Method and system for constructing life prediction model of electric power microsensor

Abstract

The invention provides an electric power micro sensor life prediction model construction method and system, and the method comprises the steps: determining a plurality of acceleration test stages and acceleration factor combinations corresponding to the acceleration test stages based on the obtained historical degradation state information of a sensor; and carrying out an acceleration test on the sensor based on the acceleration factor combination corresponding to the plurality of acceleration test stages to obtain an acceleration test data set, and training the life prediction model by taking the acceleration test data set as an input and taking the residual life of the sensor as an output to obtain a trained life prediction model. The complex working condition of the electric power microsensor in the actual working environment is simulated more truly in a multi-stage multi-acceleration-factor combination mode, interference and mutual influence between different failure mechanisms are avoided, and the accuracy and comprehensiveness of life evaluation are improved.

Description

Technical Field

[0001] The present invention relates to the field of power micro sensors, and in particular to a method and system for constructing a life prediction model for a power micro sensor. Background Art

[0002] With the continued growth of energy demand and the continuous upgrading of power systems, the level of intelligence and automation in power systems has significantly increased. Against this backdrop, power microsensors, as a crucial component in key scenarios such as operation monitoring, fault diagnosis, and power management, are increasingly expanding in scope and importance. Power microsensors offer significant advantages such as small size, fast response, low energy consumption, and high integration. They hold broad application prospects in smart substations, power distribution automation systems, distributed energy management platforms, electric vehicle charging infrastructure, and high-voltage transmission monitoring equipment. However, in practical applications, power microsensors face complex and changing environmental conditions, including the combined influence of high and low temperatures, humidity fluctuations, mechanical vibration, electromagnetic interference, and other factors. These complex environmental conditions pose significant challenges to the performance stability and service life of the sensors.

[0003] At present, in view of the complexity of the power system environment, the existing life assessment still has the following three shortcomings:

[0004] 1) The model cannot fully reflect the complex environmental characteristics of the power system. Traditional life prediction methods are mostly based on the simple superposition of environmental factors, such as the influence of a single or a few factors such as temperature and humidity. Although existing machine learning methods can process multi-factor data, because their models are too dependent on data training sets, they often fail to consider the complex coupling between multiple factors, resulting in insufficient accuracy in life prediction. Especially in environments with multiple factors such as high and low temperature cycles, moisture-heat coupling, and electromagnetic interference, existing models are often unable to effectively simulate the nonlinear degradation process of sensors, resulting in large deviations in prediction results and an inability to provide accurate guidance for sensor selection, maintenance, and replacement.

[0005] 2) Limitations of accelerated test methods. Most current accelerated test methods are designed based on the characteristics of traditional sensors and fail to optimize the unique structure and operating principle of power microsensors. Innovations in materials and structures of power microsensors, such as the use of MEMS technology and nanomaterials, have increased their sensitivity to tiny environmental changes. Therefore, traditional extreme test conditions such as high temperature and overvoltage often cause nonlinear aging of materials or generate additional failure modes, thereby distorting the test results. This test method with an excessively high acceleration ratio makes it difficult to truly reflect the service life of the sensor under actual working conditions, affecting the accuracy and effectiveness of the life assessment.

[0006] 3) Lack of dynamically adjustable test conditions. Existing accelerated test methods typically use fixed acceleration conditions, making it difficult to simulate the dynamic environmental changes that power microsensors experience in actual use. Accelerated test conditions are overly simple and fail to optimize for the dynamic changes that sensors may experience in complex and changing operating environments. As a result, test results often fail to accurately reflect the actual performance of the sensor, which in turn affects the reliability of lifespan assessments.

[0007] Therefore, existing lifespan assessment methods have significant deficiencies in model accuracy, test specificity, and result authenticity. This is especially true for the unique structure and operating conditions of power microsensors, where existing technologies are limited in effectiveness. Therefore, a comprehensive lifespan assessment method is urgently needed to accurately assess the lifespan of power microsensors. Summary of the Invention

[0008] In order to solve the problem that the existing technology has limitations in accelerated testing of power micro sensors and lacks dynamically adjusted test conditions, the present invention proposes a method and system for constructing a life prediction model for power micro sensors.

