A method, apparatus, and medium for predicting a degradation path of a temperature and humidity sensitive device or assembly

Abstract

This invention provides a method, device, and medium for predicting the degradation path of temperature and humidity sensitive devices or components. The method includes: acquiring accelerated degradation test data of the device or component under test; preprocessing the acquired accelerated degradation test data to generate a training set; constructing a physical information neural network based on the Peck model, wherein the physical information neural network is data-driven and constrained by the temperature and humidity coupling equation; constructing a hybrid loss function; inputting the training set into the constructed physical information neural network for training until the hybrid loss function is minimized, thereby obtaining a trained physical information neural network; inputting temperature and humidity values ​​and time series under normal operating conditions into the trained physical information neural network to predict the degradation path of the device or component. This invention can effectively improve the prediction capability under few sample conditions and the extrapolation accuracy across stress levels.

Description

Technical Field

[0001] This invention relates to the field of device or component reliability assessment and lifetime prediction technology, and in particular to a method, device and medium for predicting the degradation path of temperature and humidity sensitive devices or components. Background Technology

[0002] With the development of modern industrial technology, various electronic products (such as electronic equipment, photovoltaic modules, and aerospace components) are inevitably exposed to the natural environment for extended periods during their service life. Temperature and humidity are the two most significant environmental stress factors leading to performance degradation of devices or components. Under the coupled effects of these two stresses, chemical oxidation, electrochemical corrosion, or physical aging often occur inside the devices or components, resulting in reduced reliability or even failure.

[0003] To assess the lifespan of such temperature and humidity-sensitive devices or components, it is essential to obtain degradation data throughout their entire lifespan. However, under normal environmental stress conditions, the processes of corrosion in electronic components and aging in polymer materials are often quite slow. Direct natural aging tests are too time-consuming and cannot meet the rapid iteration requirements of device or component development. Therefore, accelerated degradation testing is commonly used in engineering. This involves increasing the level of environmental stress (such as high temperature and high humidity) to accelerate the degradation of devices or components and obtain accelerated degradation data. Then, accelerated models are used to extrapolate the lifespan under normal stress.

[0004] Currently, in lifetime prediction research based on accelerated degradation data, traditional methods mainly include stochastic process models, nonlinear regression models, and cumulative damage models. Xiang et al. [S. Xiang, G. Li, F. Zhou, S. Huang, Z. Xu, G. Jia, F. Niu, Life prediction of ground-wall insulation material inelectric motors based on accelerated degradation test under different failure mechanisms, IEEE Journal of Emerging and Selected Topics in Industrial Electronics 6 (2025) 654–663.] constructed a Wiener inverse Gaussian mixture model to predict the maximum partial discharge of polyimide films and combined it with the Arrhenius model to achieve lifetime assessment of the material at different temperatures. Liu et al. [Q. Liu, W. Shi, Z. Chen, K. Li, H. Liu, S. Li, Rubber accelerated aging life prediction by Peck model considering initial hardness influence, Polymer Testing 80 (2019) 106132.] introduced the Peck model to modify the traditional kinetic degradation path equation for rubber aging, comprehensively considering the dual effects of initial hardness and temperature on the material, and used Monte Carlo simulation to quantify the influence of hardness changes on life distribution, achieving accurate life prediction under normal stress. Escobar et al. [LA Escobar, WQ Meeker, A Review of Accelerated Test Models, Statistical Science 21 (2006.] used accelerated degradation test data to fit a cumulative damage model, and then inferred the life distribution under normal stress levels. This method is widely used in mechanical fatigue and wear. However, although these traditional models have clear physical meaning, they often lack accuracy when dealing with complex multi-stress coupling and nonlinear degradation processes due to oversimplification. It exhibits certain limitations when extrapolating to other stress conditions and has limited ability to describe individual material differences.

[0005] In recent years, machine learning methods such as random forests, support vector machines, and artificial neural networks have been introduced into the field of degradation path prediction. Among all machine learning methods, artificial neural networks are the most widely used and have proven valuable for problems lacking accurate physical models. However, as a purely data-driven model, the prediction accuracy of neural networks is greatly affected by the size of the training dataset. Applying artificial neural networks to lifetime prediction with limited samples may lead to the following problems: (1) the relationship between lifetime and different stress conditions is directly formed by the training samples, which may violate physical laws; (2) due to insufficient training samples, artificial neural networks may overfit, resulting in poor generalization performance, especially in extrapolation prediction under unknown stress conditions.

