, or materials to lose their original functions or performances during use. In typical failure mechanism analysis, based on failure data, it can be determined whether the failure process of the target product conforms to each typical failure mechanism. When the judgment result is in line, the environmental stress related to the typical failure mechanism can be determined as the sensitive stress of the target product. Exemplarily, typical failure mechanisms can include electromigration failure mechanism, corrosion failure mechanism, fatigue failure mechanism, wear failure mechanism, dielectric breakdown failure mechanism, and polymer material aging failure mechanism. If it can be determined according to the failure data that the failure process of the target product conforms to the electromigration failure mechanism and the corrosion failure mechanism, then the sensitive stress corresponding to the electromigration failure mechanism and the sensitive stress corresponding to the corrosion failure mechanism can both be determined as the sensitive stress of the target product.

[0058] In typical failure mechanism analysis, based on failure data, it can also be determined which components or locations have the highest frequency of causing the failure of the target product, and the devices corresponding to these components or locations are determined as the key devices of the target product. Exemplarily, if the failure data indicates that the components with the highest frequency of causing the failure of the target product are the instrument housing exposed to the atmospheric environment, the sensor in direct contact with the working medium, and the circuit system sealed inside the instrument, then the instrument housing, the sensor, and the circuit system can be determined as the key devices of the target product.

[0059] S102. Conduct cluster analysis on the failure data related to the key devices to determine the failure contribution rate of each sensitive stress to the target product.

[0060] Specifically, the failure data related to the key devices refers to data such as product life, environmental stress conditions related to the key devices of the target product, and failure situations. Cluster analysis refers to the analysis process of grouping a set of failure data into multiple classes or clusters composed of similar objects. Among them, the cluster analysis in this application is carried out according to environmental characteristics (i.e., sensitive stress), and one data point in the cluster analysis corresponds to the failure data of a target product sample. Cluster analysis can be implemented using a clustering algorithm. Exemplarily, the clustering algorithm can include the quick clustering analysis algorithm. According to the clustering result, the number of data points corresponding to each sensitive stress of the target product can be determined. Furthermore, by taking the ratio of the number of data points corresponding to each sensitive stress to the total number of data points, the failure contribution rate of each sensitive stress to the target product can be determined.

[0061] S103. Conduct multi-stress accelerated life tests on the target product according to the failure contribution rate corresponding to each sensitive stress.

[0062] Specifically, a multi-stress accelerated life test refers to an accelerated life test on a target product under the coupling condition of multiple stresses. According to the failure contribution degree of the sensitive stresses determined in step S102, several sensitive stresses are selected as test conditions. Exemplarily, several sensitive stresses with the highest failure contribution degree can be selected as test conditions. The multi-stress accelerated life test can set multiple groups according to different stress combinations. Exemplarily, when selecting temperature, humidity, and current as the test conditions for the three sensitive stresses, three groups of tests, namely, a temperature + humidity variable group, a temperature + current variable group, and a humidity + current variable group, can be set by combining the three sensitive stresses in pairs. The multi-stress accelerated life test can, based on limited failure data, obtain more failure data of the target product under various different conditions through test and test data fitting and derivation.

[0063] S104. Use the test data of the multi-stress accelerated life test to train the neural network model.

[0064] Specifically, the test data of the multi-stress accelerated life test in step S103 can include both the real test data measured and recorded during the process of making the product fail or degenerate in the multi-stress accelerated life test, and the pseudo-test data determined according to the mathematical relationship derived by establishing the relationship between multi-stresses and product life using statistical analysis methods and life models. The neural network model is trained using a large amount of real test data and pseudo-test data. The input layer of the neural network is the environmental stress and reliability parameters of the target product, and the output layer is the accelerated test life of the target product.

[0065] S105. Use the trained neural network model to predict the accelerated test life of the target product under the target stress condition.

[0066] Specifically, input the target stress condition into the trained neural network model, and use the neural network model to predict the accelerated test life of the target product under the target stress condition.

