Method, device and equipment for predicting service life of energetic device and storage medium

Abstract

The application discloses an energetic device life prediction method, device and equipment and a storage medium. The method comprises the following steps: acquiring performance parameters and storage environment parameters of an energetic device; identifying key degradation factors from the performance parameters and the storage environment parameters by using mutual information analysis; constructing a physically constrained neural network model based on the key degradation factors, wherein the physically constrained neural network model comprises constraint conditions based on physical knowledge; and predicting the life of the energetic device by using the physically constrained neural network model to obtain a prediction result. Compared with the prior art, the application effectively improves the accuracy and reliability of the life prediction of the energetic device by identifying the key degradation factors from the performance parameters and the storage environment parameters by using the mutual information analysis and predicting the life of the energetic device by using the physically constrained neural network model to obtain the prediction result.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method, apparatus, device, and storage medium for predicting the lifetime of energetic devices. Background Technology

[0002] Energetic devices are devices or components that contain energetic materials (such as gunpowder, propellants, etc.) and can release a large amount of energy and do work through their own rapid chemical reactions. Their performance reliability after long-term storage is a core indicator related to safety and application. Because energetic materials undergo complex physicochemical changes during storage, their performance degradation process is affected by multiple factors such as time and environmental conditions, making it extremely challenging to accurately predict their lifespan.

[0003] Most current technologies employ accelerated life testing, subjecting energetic devices to harsh environments such as high temperature and humidity to accelerate the performance degradation of energetic materials and predict their lifetime. However, accelerated life testing environments are singular and cannot fully represent long-term natural environmental fluctuations, resulting in poor model robustness. Furthermore, the commonly used Arrhenius model relies on extrapolating high-stress data, making its predictions of environmental temperature and humidity relatively simple. This model also performs poorly on complex, non-monotonic data with varying lifecycles, leading to poor fitting results. In addition, the characteristics of energetic devices dictate that their performance testing is often destructive, meaning the available effective data sample size is very limited. Under small sample conditions, purely data-driven machine learning models are prone to overfitting, while traditional physical models struggle to accurately describe complex nonlinear degradation processes. This contradiction makes it difficult for existing methods to balance accuracy and reliability in engineering practice, resulting in low accuracy and reliability for lifetime prediction of energetic devices.

[0004] Therefore, there is an urgent need for a method to predict the lifetime of energetic devices, which can improve the accuracy and reliability of lifetime prediction. Summary of the Invention

[0005] The main objective of this invention is to provide a method, apparatus, device, and storage medium for predicting the lifetime of energetic devices, aiming to solve the technical problem of low accuracy and reliability in the prediction of the lifetime of energetic devices in the prior art.

[0006] To achieve the above objectives, the present invention provides a method for predicting the lifetime of energetic devices, the method comprising the following steps: Obtain the performance parameters and storage environment parameters of energetic devices; Key degradation factors are identified from the performance parameters and the storage environment parameters using mutual information analysis; A physical constraint neural network model is constructed based on the aforementioned key degradation factors, and the physical constraint neural network model includes constraints based on physical knowledge. The lifetime of the energetic device is predicted using the physical constraint neural network model, and the prediction results are obtained.

[0007] Optionally, the step of identifying key degradation factors from the performance parameters and the storage environment parameters using mutual information analysis includes: The mutual information value between the performance parameter and the storage environment parameter is determined using a nonparametric method based on K-nearest neighbor estimation. The mutual information values ​​are subjected to a significance test to obtain the significance test results; Key degradation factors are identified from the performance parameters and the storage environment parameters based on the mutual information values ​​and the significance test results.

[0008] Optionally, the step of performing a significance test on the mutual information value to obtain the significance test result includes: Based on the mutual information values, a hypothesis test is constructed; The test results of the hypothesis test are corrected using the multiple hypothesis test correction method to obtain the corrected test results; The corrected test result is compared with the significance level threshold, and the significance test result of the mutual information value is determined based on the comparison result.

[0009] Optionally, the step of constructing a physically constrained neural network model based on the key degradation factor includes: Construct a neural network structure that takes the key degradation factor as input and the target degradation parameter among the performance parameters as output; Physical constraints are added to the neural network structure to obtain an initial physical constraint neural network model, wherein the physical constraints include initial state constraints and failure boundary constraints; Construct the target loss function of the initial physical constraint neural network model based on the physical constraints; Based on the target loss function, the initial physical constraint neural network model is trained using the training dataset to obtain the physical constraint neural network model.

[0010] Optionally, the step of predicting the lifetime of the energetic device using the physical constraint neural network model to obtain the prediction result includes: Based on the physical constraint neural network model, multiple performance degradation trajectories are generated using the Monte Carlo simulation method. The average failure time of the energetic device is determined based on the multiple performance degradation trajectories. The safe storage lifetime of the energetic device is determined based on the mean failure time, and the safe storage lifetime is used as the prediction result.

[0011] Optionally, the step of obtaining the performance parameters and storage environment parameters of the energetic device includes: The pressure-time curve of the energetic device is obtained through pressure testing; Pressure-related parameters and time-related parameters are extracted from the pressure-time curve. The pressure-related parameters include peak pressure, and the time-related parameters include pressure build-up time and peak pressure time. The peak pressure, the pressure build-up time, and the peak pressure time are used as performance parameters. The storage year, storage location, and leakage rate of the energetic device are obtained, and the storage year, storage location, and leakage rate are used as storage environment parameters.