[0009] In a first aspect, a method for constructing a life prediction model for a power microsensor is provided, comprising:

[0010] Determining multiple accelerated test stages and acceleration factor combinations corresponding to each accelerated test stage based on the acquired historical degradation state information of the sensor;

[0011] performing an accelerated test on the sensor based on a combination of acceleration factors corresponding to a plurality of accelerated test stages to obtain an accelerated test data set;

[0012] The life prediction model is trained with the accelerated test data set as input and the remaining life of the sensor as output to obtain a trained life prediction model, wherein the remaining life of the sensor is calculated based on the accelerated test data set.

[0013] Preferably, the determining of a plurality of accelerated test stages and a combination of acceleration factors corresponding to each accelerated test stage based on the acquired historical degradation state information of the sensor includes:

[0014] Setting up multiple accelerated test stages based on the acquired historical degradation status information of the sensor;

[0015] Based on the historical degradation state information of the sensor, the acceleration factor dynamic function is used to configure the initial acceleration factor combination corresponding to multiple acceleration test stages;

[0016] An objective function is constructed with the goal of minimizing the accelerated test time, and the objective function is solved using a preset model solving method to obtain configuration information of the acceleration factor combination corresponding to each accelerated test stage;

[0017] The acceleration factor combination corresponding to each accelerated test stage is generated based on the configuration information of the acceleration factor combination corresponding to each accelerated test stage and the initial acceleration factor combination.

[0018] Preferably, the calculation formula of the acceleration factor dynamic function is as follows:

[0019]

[0020] Among them, Φ(k) represents the dynamic function of the acceleration factor of stage k, Indicates the acceleration factor corresponding to the temperature factor of the kth stage, Indicates the acceleration factor corresponding to the humidity factor of the kth stage, represents the acceleration factor corresponding to the mechanical stress factor of stage k, Indicates the acceleration factor corresponding to the electromagnetic interference factor in the kth stage.

[0021] Preferably, the calculation formula of the acceleration factor corresponding to the temperature factor is as follows:

[0022]

[0023] Among them, E a is the activation energy, k B is the Boltzmann constant, T is the absolute temperature, T acc Indicates the absolute temperature of the accelerated test conditions, T norm Indicates the absolute temperature under normal test conditions, AF T Indicates the acceleration factor of temperature factor.

[0024] Preferably, the calculation formula of the acceleration factor corresponding to the humidity factor is as follows:

[0025] AF RH =exp[β(RH acc -RH norm )]

[0026] Among them, β is the humidity sensitivity coefficient, RH is the relative humidity, RH acc Indicates the relative humidity of the accelerated test conditions, RH norm Indicates the relative humidity under normal test conditions, AF RH Indicates the acceleration factor of humidity.

[0027] Preferably, the calculation formula of the acceleration factor corresponding to the mechanical stress factor is as follows:

[0028]

[0029] Among them, Δε accIndicates accelerated test conditions, Δε norm represents normal working conditions, p represents a constant related to the material, AF M Indicates the acceleration factor of the mechanical stress factor.

[0030] Preferably, the calculation formula of the acceleration factor corresponding to the electromagnetic interference factor is as follows:

[0031]

[0032] Among them, E acc represents the electromagnetic stress under accelerated conditions, E norm represents the electromagnetic stress under normal conditions, m is an index related to the material aging mechanism, AF EM Indicates the acceleration factor corresponding to the electromagnetic interference stress factor.

[0033] Preferably, the calculation formula of the objective function is as follows:

[0034]

[0035] Among them, t k is the duration of the k-th stage experiment, P modlei It represents the probability that the i-th failure mode is stimulated to a certain confidence level under the test design, δ is the pre-set threshold requirement, K is the total number of test stages, and Φ(k) represents the dynamic function of the acceleration factor of stage k.