[0006] Therefore, there is an urgent need for a degradation modeling method that can both utilize the powerful nonlinear mapping capability of neural networks and follow the laws of physical failure, in order to solve the problem of degradation path prediction and cross-stress extrapolation of temperature and humidity sensitive devices or components under the dual stress acceleration conditions of temperature and humidity. Summary of the Invention

[0007] This application provides a method, device, and medium for predicting the degradation path of temperature and humidity sensitive devices or components, in order to solve the problems in the background art.

[0008] Other features and advantages of this application will become apparent from the following detailed description, or may be learned in part from practice of this application.

[0009] According to a first aspect of the embodiments of this application, a method for predicting the degradation path of temperature and humidity sensitive devices or components is provided, including: Acquire accelerated degradation test data for the device or component under test; The acquired accelerated degradation test data are preprocessed to generate a training set; A physical information neural network based on the Peck model is constructed. The physical information neural network is data-driven and constrained by the temperature and humidity coupling equation. Construct a hybrid loss function; The training set is input into the constructed physical information neural network for training until the hybrid loss function is minimized, thus obtaining the trained physical information neural network; By inputting temperature and humidity values ​​and time series under normal operating conditions into a trained physical information neural network, the degradation path of a device or component can be predicted.

[0010] According to one embodiment of this application, obtaining accelerated degradation test data of the device under test or component specifically includes: Multiple combinations of different temperatures and relative humidity are set as accelerated stress conditions, and the performance degradation of the device or component under test is measured periodically to obtain raw accelerated degradation test data including test time, accelerated stress conditions, and corresponding performance degradation observations.

[0011] According to one embodiment of this application, the preprocessing specifically includes: The accelerated degradation test data are mapped to a predetermined interval using the min-max normalization method to form a training set.

[0012] According to one embodiment of this application, the construction of a physical information neural network based on the Peck model specifically includes: A physical information neural network based on a deep neural network model is constructed. The input layer receives normalized temperature, relative humidity, and time. The network contains several hidden layers, each with several neurons, and uses a nonlinear activation function to capture complex nonlinear relationships. The output layer outputs the predicted performance degradation. The physical information neural network uses the temperature-humidity coupling relationship established by the Peck model for physical constraint. The standard lifetime equation of the Peck model is as follows:

[0013] in, For standard degradation rate, For temperature, Relative humidity; , , These are the undetermined physical parameters in the multi-stress acceleration model.

[0014] According to one embodiment of this application, the hybrid loss function is:

[0015] in, For a mixed loss function, For data-driven loss terms, For physical information loss, The hyperparameter of the physical information loss term is used to balance the importance of data-driven and physical information loss. The data-driven loss term measures the difference between the neural network prediction and the actual experimental observation. The physical information loss term measures whether the output of the neural network conforms to the physical constraints described by the Peck model.

[0016] According to one embodiment of this application, the data-driven loss term is specifically:

[0017] in, For data-driven loss terms, The total number of training set data. Let be the predicted value output by the network for the i-th training sample. is the actual experimental observation value corresponding to the i-th training sample.

[0018] According to one embodiment of this application, the physical information loss term construction process includes: Predicted values ​​output by a computational physical information neural network The partial derivative with respect to time yields the predicted degradation rate. ; The standard degradation rate was calculated using the standard lifetime equation of the Peck model. Construct the physical residual equations between the predicted degradation rate and the standard degradation rate:

[0019] The physical information loss term is the mean square value of the physical residual:

[0020] in, For physical information loss, The number of placement points used to calculate physical constraints is randomly sampled and generated within the input space.

[0021] According to one embodiment of this application, the step of inputting the training set into the constructed physical information neural network for training until the hybrid loss function is minimized to obtain the trained physical information neural network specifically includes: The constructed training set and generated configuration points are input into the network, and the composite loss function is minimized using a gradient-based optimization algorithm. During training, the weight parameters of the physical information neural network and the undetermined physical parameters in the Peck model are updated synchronously as training variables.

[0022] According to a second aspect of the embodiments of this application, an electronic device is provided, comprising: At least one processor; and a memory communicatively connected to said at least one processor; The memory stores instructions that can be executed by the at least one processor, and the at least one processor executes the instructions stored in the memory to perform the method described in the first aspect.