[0012] Optionally, the step of obtaining the storage year, storage location, and leakage rate of the energetic device, and using the storage year, storage location, and leakage rate as storage environment parameters, includes: Obtain the storage year, storage location, and leakage rate of the energetic device; The storage locations are classified by region based on preset geographical classification rules to obtain the region classification results; The storage locations are coded according to the regional classification results to obtain coded storage locations; The storage year, the leakage rate, and the coded storage location are used as storage environment parameters.

[0013] Furthermore, to achieve the above objectives, the present invention also proposes an energetic device lifetime prediction device, the device comprising: The parameter acquisition module is used to acquire the performance parameters and storage environment parameters of the energetic devices; A parameter analysis module is used to identify key degradation factors from the performance parameters and the storage environment parameters using mutual information analysis; The model building module is used to build a physical constraint neural network model based on the key degradation factor, wherein the physical constraint neural network model includes constraints based on physical knowledge. The lifetime prediction module is used to predict the lifetime of the energetic device using the physical constraint neural network model and obtain the prediction result.

[0014] Furthermore, to achieve the above objectives, the present invention also proposes an energetic device lifetime prediction device, the device comprising: a memory, a processor, and an energetic device lifetime prediction program stored in the memory and executable on the processor, the energetic device lifetime prediction program being configured to implement the steps of the energetic device lifetime prediction method as described above.

[0015] In addition, to achieve the above objectives, the present invention also proposes a storage medium storing a power device lifetime prediction program, wherein when the power device lifetime prediction program is executed by a processor, it implements the steps of the power device lifetime prediction method described above.

[0016] This invention discloses a method for acquiring performance parameters and storage environment parameters of energetic devices; using mutual information analysis to identify key degradation factors from these parameters; constructing a physically constrained neural network model based on these key degradation factors, the model including constraints based on physical knowledge; and using this model to predict the lifetime of the energetic device, obtaining prediction results. Because this invention uses mutual information analysis to identify key degradation factors from performance parameters and storage environment parameters, and utilizes a physically constrained neural network model to predict the lifetime of energetic devices, it effectively improves the accuracy and reliability of lifetime prediction for energetic devices compared to existing technologies. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating the first embodiment of the energetic device lifetime prediction method of the present invention; Figure 2 This is an example diagram showing the performance parameters and storage environment parameters of the energetic device in the energetic device lifetime prediction method of the present invention; Figure 3 This is a flowchart illustrating the second embodiment of the energetic device lifetime prediction method of the present invention; Figure 4 This is a schematic diagram of the mutual information analysis matrix between variables in the energetic device lifetime prediction method of the present invention; Figure 5 This is a flowchart illustrating the third embodiment of the energetic device lifetime prediction method of the present invention; Figure 6 This is a structural block diagram of the first embodiment of the energetic device lifetime prediction device of the present invention; Figure 7 This is a schematic diagram of the structure of an energetic device lifetime prediction device in the hardware operating environment involved in the embodiments of the present invention.

[0018] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0019] It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of the invention.

[0020] This invention provides a method for predicting the lifetime of energetic devices, referring to... Figure 1 , Figure 1This is a flowchart illustrating the first embodiment of the energetic device lifetime prediction method of the present invention.

[0021] In this embodiment, the energetic device lifetime prediction method includes steps S10 to S40: Step S10: Obtain the performance parameters and storage environment parameters of the energetic device.

[0022] It should be noted that the executing entity in this embodiment can be a computer server device with data processing, network communication, and program execution functions applied in the scenario of energetic device lifetime prediction, such as a server, tablet computer, personal computer, etc., or an electronic device capable of realizing the above functions (such as an energetic device lifetime prediction device). The following uses a system including an energetic device lifetime prediction device (hereinafter referred to as the system) as an example to illustrate this embodiment and the following embodiments.

[0023] It should be understood that energetic devices refer to devices or components that contain energetic materials (such as gunpowder, propellants, etc.) and can release a large amount of energy and do work through their own rapid chemical reactions. During long-term storage, the energetic materials inside these devices will undergo slow chemical degradation or physical structural changes (collectively referred to as aging) due to environmental factors (such as humidity and temperature), resulting in a decrease in their performance (such as burning rate and pressure output), and ultimately they may be unable to perform their intended function, i.e., they will fail.

[0024] Understandably, performance parameters can refer to indicators that directly reflect and quantify the operational efficiency and dynamic characteristics of energetic devices after triggering; these parameters are direct evidence for assessing whether they have degraded. Storage environment parameters can refer to external conditions and historical records that may affect the performance degradation process of energetic devices during storage; these parameters are the triggers or stresses that lead to performance degradation.

[0025] In this embodiment, for example, referring to Figure 2 Performance parameters can be obtained through pressure-time curves (Pt curves) from pressure tests, and parameters extracted from these curves include peak pressure, pressure build-up time, and peak pressure time. Storage environment parameters include storage year (which can be obtained from warehouse records), storage location, and leakage rate (which can be obtained from helium testing reports).

[0026] In a specific implementation, a pressure-time curve of the energetic device is obtained through pressure testing; pressure-related parameters and time-related parameters are extracted from the pressure-time curve, wherein the pressure-related parameters include peak pressure, and the time-related parameters include pressure build-up time and peak pressure time; the peak pressure, pressure build-up time, and peak pressure time are used as performance parameters; the storage year, storage location, and leakage rate of the energetic device are obtained, and the storage year, storage location, and leakage rate are used as storage environment parameters.