[0036] Preferably, performing an accelerated test on the sensor based on a combination of acceleration factors corresponding to a plurality of accelerated test stages to obtain an accelerated test data set includes:

[0037] adjusting an accelerated test environment corresponding to each accelerated test stage based on acceleration factors in different acceleration factor combinations corresponding to the multiple accelerated test stages, and performing an accelerated test on the sensor based on the adjusted accelerated test environment in each accelerated test stage to obtain degradation status information of the sensor in each accelerated test stage;

[0038] An accelerated test data set is generated based on the acquired environmental factor information, acceleration factor combination, and sensor degradation status information corresponding to each accelerated test stage.

[0039] Preferably, the training process of the lifespan prediction model includes:

[0040] The environmental factor information, sensor degradation status information, and acceleration factor of each accelerated test stage in the accelerated test data set are combined into a feature matrix as input, and the environmental factor information of each accelerated test stage in the accelerated test data and the sensor degradation status information corresponding to the acceleration factor are converted into remaining life information as output. Random forest is used to train the life prediction model to generate a trained life prediction model.

[0041] Preferably, after obtaining the trained lifespan prediction model, the method further includes:

[0042] Collect the degradation status information of the sensor to be tested and the current environmental factors information;

[0043] The degradation state information and the current environmental factor information are input into the trained life prediction model to obtain the remaining life prediction information of the sensor to be tested.

[0044] In a second aspect, the present invention provides a system for constructing a life prediction model for a power microsensor, comprising:

[0045] A determination module, configured to determine a plurality of accelerated test stages and a combination of acceleration factors corresponding to each accelerated test stage based on the acquired historical degradation state information of the sensor;

[0046] an acceleration module, configured to perform an acceleration test on the sensor based on a combination of acceleration factors corresponding to a plurality of acceleration test stages to obtain an acceleration test data set;

[0047] The training module is used to train the life prediction model using the accelerated test data set as input and the remaining life of the sensor as output to obtain a trained life prediction model, wherein the remaining life of the sensor is calculated based on the accelerated test data set.

[0048] Preferably, the determining module is further configured to:

[0049] Setting a plurality of accelerated test stages and accelerated test end conditions of each accelerated test stage based on the acquired historical degradation state information of the sensor;

[0050] Based on the historical degradation state information of the sensor, the acceleration factor dynamic function is used to configure the initial acceleration factor combination corresponding to multiple acceleration test stages;

[0051] An objective function is constructed with the goal of minimizing the accelerated test time, and the objective function is solved using a preset model solving method to obtain configuration information of the acceleration factor combination corresponding to each accelerated test stage;

[0052] The acceleration factor combination corresponding to each accelerated test stage is generated based on the configuration information of the acceleration factor combination corresponding to each accelerated test stage and the initial acceleration factor combination.

[0053] Preferably, the calculation formula of the acceleration factor dynamic function in the determination module is as follows:

[0054]

[0055] Among them, Φ(k) represents the dynamic function of the acceleration factor of stage k, Indicates the acceleration factor corresponding to the temperature factor of the kth stage, Indicates the acceleration factor corresponding to the humidity factor of the kth stage, represents the acceleration factor corresponding to the mechanical stress factor of stage k, Indicates the acceleration factor corresponding to the electromagnetic interference factor in the kth stage.

[0056] Preferably, the calculation formula of the acceleration factor corresponding to the temperature factor in the determination module is as follows:

[0057]

[0058] Among them, E a is the activation energy, k B is the Boltzmann constant, T is the absolute temperature, T acc Indicates the absolute temperature of the accelerated test conditions, T norm Indicates the absolute temperature under normal test conditions, AF T Indicates the acceleration factor of temperature factor.

[0059] Preferably, the calculation formula of the acceleration factor corresponding to the humidity factor in the determination module is as follows:

[0060] AF RH =exp[β(RH acc -RH norm )]

[0061] Among them, β is the humidity sensitivity coefficient, RH is the relative humidity, RH acc Indicates the relative humidity of the accelerated test conditions, RH norm Indicates the relative humidity under normal test conditions, AF RH Indicates the acceleration factor of humidity.