[0023] According to a third aspect of the embodiments of this application, a computer-readable storage medium is provided for storing instructions that, when executed, cause the method described in the first aspect to be implemented.

[0024] Compared with existing technologies, the advantages of adopting the above technical solution are as follows: This invention does not treat the neural network as a black box, but introduces the physical mechanism constraint of temperature and humidity on material degradation through the Peck model, ensuring the physical consistency of the prediction results. Compared with pure data-driven methods, this invention can maintain good training effect even with small sample data when extrapolating from high-stress acceleration conditions to low-stress normal conditions. Due to the introduction of physical equations as regularization terms, this method is suitable for engineering scenarios with high experimental costs and limited sample size. Attached Figure Description

[0025] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application. It is obvious that the drawings described below are merely some embodiments of this application, and those skilled in the art can obtain other drawings based on these drawings without any inventive effort.

[0026] Figure 1 This is a flowchart of a method for predicting the degradation path of temperature and humidity sensitive devices or components according to an embodiment of this application.

[0027] Figure 2 This is a scatter plot of the NIST dataset from an embodiment of this application.

[0028] Figure 3 This is a graph showing the evolution of the data-driven loss during iterative training in an embodiment of this application.

[0029] Figure 4 This is a diagram illustrating the evolution of physical information loss during iterative training in an embodiment of this application.

[0030] Figure 5 This is a diagram showing the global prediction results of an embodiment of this application.

[0031] Figure 6 This is a schematic diagram of the structure of an electronic device according to an embodiment of this application. Detailed Implementation

[0032] The embodiments of this application are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar modules or modules having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain this application, and should not be construed as limiting this application. Rather, the embodiments of this application include all variations, modifications, and equivalents falling within the spirit and scope of the appended claims.

[0033] To address the shortcomings of existing technologies, this application proposes a method for predicting the degradation path of temperature and humidity sensitive devices or components based on a physical information neural network. By embedding the Peck physical model describing the effects of temperature and humidity as a constraint term into the loss function of the neural network, the network is forced to learn the physical laws of degradation of the device or component under test in a dual-stress coupling environment of temperature and humidity while ensuring the accuracy of data fitting. This improves the prediction ability under few sample conditions and the extrapolation accuracy across stress levels.

[0034] Please refer to Figure 1 The specific steps of this method for predicting the degradation path of temperature and humidity sensitive devices or components are as follows: S100: Obtain accelerated degradation test data for the device or component under test.

[0035] In this embodiment, accelerated degradation tests can be designed and conducted on-site for the device or component under test (e.g., coating materials, electronic components, etc.) to obtain accelerated degradation test data, or existing accelerated degradation test data can be collected.

[0036] Specifically, during the test, multiple different combinations of temperature (T) and relative humidity (RH) need to be set as accelerated stress conditions. By periodically measuring the key degradation characteristics of the device or component under test, raw accelerated degradation test data containing the test time t, accelerated stress conditions (T, RH), and corresponding performance degradation observations y are obtained.

[0037] S200. Preprocess the acquired accelerated degradation test data to generate a training set.

[0038] Since temperature, humidity, and time have different dimensions and physical meanings, and their numerical values ​​vary significantly, directly inputting them into a neural network would lead to training difficulties. A min-max normalization method is used to normalize the original accelerated degradation test data, mapping the input variables (T,RH,t) and output variable (y) to a specified interval (e.g., [0,1]), thus constructing a training set for the accelerated degradation test data.

[0039] S300. Construct a physical information neural network based on the Peck model. The physical information neural network is data-driven and constrained by the temperature and humidity coupling equation.

[0040] In this embodiment, a physical information neural network is constructed based on a deep neural network model. This network not only relies on data-driven mechanisms but is also constrained by physical equations.

[0041] Specifically, the constructed physical information neural network architecture includes: the input layer receiving normalized temperature. T relative humidity RH and time tThe network structure contains several hidden layers, each containing a number of neurons, and uses non-linear activation functions to capture complex non-linear relationships. The output layer outputs the predicted performance degradation. .