[0027] It's important to understand that peak pressure refers to the maximum pressure value (in MPa) generated by an energetic device during operation. This is a core indicator of its work capacity, and a drop in pressure usually directly indicates performance degradation. Build-up time refers to the time (in ms) required for the pressure to rise to a specific percentage (such as a certain percentage of peak pressure) from the start of operation, reflecting the initial rate of energy release. Peak pressure time refers to the time (in ms) required to reach peak pressure from the start of operation, reflecting the overall rate of energy release.

[0028] It should be understood that storage year can refer to the length of time the energetic device has been stored (unit: years), and time is the most fundamental factor leading to material aging. Storage location can refer to the geographical location of the energetic device. Leakage rate can refer to the quantitative indicator of the sealing performance of the energetic device (unit: Pa·m³ / s). Poor sealing can lead to the intrusion of external moisture, accelerating the degradation reactions such as hydrolysis of the internal energetic material.

[0029] It should be noted that, in order to provide effective input for mutual information analysis and ensure the accuracy of key degradation factors, in the specific implementation, the step of obtaining the storage year, storage location, and leakage rate of the energetic device, and using the storage year, storage location, and leakage rate as storage environment parameters, includes: obtaining the storage year, storage location, and leakage rate of the energetic device; classifying the storage location by region based on a preset geographical classification rule to obtain a regional classification result; encoding the storage location according to the regional classification result to obtain an encoded storage location; and using the storage year, leakage rate, and encoded storage location as storage environment parameters.

[0030] In practice, the storage location is further processed into a categorical variable (achieved through 0-1 coding) such as whether it is located in a "coastal area" to characterize the potential impact of corrosive environments such as high humidity and high salinity. Since mutual information analysis requires inputs to be computationally usable variables, the coded storage location becomes a valid discrete variable with clear physical meaning, allowing for reliable mutual information value calculations along with other parameters (such as storage year and leakage rate). Without coding the storage location, this critical environmental factor cannot be included in the quantitative analysis, potentially leading to the omission of key degradation factors.

[0031] It should be explained that the preset geographic classification rule can be a rule that maps location names to whether they are coastal or not. The regional classification results include coastal and non-coastal areas. The coded storage location can refer to a numeric variable with a value of 1 or 0 after processing using 0-1 encoding. For example, the classification result for coastal areas can be encoded as 1, and the classification result for non-coastal areas can be encoded as 0.

[0032] It should be understood that there may be a large number of different storage locations, and direct comparison is meaningless. The coding process places all devices under a uniform environmental stress level for comparison (for example, all coastal areas are coded as 1, considered to have similar high-severity environments). This eliminates analytical bias caused by different location names, allowing data from devices from different warehouses but with the same environmental type to be merged for analysis. This enhances data consistency and comparability, and provides a guarantee for subsequent accurate mutual information analysis.

[0033] Step S20: Use mutual information analysis to identify key degradation factors from the performance parameters and the storage environment parameters.

[0034] It should be noted that mutual information analysis is a statistical method based on information theory used to quantify the degree of interdependence between two variables. Its core is to quantify the amount of information about one variable that can be obtained by observing another, providing a general framework for complex information systems. By quantifying the amount of information transferred between variables, key factors affecting the performance degradation of energetic devices can be identified, providing a basis for lifetime prediction and storage strategies. Its theoretical formula is as follows: ; In the formula, X and Y represent the performance parameter and the storage environment parameter, respectively, and x and y represent the specific values ​​of the performance parameter and the storage environment parameter, respectively. This represents the mutual information value between X and Y. Describes the joint probability distribution. This represents the marginal probability distribution.

[0035] Key degradation factors refer to environmental parameters that have a significant and dominant impact on the performance degradation of energetic devices, as identified through mutual information analysis.

[0036] For example, among multiple storage environment parameters such as storage year, storage location, and leakage rate, analysis revealed that the storage year is the most critical variable driving peak pressure degradation. This means that when constructing predictive models, this factor should be given priority or even considered primarily to simplify the model structure, avoid noise interference from irrelevant variables, and improve the modeling accuracy and robustness under small sample data.

[0037] Understandably, mutual information analysis (MIA) automatically and quantitatively identifies the most critical factors for performance degradation, effectively filtering out interference from insignificant variables, allowing the model to focus on core influencing factors, simplifying the model structure while enhancing its robustness against interference.

[0038] It should be understood that pairwise mutual information values ​​between storage environment parameters and performance parameters can be calculated. Then, statistical significance tests are performed on the calculated mutual information values. Finally, based on the magnitude of the mutual information values ​​and the results of the statistical significance tests, one or more parameters most significantly associated with performance degradation are selected from the storage environment parameters as key degradation factors. For example, mutual information analysis revealed that the storage year is a key degradation factor affecting the peak pressure Pmax. Subsequently, a physical constraint neural network model was constructed using the storage year as the key degradation factor.

[0039] Step S30: Construct a physical constraint neural network model based on the key degradation factor, wherein the physical constraint neural network model includes constraints based on physical knowledge.