[0062] Preferably, the calculation formula of the acceleration factor corresponding to the mechanical stress factor in the determination module is as follows:

[0063]

[0064] Among them, Δε accIndicates accelerated test conditions, Δε norm represents normal working conditions, p represents a constant related to the material, AF M Indicates the acceleration factor of the mechanical stress factor.

[0065] Preferably, the calculation formula of the acceleration factor corresponding to the electromagnetic interference factor in the determination module is as follows:

[0066]

[0067] Among them, E acc represents the electromagnetic stress under accelerated conditions, E norm represents the electromagnetic stress under normal conditions, m is an index related to the material aging mechanism, AF EM Indicates the acceleration factor corresponding to the electromagnetic interference stress factor.

[0068] Preferably, the calculation formula of the objective function in the determination module is as follows:

[0069]

[0070] Among them, t k is the duration of the k-th stage experiment, P modlei It represents the probability that the i-th failure mode is stimulated to a certain confidence level under the test design, δ is the pre-set threshold requirement, K is the total number of test stages, and Φ(k) represents the dynamic function of the acceleration factor of stage k.

[0071] Preferably, the acceleration module is further configured to:

[0072] adjusting an accelerated test environment corresponding to each accelerated test stage based on acceleration factors in different acceleration factor combinations corresponding to the multiple accelerated test stages, and performing an accelerated test on the sensor based on the adjusted accelerated test environment in each accelerated test stage to obtain degradation status information of the sensor in each accelerated test stage;

[0073] An accelerated test data set is generated based on the acquired environmental factor information, acceleration factor combination, and sensor degradation status information corresponding to each accelerated test stage.

[0074] Preferably, the training process of the life prediction model in the training module includes:

[0075] The environmental factor information, sensor degradation status information, and acceleration factor of each accelerated test stage in the accelerated test data set are combined into a feature matrix as input, and the environmental factor information of each accelerated test stage in the accelerated test data and the sensor degradation status information corresponding to the acceleration factor are converted into remaining life information as output. Random forest is used to train the life prediction model to generate a trained life prediction model.

[0076] Preferably, the system further comprises:

[0077] Model application module, used to collect degradation status information of the sensor to be tested and current environmental factors information;

[0078] The degradation state information and the current environmental factor information are input into the trained life prediction model to obtain the remaining life prediction information of the sensor to be tested.

[0079] In another aspect, the present application further provides an electronic device, comprising: at least one processor and a memory; the memory and the processor are connected via a bus;

[0080] The memory is used to store one or more programs;

[0081] When the one or more programs are executed by the at least one processor, the above-mentioned method for constructing a power microsensor life prediction model is implemented.

[0082] On the other hand, the present application also provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed, it implements the method for constructing a power microsensor life prediction model as described above.

[0083] Compared with the prior art, the present invention has the following beneficial effects:

[0084] The present invention provides a method and system for constructing a life prediction model for an electric microsensor. The construction method determines multiple accelerated test stages and acceleration factor combinations corresponding to each accelerated test stage based on the acquired historical degradation state information of the sensor, and performs accelerated tests on the sensor based on the acceleration factor combinations corresponding to the multiple accelerated test stages to obtain an accelerated test data set. The accelerated test data set is then used as input, and a life prediction model is trained with the remaining life of the sensor as output to obtain a trained life prediction model. The trained life prediction model can achieve accurate prediction of the remaining life of the sensor.