[0042] Furthermore, this physical information neural network uses the temperature-humidity coupling relationship established by the Peck model to constrain physical laws. The Peck model is a temperature-humidity coupled accelerated lifetime model, the core idea of ​​which is to combine the Arrhenius model and the power-law model in a product form, assuming that the failure rate is simultaneously affected by temperature and humidity. In this embodiment, the standard lifetime equation of the Peck model is expressed in the following form:

[0043] in, For standard degradation rate, For temperature, Relative humidity; , , These are the undetermined physical parameters in the multi-stress acceleration model. The equation shows that the degradation rate exhibits a power-law relationship with humidity and an exponential relationship with temperature.

[0044] S400, Construct a hybrid loss function.

[0045] Corresponding to the constructed physical information neural network, the loss function constructed in this embodiment is a hybrid loss function. A composite loss function is defined to guide the update of the neural network parameters. This function is driven by the data loss term. and physical information loss item Weighted composition:

[0046] in, This is a hyperparameter for the physical information loss weight, used to balance the importance of data-driven factors and physical information.

[0047] To more clearly illustrate the constructed hybrid loss function, each loss term will be explained below.

[0048] Data-driven loss term Used to measure neural network predictions In this embodiment, the difference between the observed value y and the actual experimental value is calculated using the mean squared error (MSE):

[0049] in, For data-driven loss terms, The total number of training set data. Let be the predicted value output by the network for the i-th training sample. is the actual experimental observation value corresponding to the i-th training sample.

[0050] Physical information loss item This is used to measure whether the output of the neural network conforms to the physical laws described by the Peck model. In this embodiment, the predicted value of the physical information neural network output is first calculated directly using automatic differentiation technology. The partial derivative with respect to time yields the predicted degradation rate. .

[0051] Then, the standard degradation rate was calculated using the standard lifetime equation of the Peck model. .

[0052] Next, the physical residual equations for the predicted degradation rate and the standard degradation rate are constructed:

[0053] Finally, the physical information loss term is defined as the mean square value of the physical residual:

[0054] in, For physical information loss, The number of configuration points used to calculate physical constraints is randomly sampled and generated within the input space, without requiring corresponding real label values.

[0055] S500. Input the training set into the constructed physical information neural network for training until the hybrid loss function is minimized, and the trained physical information neural network is obtained.

[0056] After constructing the physical information neural network and defining the hybrid loss function, the model can be trained and its parameters optimized using the dataset. Specifically, the constructed training set and generated configuration points are input into the network, and then a gradient-based optimization algorithm is used to minimize the composite loss function. Training can be completed in this way.

[0057] It should be noted that during the training process, the weight parameters of the physical information neural network and the undetermined physical parameters in the Peck model are... , , This process updates synchronously as a training variable. It not only achieves the fitting of the degradation path but also completes the inversion identification of unknown physical parameters.

[0058] S600: By inputting temperature and humidity values ​​under normal operating conditions and time series data into the trained physical information neural network, the degradation path of a device or component can be predicted.

[0059] Finally, based on the trained physical information neural network, by inputting the temperature and humidity values ​​under normal operating conditions (low stress level) and the corresponding time series, the degradation path of the device or component under normal stress conditions can be predicted.

[0060] To more clearly illustrate the degradation path prediction method proposed in this application, the following detailed implementation process of the method is explained using coating materials as an example.

[0061] Step A: Obtain accelerated degradation test data of coating materials This embodiment uses the NIST (National Institute of Standards and Technology) dataset on photodegradation of coating materials as an example to verify the effectiveness of the method proposed in this invention. This dataset contains measurement data on the change in gloss of coating materials with exposure time under different temperature and humidity conditions.

[0062] Step B: Data Preprocessing Five datasets under accelerated aging conditions were selected from the NIST dataset. The first 80% of the data in each dataset was used as the training set, and the last 20% was used as the validation set. The specific acceleration conditions are as follows: Condition A: Temperature 35℃, relative humidity 0%; Condition B: Temperature 35℃, relative humidity 25%; Condition C: Temperature 45℃, relative humidity 25%; Condition D: Temperature 35℃, relative humidity 50%; Condition E: Temperature 45℃, relative humidity 50%.

[0063] Meanwhile, to verify the model's extrapolation ability, a set of data under relatively low stress conditions (temperature 25℃, relative humidity 0%) was selected as the test set. The scatter plot of the dataset is shown below. Figure 2 As shown. The minimum-maximum normalization method is used to map it to the [0,1] interval. In order to prevent gradient problems in the power-law calculation in the Peck model, the normalized minimum humidity is set to 0.1 (instead of 0).