[0040] It's important to note that Physically Constrained Neural Networks (PINNs) are advanced machine learning models. Their core meaning lies in integrating known scientific principles describing physical systems (such as laws, equations, and constraints) into the neural network's learning process, rather than solely relying on data-driven approaches. Unlike traditional neural networks, which are essentially "black box" prediction models, PINN guides and restricts the network's learning direction by using physical knowledge as constraints. This ensures that the network's predictions not only fit the limited training data but also conform to fundamental physical laws. Therefore, PINN effectively prevents the model from learning erroneous patterns that violate common sense physics, thereby improving the reliability and generalization ability of predictions.

[0041] It should be understood that physics-based constraints can refer to rules or boundary conditions extracted from the physical mechanisms of degradation in energetic devices and encoded into a neural network model. In this embodiment, physics-based constraints may include initial state constraints and failure boundary constraints.

[0042] It should be noted that initial state constraints refer to the physical conditions that the system must satisfy at the initial moment (usually t=0), that is, the performance state of the energetic device at the initial moment of storage. Failure boundary constraints refer to the physical boundaries that the system cannot violate throughout the entire evolution process, that is, the performance indicators of the energetic device cannot be lower than or higher than a certain physically meaningful critical value.

[0043] In a specific implementation, the step of constructing a physical constraint neural network model based on the key degradation factor includes: constructing a neural network structure with the key degradation factor as input and the target degradation parameter in the performance parameters as output; adding physical constraints to the neural network structure to obtain an initial physical constraint neural network model, wherein the physical constraints include initial state constraints and failure boundary constraints; constructing a target loss function for the initial physical constraint neural network model based on the physical constraints; and training the initial physical constraint neural network model using a training dataset based on the target loss function to obtain a physical constraint neural network model.

[0044] It should be noted that the neural network structure adopts a three-layer fully connected structure (input layer → hidden layer 1 → hidden layer 2 → output layer), and the corresponding mathematical expression is as follows:

[0045] In the formula, y represents the model output and x represents the model input. Represents a neural network function. These are the weight matrices for the first, second, and third layers (output layer) of the neural network, respectively. These are the bias terms for the first, second, and third layers (output layer) of the neural network, respectively. This represents the activation function.

[0046] Input layer: ; Hidden layer 1 (16 neurons): ; Hidden layer 2 (8 neurons): ; Output layer (1 neuron):

[0047] in, This represents the activation function. This is the output of the first layer. This is the output result of the second layer. This is the output of the third layer, i.e., the model output.

[0048] It should be noted that the initial pressure constraint forces the output of the initial pressure value at t=0 (e.g., 17.804 MPa). ; Failure boundary constraint: The penalty prediction value is lower than the failure threshold (e.g., 14.24 MPa). ; Where λ is the constraint strength coefficient, for example, λ=1.0, For the initial loss, For failure boundary loss, This represents the predicted output value of the neural network model when the input is t.

[0049] By combining data fitting error and physical constraints, a target loss function is constructed. : ; In the formula, N represents the total number of samples (data points) in the training dataset; Indicates at a point in time The actual measured value, that is, the value obtained through destructive testing during storage time. Peak pressure of the energetic device; This represents the model prediction value for the i-th sample; and These are the weight coefficients for the initial state constraints and the failure boundary constraints, respectively. This represents the L2 regularization term, also known as weight decay.

[0050] It should be noted that the L2 regularization term can prevent the model from overfitting to the training data during training (i.e., the model becomes too complex, memorizes data noise, and loses its generalization ability). It forces the model to remain relatively simple by penalizing larger model weight values ​​W. β is a hyperparameter of the regularization strength.

[0051] It should be added that, in the process of training the initial physical constraint neural network model using the training dataset, the training optimization method for obtaining the physical constraint neural network model is to update the weights using the Adaptive Moment Estimation (Adam) optimizer: ; In the formula, , , and These are hyperparameters, for example, =0.01, =0.9, =0.999, ; Represents the target loss function The gradient; and These are the first-moment estimates (i.e., the exponential moving average of the gradient) at time t-1 (the previous iteration) and time t (the current iteration), respectively. and These are the second-moment estimates (i.e., the exponential moving average of the squared gradient) for time t-1 and time t, respectively. and These are the model weight parameters at time t-1 (before the update) and time t (after the update), respectively.

[0052] Physically constrained neural network models, also known as degradation models, assume that the optimal parameter set is obtained after 500 training iterations. The degradation model is represented as:

[0053] Step S40: Use the physical constraint neural network model to predict the lifetime of the energetic device and obtain the prediction result.

[0054] In practical implementation, multiple performance degradation trajectories can be generated based on a physical constraint neural network model; the average failure time of the energetic device can be determined based on the multiple performance degradation trajectories; the safe storage life of the energetic device can be determined based on the average failure time, and the safe storage life can be used as the prediction result.

[0055] It should be understood that the performance degradation trajectory is a virtual, simulated curve that depicts the predicted path of how key performance characteristics (such as peak pressure) of a single energetic device change over time, from the start of storage until failure.

[0056] Accordingly, mean time to failure (MTBF) can refer to a statistically expected value calculated based on all the aforementioned performance degradation trajectories. It represents the average length of time that these energetic devices can operate normally under simulated conditions. Safe storage lifetime is the time obtained by dividing the MTBF by a safety factor greater than 1; it is a conservative lifetime indicator with a margin of error.