[0067] The life prediction method for accelerated test products provided in this embodiment is based on the failure data of the target product, conducts a typical failure mechanism analysis to determine the key components and sensitive stresses of the target product. Cluster analysis is performed on the failure data related to the key components to determine the failure contribution rate of each sensitive stress to the target product. According to the failure contribution rate corresponding to each sensitive stress, a multi-stress accelerated life test of the target product is carried out. The experimental data of the multi-stress accelerated life test are used to train the neural network model. The trained neural network model is used to predict the accelerated test life of the target product under the target stress conditions, realizing the life prediction of the accelerated test product. This method is based on physics of failure, uses the multi-stress accelerated life test to train the neural network model, and pays attention to the influence of the coupling effect between environmental stresses on the product, making the model more in line with the actual application process of the product and greatly improving the accuracy of life prediction.

[0068] Figure 2 FIG. is a schematic flow chart of another life prediction method for accelerated test products provided in an embodiment of the present invention. On the basis of the foregoing embodiment, with reference to Figure 2 , the life prediction method for accelerated test products includes:

[0069] S201. Determine the key components of the target product according to the failure situation.

[0070] Specifically, the failure data of the target product includes environmental stress conditions, product operation time, and failure situations. Among them, the failure situation may include the failure location, and the failure location can indicate the component in the target product where the failure occurs. According to the failure location in the failure data of each target product, the high-frequency damaged components of the target product can be determined, and the high-frequency damaged components can be determined as the key components of the target product. Exemplarily, the failure data of 1000 mobile phones of model x shows that: 702 mobile phones have motherboard failures, 681 mobile phones have display screen damage failures, 54 mobile phones have housing damage failures,..., 41 mobile phones have interface failures. Among them, the failure frequencies of the motherboard and the display screen are the highest, much higher than other components in the target product. Then, according to the failure situation, the key components of the X model mobile phone can be determined as the motherboard and the display screen.

[0071] S202. Determine the relevant failure mechanisms according to the product operation time and the environmental stress conditions related to the key components, in combination with the failure time calculation formula of the typical failure mechanism.

[0072] Specifically, the product operation time refers to the time period from the first activation of the product's function to its failure (for non-failed products, it corresponds to the current time). Different types of target products may correspond to different typical failure mechanisms. Exemplarily, in the case where the target product is an electronic product, its corresponding typical failure mechanisms may include the electromigration failure mechanism and the corrosion failure mechanism. The formula for calculating the failure time of the electromigration failure mechanism is where A 0 is the correlation coefficient related to the process flow and component materials, and this coefficient varies with different device manufacturers; n is the current density exponent, which is related to the device material; Q is the activation energy, and in the electromigration failure mechanism, its value is related to the material of the product contact area. The formula for calculating the failure time of the corrosion failure mechanism is b is the humidity-related coefficient; RH is the relative humidity, expressed as a percentage; Q is the activation energy, and in the electromigration failure mechanism, its value is basically equal to the activation energy of liquid corrosion, and it is related to both the liquid properties and the metal material.

[0073] According to the formula for calculating the failure time, the relevant stresses of the typical failure mechanism and the relative relationship between the failure time and the relevant stresses can be determined. Furthermore, based on the relative relationship between the product operation time and the environmental stress conditions in the failure data, it can be judged whether the target product conforms to the relevant stresses in each typical failure mechanism and the relative relationship between the failure time and the relevant stresses. If so, it is determined that this typical failure mechanism is the relevant failure mechanism of the target product. Exemplarily, assuming that according to the formula for calculating the failure time, it can be determined that the sensitive forces corresponding to the electromigration failure mechanism are current and temperature, and the failure time is positively correlated with both current and temperature as preset. And based on the product operation time in the failure data and the environmental stress conditions related to the key components, it can be determined that the failure time of the target product is also positively correlated with both current and temperature as preset. Then it can be determined that the electromigration failure mechanism is the relevant failure mechanism of the target product.

[0074] S203. Determine the sensitive stress for the target product according to the relevant stresses of the relevant failure mechanism.

[0075] Specifically, the relevant stresses of the relevant failure mechanism refer to the environmental stresses that have an inevitable correlation with the failure time of the relevant failure mechanism. Exemplarily, the relevant stresses of the corrosion failure mechanism include environmental humidity and device materials.