[0064] Step C: Construct a physical information neural network based on the Peck model A Physical Information Neural Network (PINN) based on residual connections is constructed. The input layer contains 3 nodes (corresponding to normalized time t, temperature T, and relative humidity RH, respectively), and the output layer contains 1 node (predicting the normalized degradation). The hidden layer consists of 4 cascaded residual blocks with the number of neurons decreasing sequentially to 256, 128, 64, and 32. The LeakyReLU activation function is selected.

[0065] Step D: Construct a composite loss function Because there is a quantitative difference between the degradation rate and the amount of degradation, we set the maximum physical constraint weight to 1400. Furthermore, we implemented a dynamic weighting strategy, gradually increasing the physical constraint weight from 1 to 1400. This is because the neural network strives to simultaneously satisfy data fitting requirements and complex differential equations during the initial training phase, and dynamic weighting helps prevent the training process from getting trapped in local optima. This approach balances the initial data fitting requirements with the subsequent physical consistency requirements.

[0066] Step E: Model Training and Parameter Optimization Using the training set obtained in step B, construct the total loss function, which includes MSE data loss and Peck physical residual loss, according to step D. Optimize using the Adam optimizer, with an initial learning rate set to... The number of iterations is set to 300. The unknown parameters in the Peck model ( , , The data is initialized with reasonable random values ​​and set as a trainable variable. During training, the neural network continuously adjusts its weights to fit the observed data while trying to satisfy the constraints of the Peck model. During model training, the co-evolution of data-driven loss and physical information loss reflects the optimization dynamics and inherent consistency of the physical information neural network. Figure 3 As shown, during the early training phase (approximately 50 epochs), the data-driven loss value rapidly decreased from about 100 to... The magnitude was on the order of magnitude, and gradually stabilized at a low level during subsequent training. Meanwhile, physical information loss (such as...) Figure 4 The model exhibits a similar rapid convergence trend, indicating that while fitting the observed data, it continuously strengthens its adherence to the embedded physical laws. The simultaneous reduction and eventual maintenance of both losses at low magnitudes demonstrate that the constructed model can effectively coordinate data-driven learning and physical constraints, achieving stable optimization with limited samples without significant overfitting or violation of physical mechanisms. This consistent convergence behavior of the dual losses not only reflects the model's strong fitting ability under training conditions but also lays a foundation for physical consistency in reliable extrapolation predictions under unknown stress conditions.

[0067] Step F: Degradation Path Prediction and Extrapolation After training, the temperature and humidity conditions and time series data from the test set are input into the PINN model to obtain the predicted degradation path. To intuitively evaluate the model's generalization ability and physical consistency, the global prediction results under dual stress conditions are shown below. Figure 5As shown, the solid and dashed lines represent the model's fit on the training and validation sets, respectively, demonstrating a high degree of consistency with experimental observations under the five training conditions. Importantly, the figure highlights the model's performance on the extrapolated test set ([0%, 25°C]), indicated by blue markers. Although this condition was completely excluded from the training process, PINN's predicted path (red dashed line) accurately followed the unseen experimental data, validating that the embedded Peck model successfully guides the neural network to learn the underlying physical mechanisms, rather than simply memorizing data points. Furthermore, Figure 5 The illustrations in the figure show the identified physical parameters ( This further confirms the interpretability of the model. As a supplement to the evaluation, Table 1 provides a quantitative assessment of the model's accuracy. Across all datasets, the mean squared error remains... The lower order of magnitude.

[0068] Table 1. Average MSE for each dataset

[0069] As can be seen from the above embodiments, the degradation path prediction method for temperature and humidity sensitive devices or components based on physical information neural networks proposed in this invention effectively integrates the advantages of physical mechanisms and data-driven approaches, solves the limitations of traditional methods in small sample and extrapolation prediction scenarios, and provides an accurate and reliable new approach for the life assessment of temperature and humidity sensitive devices or components.