[0057] In practical implementation, a physically constrained neural network (PINN) model is used to predict the lifetime of energetic devices. The PINN model is also validated using real-world data. The introduction of physical constraints transforms the PINN model from a purely data-driven approach to a physics-guided learning model, significantly improving extrapolation reliability. Real-world data validation also reveals that environmental and measurement errors severely impact the fitting performance of traditional models on small datasets. The lightweight model training demonstrates excellent fitting capabilities for small datasets. Through initial pressure constraints and failure boundary constraints, the model strictly adheres to physical facts, resulting in strong applicability.

[0058] This embodiment discloses the acquisition of performance parameters and storage environment parameters of energetic devices; the use of mutual information analysis to identify key degradation factors from these parameters; the construction of a physically constrained neural network model based on these key degradation factors, the model including constraints based on physical knowledge; and the prediction of the lifetime of the energetic device using this model. Because this embodiment uses mutual information analysis to identify key degradation factors from performance parameters and storage environment parameters, and utilizes a physically constrained neural network model to predict the lifetime of the energetic device, it effectively improves the accuracy and reliability of lifetime prediction for energetic devices compared to existing technologies.

[0059] refer to Figure 3 , Figure 3 This is a flowchart illustrating the second embodiment of the energetic device lifetime prediction method of the present invention.

[0060] Based on the first embodiment described above, in this embodiment, step S20 includes steps S201 to S203: Step S201: Determine the mutual information value between the performance parameter and the storage environment parameter based on the nonparametric method of K-nearest neighbor estimation.

[0061] Step S202: Perform a significance test on the mutual information value to obtain the significance test result.

[0062] Step S203: Identify key degradation factors from the performance parameters and the storage environment parameters based on the mutual information value and the significance test result.

[0063] In practical implementation, storage location, storage year, and leakage rate are used as influencing factors, while pressure build-up time, peak pressure time, and peak pressure are used as dependent variables. The degree of influence on parameter sensitivity is analyzed using mutual information analysis. The mutual information (MI) value between the performance parameters and the storage environment parameters can be determined based on the nonparametric method of K-nearest neighbor estimation. The nonparametric method of K-nearest neighbor estimation can capture any form of dependency (linear or nonlinear) without assumptions, ensuring that no potentially important correlations are overlooked.

[0064] It should be explained that, due to the random fluctuations in the data itself, measurement errors, or chance, the calculated mutual information value may be greater than 0 because of two completely unrelated variables. Without significance testing, an insignificant variable might be mistakenly treated as a key degradation factor, leading to noise in the subsequent model and inaccurate predictions.

[0065] To further improve the accuracy of identifying key degradation factors, the step of performing a significance test on the mutual information value to obtain a significance test result includes: constructing a hypothesis test based on the mutual information value; correcting the test result of the hypothesis test using a multiple hypothesis test correction method to obtain a corrected test result; comparing the corrected test result (p-value) with a significance level threshold (e.g., α = 0.05), and determining the significance test result of the mutual information value based on the comparison result.

[0066] For example, if the p-value < α (0.05) means that the probability of the event occurring by chance is too small (e.g., p = 0.01, i.e., a 1% probability), which is below the significance level threshold (0.05), then it is considered that such a low-probability event is unlikely to occur. Therefore, there is strong enough evidence to reject the null hypothesis that they are unrelated, and the significance test result is that the association is significant.

[0067] For example, if the p-value ≥ α (0.05) means that the probability of the accidental occurrence is not small enough (e.g., p = 0.1, i.e., a 10% probability), and is still within the acceptable range of chance, then there is not enough strong evidence to prove that they are related, and the significance test result is that the association is not significant.

[0068] It should be understood that the significance level threshold is a rigorous quality control standard used for decision-making. It ensures that the correlation between all variables identified as key degradation factors and performance is not random or accidental, but rather a genuine correlation with high statistical confidence. This provides a reliable scientific basis for the subsequent construction of physically constrained neural network models, avoiding the introduction of noise and random factors into the model, thereby fundamentally improving the accuracy and reliability of the entire energetic device lifetime prediction method.

[0069] It should be noted that the Bonferroni correction method, by adjusting the judgment criteria (such as making the threshold of the p-value more stringent), significantly reduces the risk of misjudging accidental coincidences as real associations. This makes the conclusions of the key degradation factors that are finally identified highly credible and avoids contaminating the subsequent physical constraint neural network model due to the misinclusion of irrelevant variables.

[0070] For example, refer to Figure 4 , Figure 4 This is a schematic diagram of the mutual information analysis matrix between variables in the energetic device lifetime prediction method of this invention. Mutual information analysis is used to analyze the mutual information of six indicators: leakage rate, sampling location (i.e., storage location), year (i.e., storage year), pressure build-up time, peak pressure time, and peak pressure. The mutual information matrix diagram is generated using MATLAB, as shown below. Figure 4As shown in the figure, the key focus should be on the correlation between storage environment parameters (storage year, leakage rate logleak, storage location iscoastal) and performance parameters (build-up time td, peak pressure time tf, peak pressure pmax) (as shown in the 3×3 matrix in the lower left corner of the figure). The results show that the mutual information value between storage year (year) and peak pressure pmax is the highest and statistically significant (MI = 0.83, p < 0.05, Bonferroni correction), indicating that storage year is the dominant factor in peak pressure degradation. Although the mutual information value between leakage rate logleak and build-up time td is also relatively high (MI = 1.02, p < 0.05), considering that the measured variation range of build-up time td is extremely small (≤ 0.04 ms) and its impact on product performance is negligible in engineering, it was excluded in subsequent modeling. Due to the complex and difficult-to-trace nature of the storage environment history, and the low and insignificant (p>0.05) mutual information values ​​between the storage location (iscoastal) and the pressure build-up time (td), peak pressure time (tf), and peak pressure (pmax) (0.10, 0.17, 0.21), the data indicates that within the data range of this invention, the storage location (high humidity / high salinity) has no significant impact on the degradation of key performance indicators of this energetic device. Based on mutual information analysis, the storage year was identified as a key degradation factor, and the storage year (year) was selected as the independent variable, with the peak pressure (pmax) as the dependent variable for modeling research.