[0076] Exemplarily, steps S201, S202, and S203 can be implemented as follows: Conduct a research on historical failure data for multiple samples of electronic products, and collect the environmental stress conditions, product operation time, and failure situations of the products during stages such as the development process, qualification tests, and field use of the electronic products. Then, based on the collected data, conduct a study on typical failure mechanisms, focusing on failure mechanisms such as electromigration, stress migration, corrosion, thermal cycling and thermal fatigue, time-dependent dielectric breakdown, and hot carriers, and determine the key devices and sensitive stress types of the electronic products. For example, when conducting a failure mechanism analysis on a certain instrument, it can be determined that its three key parts are the instrument housing exposed to the atmospheric environment, the sensor in direct contact with the working medium, and the circuit system sealed inside the instrument, and the main sensitive stresses are temperature, relative humidity, and electrical stress.

[0077] S204. Use a clustering analysis algorithm to perform clustering analysis on the failure data related to the key devices according to the sensitive stresses.

[0078] Specifically, based on the typical failure mechanism analysis, conduct clustering analysis on the product failure data caused by all key devices. Among them, the clustering analysis algorithm can be the quick sampling analysis algorithm. Use the quick clustering analysis to classify the failure data related to the key devices according to the environmental characteristics (i.e., the sensitive stresses of the target product). Let the data set of k initial cluster centers classified according to the environmental characteristics be Implement the initial classification using the following principle. Denote Divide the failure samples into non-overlapping k classes. The classification principle is that each failure sample is classified into the closest initial cluster center. According to the above calculation, an initial classification is obtained. Starting from G (0) calculate a new set of cluster centers L (1) , and use the center of gravity as the new cluster center: In this way, a new set of cluster centers is obtained. Starting from L (1) classify the samples again and denote to obtain the classification Repeat this process in turn. Suppose the classification is obtained at the m-th step. In the above recurrence process, is the center of gravity of class , not the center of gravity of . When m gradually increases, the classification tends to be stable. In actual calculation, starting from a certain step m, the classification is exactly the same as , and the clustering ends.

[0079] S205. According to the results of the clustering analysis, determine the failure contribution rates of each sensitive stress to the target product.

[0080] Specifically, in the clustering analysis results, each category corresponds to a related failure mechanism of the target product. And the related failure mechanism corresponds to at least one sensitive stress. According to the ratio of the number of data points in the category corresponding to the failure mechanism of the sensitive stress to the total number of data points, the failure contribution rate of the sensitive stress can be determined. Exemplarily, assume that in the clustering analysis results, a category corresponds to sensitive forces including current density, and the number of data points within the range of this category is 241, and the total number of data points is 1000. Then the total failure contribution rate of the sensitive stress of current density is 241÷1000 = 0.241 = 24%.

[0081] S206. Determine the number of groups of the multi-stress accelerated life test according to the number of sensitive stresses.

[0082] Specifically, each group of the multi-stress accelerated life test corresponds to at least two sensitive stresses. Then when grouping the multi-stress accelerated life test, the number combination value of the sensitive stresses can be solved, where the number combination value can be the number of groups of combinations of two sensitive stresses. Determine the number of groups of the multi-stress accelerated life test to be greater than or equal to the number combination value. Exemplarily, in the case where the number of sensitive stresses is 10, calculate the number of groups of pairwise combinations of 10 sensitive stresses If the multi-stress accelerated life test only involves the coupling of two environmental stresses, then the number of test groups can be determined to be 45 groups. If the multi-stress accelerated life test also involves the coupling of three environmental stresses, then on the basis of 45 groups, the number of groups of combinations of three out of 10 sensitive stresses needs to be added which is equal to 165. According to the actual requirements of the test, by analogy and summation, the number of groups of the multi-stress accelerated life test can be calculated.

[0083] S207. Determine the number of samples in each group of the multi-stress accelerated life test according to the fitting distribution model selected for the multi-stress accelerated life test.