[0070] Based on the same technical concept, embodiments of this application also provide an electronic device that can implement the degradation path prediction method for temperature and humidity sensitive devices or components provided in the above embodiments of the present invention. In one embodiment, the electronic device can be a server, a terminal device, or other electronic equipment. Figure 6 As shown, the electronic device may include: At least one processor and a memory connected to the at least one processor. In this embodiment of the invention, the specific connection medium between the processor and the memory is not limited. Figure 6 The example used is the connection between the processor and memory via a bus. The bus... Figure 6 The connections between other components are indicated by thick lines and are for illustrative purposes only, not as limiting information. Buses can be divided into address buses, data buses, control buses, etc., but for ease of representation, [the specific bus type is not shown here]. Figure 6 The processor is represented by a single thick line, but this does not imply that there is only one bus or one type of bus. Alternatively, a processor can also be called a controller; there are no restrictions on the name.

[0071] In this embodiment of the invention, the memory stores instructions executable by at least one processor. By executing the instructions stored in the memory, the at least one processor can perform the degradation path prediction method for temperature and humidity sensitive devices or components described above. The processor can implement... Figure 6 The functions of each module in the device shown.

[0072] The processor is the control center of the device. It can connect to various parts of the control device through various interfaces and lines. By running or executing instructions stored in memory and calling data stored in memory, it can monitor the device's various functions and process data, thereby enabling overall monitoring of the device.

[0073] In an alternative design, the processor may include one or more processing units. The processor may integrate an application processor and a modem processor, wherein the application processor primarily handles the operating system, user interface, and applications, while the modem processor primarily handles wireless communication. It is understood that the modem processor may also not be integrated into the processor. In some embodiments, the processor and memory may be implemented on the same chip; in some embodiments, they may also be implemented separately on separate chips.

[0074] The processor can be a general-purpose processor, such as a CPU, digital signal processor, application-specific integrated circuit, field-programmable array, or other programmable logic device, discrete gate or transistor logic device, or discrete hardware component, capable of implementing or executing the methods, steps, and logic block diagrams disclosed in the embodiments of this invention. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the degradation path prediction method for temperature and humidity sensitive devices or components disclosed in the embodiments of this invention can be directly manifested as being executed by a hardware processor, or executed by a combination of hardware and software modules within the processor.

[0075] Memory, as a non-volatile computer-readable storage medium, can be used to store non-volatile software programs, non-volatile computer-executable programs, and modules. Memory can include at least one type of storage medium, such as flash memory, hard disk, multimedia card, card-type memory, random access memory (RAM), static random access memory (SRAM), programmable read-only memory (PROM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), magnetic memory, magnetic disk, optical disk, etc. Memory is any other medium capable of carrying or storing desired program code in the form of instructions or data structures, and accessible by a computer, but is not limited thereto. In embodiments of the present invention, memory can also be a circuit or any other device capable of implementing storage functions, used to store program instructions and / or data.

[0076] By designing and programming the processor, the code corresponding to the degradation path prediction method for temperature and humidity sensitive devices or components described in the foregoing embodiments can be embedded into the chip, thereby enabling the chip to execute the steps of the methods described in the foregoing embodiments during operation. How to design and program the processor is a technique well-known to those skilled in the art and will not be elaborated upon here.

[0077] Based on the same inventive concept, embodiments of the present invention also provide a storage medium storing computer instructions that, when executed on a computer, cause the computer to perform a degradation path prediction method for a temperature and humidity sensitive device or component as described above.

[0078] In some alternative embodiments, the present invention also provides a method for predicting the degradation path of a temperature and humidity sensitive device or component, which can also be implemented in the form of a program product including program code that, when the program product is run on a device, causes the control device to perform the steps in a method for predicting the degradation path of a temperature and humidity sensitive device or component according to various exemplary embodiments of the present invention as described in this specification.

[0079] It should be noted that although several units or sub-units of the apparatus have been mentioned in the detailed description above, this division is merely exemplary and not mandatory. In fact, according to embodiments of the invention, the features and functions of two or more units described above can be embodied in one unit. Conversely, the features and functions of one unit described above can be further divided and embodied by multiple units. Furthermore, although the operation of the method of the invention is described in a specific order in the drawings, this does not require or imply that these operations must be performed in that specific order, or that all the operations shown must be performed to achieve the desired result. Additionally or alternatively, certain steps may be omitted, multiple steps may be combined into one step, and / or one step may be broken down into multiple steps.

[0080] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0081] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a server, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0082] Program code for performing the operations of this invention can be written using any combination of one or more programming languages, including object-oriented programming languages ​​such as Java and C++, as well as conventional procedural programming languages ​​such as C or similar languages. The program code can be executed entirely on the user's computing device, partially on the user's device, as a standalone software package, partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server.