[0071] This embodiment discloses a nonparametric method based on K-nearest neighbor estimation to determine the mutual information value between the performance parameters and the storage environment parameters; performs a significance test on the mutual information value to obtain the significance test result; and identifies key degradation factors from the performance parameters and the storage environment parameters based on the mutual information value and the significance test result. Because this embodiment uses a nonparametric method based on K-nearest neighbor estimation to determine the mutual information value between the performance parameters and the storage environment parameters, then performs a significance test on the mutual information value, and finally identifies key degradation factors based on the mutual information value and the significance test result, compared with the prior art, this embodiment effectively removes redundant and interfering variables, and enhances the robustness and generalization ability of the physically constrained neural network model.

[0072] refer to Figure 5 , Figure 5 This is a flowchart illustrating the third embodiment of the energetic device lifetime prediction method of the present invention.

[0073] Based on the above embodiments, in this embodiment, step S40 includes steps S401 to S403: Step S401: Based on the physical constraint neural network model, generate multiple performance degradation trajectories using the Monte Carlo simulation method.

[0074] Step S402: Determine the average failure time of the energetic device based on the multiple performance degradation trajectories.

[0075] Step S403: Determine the safe storage lifetime of the energetic device based on the mean failure time, and use the safe storage lifetime as the prediction result.

[0076] It should be understood that Monte Carlo simulation is a mathematical calculation method that obtains numerical results through repeated random sampling. Its core idea is to simulate the possible behavior of a complex system through a large number of random experiments, thereby estimating the corresponding indicators. It quantifies the uncertainty of the prediction, allowing us to see how the lifetime of energetic devices will be distributed under all possible conditions, rather than just a single, deterministic value.

[0077] For example, Monte Carlo simulation can be used to inject random noise conforming to a predetermined distribution into the input of a physical constraint neural network model; then, based on the input after the noise injection, multiple performance degradation sequences are generated through the physical constraint neural network model; each performance degradation sequence is then treated as a performance degradation trajectory; finally, for each performance degradation trajectory, the time point at which its performance parameter first falls below a preset failure threshold is determined as a single failure time; the single failure times of all performance degradation trajectories are statistically processed to obtain the average failure time.

[0078] In the specific implementation, the failure time is defined. With reliability R(t): ; ; Then, multiple performance degradation trajectories (e.g., 1000 performance degradation trajectories) are generated based on Monte Carlo simulation, and finally, the mean time to failure and safe storage lifetime are calculated. Mean time to failure is: ; The safe storage life is: .

[0079] In the formula, The safety factor is used. In practical implementation, the mean time to failure, safe storage life, and reliability R(t) can be used as the final prediction results.

[0080] It needs to be explained that reliability R(t) represents the probability that an energetic device will still be able to work normally from the start of storage to a specified time t. The calculation of reliability R(t) transforms the physical constraint neural network model and Monte Carlo simulation into a precise language about risk and reliability that engineers and managers can directly understand and use to guide practice, thus achieving precise quantification of risk.

[0081] This embodiment discloses a method for generating multiple performance degradation trajectories using Monte Carlo simulation based on the physical constraint neural network model; determining the mean time to failure (MTBF) of the energetic device based on these trajectories; determining the safe storage lifetime of the energetic device based on the MTBF; and using the safe storage lifetime as the prediction result. Because this embodiment uses Monte Carlo simulation to generate multiple performance degradation trajectories and determines the MTBF of the energetic device based on these trajectories, it improves the scientific validity and reliability of the prediction results compared to existing technologies.

[0082] Furthermore, this embodiment of the invention also proposes a storage medium storing a lifetime prediction program for energetic devices, wherein when the lifetime prediction program for energetic devices is executed by a processor, it implements the steps of the lifetime prediction method for energetic devices as described above.

[0083] Reference Figure 6 , Figure 6 This is a structural block diagram of the first embodiment of the energetic device lifetime prediction device of the present invention.

[0084] like Figure 6 As shown, the energetic device lifetime prediction device proposed in this embodiment of the invention includes: a parameter acquisition module 501, a parameter analysis module 502, a model construction module 503, and a lifetime prediction module 504.

[0085] The parameter acquisition module 501 is used to acquire the performance parameters and storage environment parameters of the energetic device.

[0086] The parameter analysis module 502 is used to identify key degradation factors from the performance parameters and the storage environment parameters by employing mutual information analysis.

[0087] The model building module 503 is used to build a physical constraint neural network model based on the key degradation factor, and the physical constraint neural network model includes constraints based on physical knowledge.