[0084] Specifically, the fitting distribution model is a statistical technique used to find a theoretical probability distribution that is as close as possible to the distribution of a set of observed data. Its purpose is to describe the internal law of the data through a known distribution function (such as normal distribution, Poisson distribution, exponential distribution, etc.). Different fitting distribution models require different numbers of samples. Exemplarily, the fitting distribution model of the experimental data includes the Weibull distribution model, and the number of samples in each group of the multi-stress accelerated life test corresponding to the Weibull distribution model is greater than 3.

[0085] S208. Use a preset high value to conduct an accelerated life test on each group of samples to obtain the life data of the target product under different stress conditions.

[0086] Specifically, the accelerated life test includes failure tests and degradation tests. In order to complete the accelerated life test at a faster speed, a corresponding preset high value can be adopted for each sensitive stress. Exemplarily, in the test of studying the influence of the coupling conditions of temperature and humidity on the product life, high value preset temperature and preset humidity that can accelerate the failure or damage of the product can be adopted for the test.

[0087] Exemplarily, it is assumed that the accelerated life test is carried out according to the following life equation η = f(T, V, I, RH|Θ), where η is the life of the target product, T is the temperature stress, V is the vibration stress, I is the electrical stress, RH is the humidity stress, and Θ is the set of unknown parameters. After obtaining the failure time t of the jth sample in the ith group of tests ij , the set of unknown parameters Θ can be obtained by maximum likelihood estimation, so as to obtain the life equation of the product. Using the life equation and inputting different environmental stresses, the accelerated test life data of the target product can be obtained.

[0088] The life data can include real test data and extrapolated test data. The real test data refers to the determination of the failure time by detecting the failure of the target product in the accelerated degradation test. If the failure of the product cannot be detected in a long time, the failure time of the product can be extrapolated by detecting the degree of degradation of the product performance, as the extrapolated test data. For most products with performance degradation, their degradation trajectories can generally be effectively fitted by the following five degradation trajectory models. Under the condition of allowable error, the effectively fitted degradation trajectory can be regarded as the actual degradation trajectory of the product. The five degradation trajectory models include: linear model y i = α i ·t + β i , exponential model power law model logarithmic model y i = α i ·ln(t) + β i and Lloyd-Lipow model where y i represents the performance degradation parameter of the product; i represents the number of samples tested under a certain stress level; t is the test time; α i , β i are model parameters. The core purpose of degradation data processing is to estimate the unknown parameters in the model according to the measured performance degradation data in the test, and at the same time compare which model is more suitable, and finally extrapolate the failure time of the product.

[0089] The data processing process of the degradation model is as follows: linearize the degradation model, including logarithmization and parametric substitution method. Substitute the five linearized data into the linear regression equation: assume that factor x has a linear influence relationship with index y, that is, y = β0 +β 1 x + ε. To estimate β 0 and β 1 , the test observation levels X 1 ,..., X n are taken to obtain n independent observed values Y of y 1 ,..., Y n Estimation is obtained using the least squares method and Substitute the test observation levels X 1 ,..., X n into the five regression equations Compare with the observed values, and the one with the smallest error can be considered the most suitable degradation model. Using the most suitable regression equation combined with the failure threshold, the extrapolated test data of the product is given as the accelerated test life of the product.

[0090] S209. Use the test data of the multi-stress accelerated life test to train the neural network model.

[0091] Specifically, use the test data determined by the test in step S208 to train the neural network model. Figure 3 This is a training framework diagram of a neural network model provided by an embodiment of the present invention. Combining Figure 3 , the input layer is the parameters related to the accelerated test life of the target product. For example, the parameters of the input layer can include three sensitive stresses of temperature, humidity, and electrical stress and a reliability parameter, and the output layer is the accelerated test life of the target product. The number of neurons in the hidden layer depends on the degree of non-linear mapping between the input and output parameters. In the present invention, the cross-validation method is used to determine the number of neurons in the hidden layer. The cross-validation method randomly divides the data into three parts: a training set, a validation set, and a test set, with the three parts accounting for 60%, 20%, and 20% in sequence. Among them, the training set is used to train the model, the validation set is used to select the number of neurons in the hidden layer, and the test set is used to evaluate the prediction ability of the model. For hidden layers with different numbers of neurons, the mean absolute error M AE and the mean relative error M RE are used to characterize the error between the life estimation value and the life measurement value of the accelerated test. The expression of the mean absolute error M AE is The expression of the mean relative error is where X i and Y i are respectively the measured value of the product accelerated test life and the estimated value of the neural network model; i is the sample serial number of the data set; N is the number of (X i , Y i ) data.