[0083] In cases involving remote computing devices, the remote computing device can be connected to the user's computing device via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computing device (e.g., via the Internet using an Internet service provider).

[0084] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0085] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0086] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method of predicting a degradation path of a temperature and humidity sensitive device or assembly, characterized by, include: Acquire accelerated degradation test data for the device or component under test; The acquired accelerated degradation test data are preprocessed to generate a training set; A physical information neural network based on the Peck model is constructed. The physical information neural network is data-driven and constrained by the temperature and humidity coupling equation. Construct a hybrid loss function; The training set is input into the constructed physical information neural network for training until the hybrid loss function is minimized, thus obtaining the trained physical information neural network; By inputting temperature and humidity values ​​and time series under normal operating conditions into a trained physical information neural network, the degradation path of a device or component can be predicted.

2. The method of claim 1, wherein the temperature and humidity sensitive device or assembly degradation path prediction method is characterized by, The acquisition of accelerated degradation test data for the device or component under test specifically includes: Multiple combinations of different temperatures and relative humidity are set as accelerated stress conditions, and the performance degradation of the device or component under test is measured periodically to obtain raw accelerated degradation test data including test time, accelerated stress conditions, and corresponding performance degradation observations.

3. The method for predicting the degradation path of temperature and humidity sensitive devices or components according to claim 1, characterized in that, The preprocessing specifically includes: The accelerated degradation test data are mapped to a predetermined interval using the min-max normalization method to form a training set.

4. The method for predicting the degradation path of temperature and humidity sensitive devices or components according to claim 1, characterized in that, The construction of the physical information neural network based on the Peck model specifically includes: A physical information neural network based on a deep neural network model is constructed. The input layer receives normalized temperature, relative humidity, and time. The network contains several hidden layers, each with several neurons, and uses a nonlinear activation function to capture complex nonlinear relationships. The output layer outputs the predicted performance degradation. The physical information neural network uses the temperature-humidity coupling relationship established by the Peck model for physical constraint. The standard lifetime equation of the Peck model is as follows: in, For standard degradation rate, For temperature, Relative humidity; , , These are the undetermined physical parameters in the multi-stress acceleration model.

5. The method for predicting the degradation path of temperature and humidity sensitive devices or components according to claim 1, characterized in that, The hybrid loss function is: in, For a mixed loss function, For data-driven loss terms, For physical information loss, The hyperparameter of the physical information loss term is used to balance the importance of data-driven and physical information loss. The data-driven loss term measures the difference between the neural network prediction and the actual experimental observation. The physical information loss term measures whether the output of the neural network conforms to the physical constraints described by the Peck model.

6. The method for predicting the degradation path of temperature and humidity sensitive devices or components according to claim 5, characterized in that, The data-driven loss term is specifically: in, For data-driven loss terms, The total number of training set data. Let be the predicted value output by the network for the i-th training sample. is the actual experimental observation value corresponding to the i-th training sample.

7. The degradation path prediction method for temperature and humidity sensitive devices or components according to claim 6, wherein the physical information loss term construction process of the claim includes: Predicted values ​​output by a computational physical information neural network The partial derivative with respect to time yields the predicted degradation rate. ; The standard degradation rate was calculated using the standard lifetime equation of the Peck model. Construct the physical residual equations between the predicted degradation rate and the standard degradation rate: The physical information loss term is the mean square value of the physical residual: in, For physical information loss, The number of placement points used to calculate physical constraints is randomly sampled and generated within the input space.

8. The method for predicting the degradation path of temperature and humidity sensitive devices or components according to claim 1, characterized in that, The step of inputting the training set into the constructed physical information neural network for training until the hybrid loss function is minimized to obtain the trained physical information neural network specifically includes: The constructed training set and generated configuration points are input into the network, and the composite loss function is minimized using a gradient-based optimization algorithm. During training, the weight parameters of the physical information neural network and the undetermined physical parameters in the Peck model are updated synchronously as training variables.

9. An electronic device, characterized in that, include: At least one processor; and a memory communicatively connected to the at least one processor; The memory stores instructions executable by the at least one processor, which executes the instructions stored in the memory to perform the method as described in any one of claims 1 to 8.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium is used to store instructions that, when executed, cause the method as described in any one of claims 1 to 8 to be implemented.

Images