[0088] The lifetime prediction module 504 is used to predict the lifetime of the energetic device using the physical constraint neural network model and obtain the prediction result.

[0089] The parameter acquisition module 501 is further configured to acquire the pressure-time curve of the energetic device through pressure testing; extract pressure-related parameters and time-related parameters from the pressure-time curve, wherein the pressure-related parameters include peak pressure and the time-related parameters include pressure build-up time and peak pressure time; use the peak pressure, pressure build-up time and peak pressure time as performance parameters; acquire the storage year, storage location and leakage rate of the energetic device, and use the storage year, storage location and leakage rate as storage environment parameters.

[0090] The parameter acquisition module 501 is further configured to acquire the storage year, storage location, and leakage rate of the energetic device; classify the storage location into regions based on preset geographical classification rules to obtain regional classification results; encode the storage location according to the regional classification results to obtain coded storage locations; and use the storage year, leakage rate, and coded storage location as storage environment parameters.

[0091] The model building module 503 is further configured to construct a neural network structure with the key degradation factor as input and the target degradation parameter among the performance parameters as output; add physical constraints to the neural network structure to obtain an initial physical constraint neural network model, wherein the physical constraints include initial state constraints and failure boundary constraints; construct a target loss function for the initial physical constraint neural network model based on the physical constraints; and train the initial physical constraint neural network model using a training dataset based on the target loss function to obtain a physical constraint neural network model.

[0092] This device embodiment discloses the acquisition of performance parameters and storage environment parameters of energetic devices; the use of mutual information analysis to identify key degradation factors from the performance parameters and storage environment parameters; the construction of a physically constrained neural network model based on the key degradation factors, the physically constrained neural network model including constraints based on physical knowledge; and the prediction of the lifetime of the energetic device using the physically constrained neural network model to obtain prediction results. Because this device embodiment uses mutual information analysis to identify key degradation factors from performance parameters and storage environment parameters, and uses a physically constrained neural network model to predict the lifetime of energetic devices to obtain prediction results, compared with the prior art, this device embodiment effectively improves the accuracy and reliability of lifetime prediction for energetic devices.

[0093] Based on the first embodiment of the energetic device lifetime prediction device of the present invention, a second embodiment of the energetic device lifetime prediction device of the present invention is proposed.

[0094] In this embodiment, the parameter analysis module 502 is further configured to determine the mutual information value between the performance parameter and the storage environment parameter using a nonparametric method based on K-nearest neighbor estimation; perform a significance test on the mutual information value to obtain a significance test result; and identify key degradation factors from the performance parameter and the storage environment parameter based on the mutual information value and the significance test result.

[0095] The parameter analysis module 502 is further configured to construct a hypothesis test based on the mutual information value; correct the test result of the hypothesis test using the multiple hypothesis test correction method to obtain the corrected test result; compare the corrected test result with the significance level threshold, and determine the significance test result of the mutual information value based on the comparison result.

[0096] Other embodiments or specific implementations of the energetic device lifetime prediction device of the present invention can be referred to the above-described method embodiments, and will not be repeated here.

[0097] This application provides a device for predicting the lifetime of a powered device. The device includes: at least one processor; and a memory communicatively connected to the at least one processor. The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the device lifetime prediction method in the above embodiment 1.

[0098] The following is for reference. Figure 7 The diagram illustrates a structural schematic suitable for implementing the energetic device lifetime prediction device in the embodiments of this application. The energetic device lifetime prediction device in the embodiments of this application may include, but is not limited to, mobile terminals such as mobile phones, laptops, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Description), PMPs (Portable Media Players), in-vehicle terminals (e.g., in-vehicle navigation terminals), and fixed terminals such as digital TVs and desktop computers. Figure 7 The illustrated device for predicting the lifetime of energetic devices is merely an example and should not be construed as limiting the functionality and scope of the embodiments of this application.

[0099] like Figure 7As shown, the device lifetime prediction equipment may include a processing unit 1001 (e.g., a central processing unit, a graphics processing unit, etc.) that can perform various appropriate actions and processes according to a program stored in a read-only memory 1002 or a program loaded from a storage device 1003 into a random access memory 1004. The random access memory 1004 also stores various programs and data required for the operation of the device lifetime prediction equipment. The processing unit 1001, the read-only memory 1002, and the random access memory 1004 are interconnected via a bus 1005. An input / output interface 1006 is also connected to the bus. Typically, the following systems can be connected to the input / output interface 1006: input devices 1007 including, for example, touchscreens, touchpads, keyboards, mice, image sensors, microphones, accelerometers, gyroscopes, etc.; output devices 1008 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 1003 including, for example, magnetic tapes, hard disks, etc.; and communication devices 1009. The communication device 1009 allows the energetic device lifetime prediction device to communicate wirelessly or wiredly with other devices to exchange data. Although energetic device lifetime prediction devices with various systems are shown in the figures, it should be understood that implementation or possession of all the systems shown is not required. More or fewer systems may be implemented alternatively.

[0100] Specifically, according to the embodiments disclosed in this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments disclosed in this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device, or installed from storage device 1003, or installed from read-only memory 1002. When the computer program is executed by processing device 1001, it performs the functions defined in the methods of the embodiments disclosed in this application.