[0092] After determining the number of neurons in the hidden layer, the neural network model can be built. To avoid interference from data in different units during model training, all data in the model needs to be normalized. The normalization processing expression is where x is the normalized parameter value, x ∈ [-1, 1]; X is the parameter value before correction, X max and X min are the maximum and minimum values of the corresponding parameter factors respectively.

[0093] Exemplarily, the neural network model includes a Back-Propagation Neural Network (abbreviated as BP neural network) model. To improve the prediction accuracy and speed, the present invention uses a genetic algorithm to optimize the BP neural network model and initialize the weights and thresholds, that is, a Genetic Algorithm-Back-Propagation Neural Network (abbreviated as GA-BP neural network model). Figure 4 FIG. is a training flowchart of a GA-BP neural network model provided by an embodiment of the present invention. Referring to Figure 4 , the training steps of the GA-BP neural network model include: S401. Determine the topological structure of the BP neural network, where the topological structure of the BP neural network includes input variables, output variables, and the number of neurons in the hidden layer. S402. After the network structure is determined, initialize the weights and thresholds of the BP neural network. S403. Encode the weights and thresholds using a genetic algorithm. S404. Initialize the population and determine the fitness function. S405. Perform selection, crossover, and mutation operations. S406. Calculate the fitness. S407. When the fitness meets the fitness condition, obtain the optimal weights and thresholds. S408. Use the accelerated test life data of the product, combined with the optimal weights and thresholds, to train the BP neural network model. S409. Calculate the estimation error of the BP neural network model using the verification data. S410. Update the weights and thresholds. S411. When the error of the BP neural network model meets the error condition, end the model training and determine the structure of the neural network model.

[0094] S210. Use the trained neural network model to predict the accelerated test life of the target product under the target stress condition

[0095] Exemplarily, inputting the target stress condition data into the trained neural network model can predict the accelerated test life of the target product under various stress coupling conditions. Figure 5 FIG. is a comparison chart of the life prediction value obtained by a life prediction method for an accelerated test product provided by an embodiment of the present invention and the measured value of the target product life obtained from the actual test. CombiningFigure 5 The results show that the GA-BP neural network model determined by the method of this application can accurately estimate the accelerated test life of the product.

[0096] The life prediction method for the accelerated test product provided in this embodiment combines existing data, the fast clustering analysis algorithm, and actual accelerated life test data to determine more and more comprehensive failure-related data, thereby realizing more comprehensive and in-depth training of the neural network. In addition, a backpropagation neural network model optimized by the genetic algorithm is selected to further improve the prediction accuracy and speed.

[0097] The embodiment of the present invention also provides a life prediction device for an accelerated test product. Figure 6 It is a schematic diagram of the composition of the life prediction device for the accelerated test product provided in the embodiment of the present invention. Refer to Figure 6 As shown in, the life prediction device 600 for the accelerated test product includes a failure mechanism analysis module 601, a clustering analysis module 602, an accelerated life test module 603, a model training module 604, and a prediction module 605. The failure mechanism analysis module 601 is used to carry out typical failure mechanism analysis based on the failure data of the target product to determine the key components and sensitive stresses of the target product. The clustering analysis module 602 is used to perform clustering analysis on the failure data related to the key components to determine the failure contribution rate of each sensitive stress to the target product. The accelerated life test module 603 is used to carry out multi-stress accelerated life tests on the target product according to the failure contribution rate corresponding to each sensitive stress. The model training module 604 is used to train the neural network model with the test data of the multi-stress accelerated life test. The prediction module 605 is used to predict the accelerated test life of the target product under the target stress conditions by using the trained neural network model.

[0098] The life prediction device for the accelerated test product provided in the embodiment of the present invention can execute the life prediction method for the accelerated test product provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects for executing the method.