[0101] The energetic device lifetime prediction device provided in this application, employing the energetic device lifetime prediction method in the above embodiments, can solve the technical problem of low accuracy and reliability in energetic device lifetime prediction in the prior art. Compared with the prior art, the beneficial effects of the energetic device lifetime prediction device provided in this application are the same as those of the energetic device lifetime prediction method provided in the above embodiments, and other technical features in this energetic device lifetime prediction device are the same as those disclosed in the method of the previous embodiment, and will not be repeated here.

[0102] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.

[0103] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0104] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or system that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or system. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or system that includes that element.

[0105] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0106] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as read-only memory / random access memory, magnetic disk, optical disk) and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0107] The above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Any equivalent structural or procedural transformations made based on the description and drawings of the present invention, or direct or indirect applications in other related technical fields, are similarly included within the protection scope of the present invention.

Claims

1. A method for predicting the lifetime of energetic devices, characterized in that, The method includes: Obtain the performance parameters and storage environment parameters of energetic devices; Key degradation factors are identified from the performance parameters and the storage environment parameters using mutual information analysis; A physical constraint neural network model is constructed based on the aforementioned key degradation factors, and the physical constraint neural network model includes constraints based on physical knowledge. The lifetime of the energetic device is predicted using the physical constraint neural network model, and the prediction results are obtained.

2. The method for predicting the lifetime of energetic devices as described in claim 1, characterized in that, The step of identifying key degradation factors from the performance parameters and the storage environment parameters using mutual information analysis includes: The mutual information value between the performance parameter and the storage environment parameter is determined using a nonparametric method based on K-nearest neighbor estimation. The mutual information values ​​are subjected to a significance test to obtain the significance test results; Key degradation factors are identified from the performance parameters and the storage environment parameters based on the mutual information values ​​and the significance test results.

3. The method for predicting the lifetime of energetic devices as described in claim 2, characterized in that, The step of performing a significance test on the mutual information value to obtain the significance test result includes: Based on the mutual information values, a hypothesis test is constructed; The test results of the hypothesis test are corrected using the multiple hypothesis test correction method to obtain the corrected test results; The corrected test result is compared with the significance level threshold, and the significance test result of the mutual information value is determined based on the comparison result.

4. The method for predicting the lifetime of energetic devices as described in claim 1, characterized in that, The step of constructing a physically constrained neural network model based on the key degradation factor includes: Construct a neural network structure that takes the key degradation factor as input and the target degradation parameter among the performance parameters as output; Physical constraints are added to the neural network structure to obtain an initial physical constraint neural network model, wherein the physical constraints include initial state constraints and failure boundary constraints; Construct the target loss function of the initial physical constraint neural network model based on the physical constraints; Based on the target loss function, the initial physical constraint neural network model is trained using the training dataset to obtain the physical constraint neural network model.

5. The method for predicting the lifetime of energetic devices as described in claim 1, characterized in that, The step of predicting the lifetime of the energetic device using the physical constraint neural network model and obtaining the prediction result includes: Based on the physical constraint neural network model, multiple performance degradation trajectories are generated using the Monte Carlo simulation method. The average failure time of the energetic device is determined based on the multiple performance degradation trajectories. The safe storage lifetime of the energetic device is determined based on the mean failure time, and the safe storage lifetime is used as the prediction result.

6. The method for predicting the lifetime of energetic devices as described in claim 1, characterized in that, The steps for obtaining the performance parameters and storage environment parameters of the energetic device include: The pressure-time curve of the energetic device is obtained through pressure testing; Pressure-related parameters and time-related parameters are extracted from the pressure-time curve. The pressure-related parameters include peak pressure, and the time-related parameters include pressure build-up time and peak pressure time. The peak pressure, the pressure build-up time, and the peak pressure time are used as performance parameters. The storage year, storage location, and leakage rate of the energetic device are obtained, and the storage year, storage location, and leakage rate are used as storage environment parameters.

7. The method for predicting the lifetime of energetic devices as described in claim 6, characterized in that, The step of obtaining the storage year, storage location, and leakage rate of the energetic device, and using the storage year, storage location, and leakage rate as storage environment parameters, includes: Obtain the storage year, storage location, and leakage rate of the energetic device; The storage locations are classified by region based on preset geographical classification rules to obtain the region classification results; The storage locations are coded according to the regional classification results to obtain coded storage locations; The storage year, the leakage rate, and the coded storage location are used as storage environment parameters.

8. A device for predicting the lifetime of energetic devices, characterized in that, The device includes: The parameter acquisition module is used to acquire the performance parameters and storage environment parameters of the energetic devices; A parameter analysis module is used to identify key degradation factors from the performance parameters and the storage environment parameters using mutual information analysis; The model building module is used to build a physical constraint neural network model based on the key degradation factor, wherein the physical constraint neural network model includes constraints based on physical knowledge. The lifetime prediction module is used to predict the lifetime of the energetic device using the physical constraint neural network model and obtain the prediction result.

9. A device for predicting the lifetime of energetic devices, characterized in that, The device includes: a memory, a processor, and an energetic device lifetime prediction program stored in the memory and executable on the processor, the energetic device lifetime prediction program being configured to implement the steps of the energetic device lifetime prediction method as described in any one of claims 1 to 7.

10. A storage medium, characterized in that, The storage medium stores a lifetime prediction program for energetic devices, which, when executed by a processor, implements the steps of the lifetime prediction method for energetic devices as described in any one of claims 1 to 7.

Images