[0099] Figure 7 It shows a schematic diagram of the structure of an electronic device that can be used to implement the embodiments of the present invention. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workbenches, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smart phones, wearable devices (such as helmets, glasses, watches, etc.) and other similar computing devices. The components shown herein, their connections and relationships, and their functions are only examples and are not intended to limit the implementation of the present invention described and / or claimed herein.

[0100] As shown Figure 7 in FIG. 700, the electronic device 700 includes at least one processor 11, and a memory communicatively connected to the at least one processor 11, such as a read-only memory (ROM) 12, a random access memory (RAM) 13, etc. The memory stores a computer program executable by the at least one processor. The processor 11 can perform various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 12 or the computer program loaded from the storage unit 18 into the random access memory (RAM) 13. In the RAM 13, various programs and data required for the operation of the electronic device 700 can also be stored. The processor 11, the ROM 12, and the RAM 13 are connected to each other via a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0101] Multiple components in the electronic device 700 are connected to the I / O interface 15, including: an input unit 16, such as a keyboard, a mouse, etc.; an output unit 17, such as various types of displays, speakers, etc.; a storage unit 18, such as a magnetic disk, an optical disk, etc.; and a communication unit 19, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 19 allows the electronic device 700 to exchange information / data with other devices via a computer network such as the Internet and / or various telecommunication networks.

[0102] The processor 11 can be various general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The processor 11 executes the various methods and processes described above, such as the method for predicting the life of an accelerated test product.

[0103] In some embodiments, the method for predicting the life of an accelerated test product can be implemented as a computer program tangibly embodied in a computer-readable storage medium, such as the storage unit 18. In some embodiments, part or all of the computer program can be loaded and / or installed onto the electronic device 700 via the ROM 12 and / or the communication unit 19. When the computer program is loaded into the RAM 13 and executed by the processor 11, one or more steps of the method for predicting the life of an accelerated test product described above can be executed. Alternatively, in other embodiments, the processor 11 can be configured to execute the method for predicting the life of an accelerated test product by any other appropriate means (e.g., by means of firmware).

[0104] The various embodiments of the systems and techniques described above in this specification can be implemented in digital electronic circuitry, integrated circuit systems, field programmable gate arrays (FPGA), application specific integrated circuits (ASIC), application specific standard products (ASSP), systems on chip (SOC), complex programmable logic devices (CPLD), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include: being implemented in one or more computer programs that are executable and / or interpretable on a programmable system including at least one programmable processor, which may be a special-purpose or general-purpose programmable processor that receives data and instructions from, and transmits data and instructions to, a storage system, at least one input device, and at least one output device.

[0105] The computer programs for implementing the methods of the present invention can be written in any combination of one or more programming languages. These computer programs can be provided to a processor of a general purpose computer, special purpose computer, or other programmable data processing apparatus, such that the computer programs, when executed by the processor, cause the functions / operations specified in the flowchart and / or block diagram to be implemented. The computer programs can be executed entirely on the machine, partly on the machine, as a stand-alone software package partly on the machine and partly on a remote machine or entirely on the remote machine or server.

[0106] In the context of the present invention, a computer-readable storage medium can be a tangible medium that can contain or store a computer program for use by or in connection with an instruction execution system, apparatus, or device. The computer-readable storage medium can include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. Alternatively, the computer-readable storage medium can be a machine-readable signal medium. More specific examples of the machine-readable storage medium would include an electrical connection based on one or more wires, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0107] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and a pointing device (e.g., a mouse or a trackball) through which the user can provide input to the electronic device. Other kinds of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0108] The systems and techniques described herein can be implemented in a computing system including backend components (e.g., as a data server), or a computing system including middleware components (e.g., an application server), or a computing system including frontend components (e.g., a user computer having a graphical user interface or a web browser through which the user can interact with an implementation of the systems and techniques described herein), or a computing system including any combination of such backend components, middleware components, or frontend components. The components of the system can be interconnected to each other by digital data communication in any form or medium (e.g., a communication network). Examples of communication networks include: local area network (LAN), wide area network (WAN), blockchain network, and the Internet.

[0109] The computing system can include a client and a server. The client and the server are generally far from each other and usually interact through a communication network. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or a cloud host, which is a host product in the cloud computing service system and solves the defects of difficult management and weak business scalability existing in traditional physical hosts and VPS services.

[0110] It should be understood that various forms of the processes shown above can be used, with steps reordered, added, or deleted. For example, the steps recited in the present invention can be executed in parallel, sequentially, or in a different order, as long as the desired results of the technical solution of the present invention can be achieved, and no limitation is imposed herein.

[0111] The above specific embodiments do not constitute a limitation on the protection scope of the present invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A method for predicting the life of an accelerated test product, characterized in that: include: Based on the failure data of the target product, conduct typical failure mechanism analysis to determine the key components and sensitive stresses of the target product; Performing cluster analysis on the failure data related to the key components to determine the failure contribution rate of each of the sensitive stresses to the target product; Conducting a multi-stress accelerated life test of the target product according to the failure contribution rate corresponding to each of the sensitive stresses; Using the test data of the multi-stress accelerated life test to train the neural network model; The trained neural network model is used to predict the accelerated test life of the target product under target stress conditions.

2. The method according to claim 1, characterized in that The failure data of the target product includes environmental stress conditions, product operation time and failure conditions; The typical failure mechanism analysis is carried out based on the failure data of the target product to determine the key components and sensitive stresses of the target product, including: According to the fault condition, determining the key component of the target product; Determine the relevant failure mechanism based on the product operation time and the environmental stress conditions related to the key components and the failure time calculation formula of the typical failure mechanism; The sensitive stress of the target product is determined according to the relevant stress of the relevant failure mechanism.

3. The method according to claim 1, characterized in that: The cluster analysis of the failure data related to the key components to determine the failure contribution rate of each sensitive stress to the target product includes: Using a cluster analysis algorithm, cluster analysis is performed on the fault data related to the key components according to the sensitive stress; According to the result of the cluster analysis, the failure contribution rate of each of the sensitive stresses to the target product is determined.

4. The method according to claim 3, characterized in that The cluster analysis algorithm includes a fast cluster analysis algorithm.

5. The method according to claim 1, characterized in that According to the failure contribution rate corresponding to each of the sensitive stresses, a multi-stress accelerated life test of the target product is carried out, including: Determining the number of groups of the multi-stress accelerated life test according to the number of the sensitive stresses; Determine the number of samples in each group in the multi-stress accelerated life test according to the experimental data fitting distribution model selected for the multi-stress accelerated life test; Using a preset high value, an accelerated life test is performed on each group of samples to obtain life data of the target product under different stress conditions, wherein the accelerated life test includes a failure test and a degradation test.

6. The method according to claim 5, characterized in that The experimental data fitting distribution model includes a Weibull distribution model, and the number of samples in each group in the multi-stress accelerated life test corresponding to the Weibull distribution model is greater than 3.

7. The method according to claim 5, characterized in that Determining the number of groups of the multi-stress accelerated life test according to the number of the sensitive stresses includes: Solving the quantitative combination value of the sensitive stress; The number of groups of the multi-stress accelerated life test is determined to be greater than or equal to the quantitative combination value.

8. The method according to claim 1, characterized in that The neural network model includes a back-propagation neural network model.

9. A life prediction device for accelerated test products, characterized in that: include: A failure mechanism analysis module, used to carry out typical failure mechanism analysis based on the failure data of the target product to determine the key components and sensitive stresses of the target product; A cluster analysis module, used to perform cluster analysis on the failure data related to the key components to determine the failure contribution rate of each sensitive stress to the target product; An accelerated life test module, used for carrying out a multi-stress accelerated life test of the target product according to the failure contribution rate corresponding to each of the sensitive stresses; A model training module, used to train a neural network model using test data of the multi-stress accelerated life test; The prediction module is used to predict the accelerated test life of the target product under target stress conditions by using the trained neural network model.

10. An electronic device, characterized in that: The electronic device comprises: at least one processor; and a memory communicatively connected to the at least one processor; wherein, The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can perform the life prediction method of an accelerated test product according to any one of claims 1 to 8.

Images