[0085] Specifically, a multi-stage and multi-acceleration factor combination is used to more realistically simulate the complex working conditions of power microsensors in actual working environments, avoiding interference and mutual influence between different failure mechanisms, and improving the accuracy and comprehensiveness of life assessment. The construction of a prediction model based on the accelerated test data set of the sensor collected during the staged accelerated test not only improves the accuracy of life prediction, but also enhances the adaptability and flexibility of the model, and can better cope with complex situations in actual applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0086] Figure 1A flowchart of a method for constructing a power microsensor life prediction model according to the present invention;

[0087] Figure 2 A schematic diagram of environmental factor coupling of the method for constructing a power microsensor life prediction model of the present invention;

[0088] Figure 3 A flow chart of a dynamic acceleration ratio optimization strategy for a method for constructing a power microsensor life prediction model according to the present invention;

[0089] Figure 4 A schematic diagram of the random forest life prediction algorithm structure of the power microsensor life prediction model construction method of the present invention;

[0090] Figure 5 This is a schematic diagram of the structure of the power microsensor life prediction model construction system of the present invention;

[0091] Figure 6 The figure is a schematic structural diagram of an electronic device of the present invention. DETAILED DESCRIPTION

[0092] The present invention addresses the key challenges of life evaluation of power microsensors in complex environments and provides a method and system for constructing a life prediction model for power microsensors. Through systematic, multi-dimensional accelerated aging and data acquisition, combined with advanced data analysis and life prediction models, this method comprehensively considers the multi-factor coupling effect of the complex environment of the power system, and combines the dynamic acceleration factor optimization strategy (Dynamic Acceleration Factor Optimization, DAFO) with the data-driven random forest life prediction algorithm. In addition, in terms of the application process and system of this life prediction model, it further provides an experimental and usage process based on the collaborative operation of multiple modules, which can be implemented under laboratory accelerated test conditions and can also be extended to life monitoring and prediction in actual operating conditions, realizing the full life cycle management of sensor life.

[0093] The present invention achieves a high-fidelity assessment of the failure processes that a sensor may experience throughout its life cycle. It can dynamically adjust acceleration conditions based on the simulation of the actual working environment and combine it with advanced data-driven algorithms to accurately assess the life of power microsensors. This will provide strong support for the optimized design of sensors, maintenance strategies and the intelligent development of power systems.

[0094] In order to better understand the present invention, the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0095] Example 1:

[0096] A method for constructing a life prediction model for power micro sensors, such as Figure 1 Shown, including:

[0097] Step 1: Determine multiple accelerated test stages and corresponding acceleration factor combinations for each accelerated test stage based on the acquired historical degradation status information of the sensor;

[0098] Step 2: performing an accelerated test on the sensor based on a combination of acceleration factors corresponding to multiple accelerated test stages to obtain an accelerated test data set;

[0099] Step 3: Using the accelerated test dataset as input and the sensor remaining life as output to train the life prediction model, thereby obtaining a trained life prediction model, wherein the sensor remaining life is calculated based on the accelerated test dataset.

[0100] Among them, the acceleration factor (AF) is used to characterize the proportion of shortened sensor life under accelerated test conditions compared to normal working conditions.

[0101] Life assessment is based on acceleration factors, which convert the observed failure time in accelerated testing into the expected life under normal operating conditions. By calculating the acceleration factor, accelerated test results can be effectively converted into life predictions under actual operating conditions, thereby improving the accuracy and reliability of the predictions. The acceleration factor intuitively reflects the characteristics of the failure model. Its quality depends primarily on its ability to accurately describe the actual failure time. Properly determining the acceleration factor is key to ensuring the effectiveness of life assessment.

[0102] In a specific embodiment, the present invention comprehensively considers the following major environmental factors for the working environment of power microsensors in practical applications: 1) Temperature change: considering the impact of the day-night temperature difference and seasonal temperature fluctuations in the power system on the sensor; 2) Humidity change: considering the seasonal changes in air humidity and its impact on sensor materials and electronic components; 3) Mechanical stress: considering the long-term impact of mechanical stress such as vibration and impact during equipment operation on the sensor; 4) Electromagnetic interference: considering the impact of various types of electromagnetic interference in the power system on sensor signal processing.

[0103] See also Figure 2 , Figure 2 It is a schematic diagram of environmental factor coupling of the method for constructing the power microsensor life prediction model of the present invention.

[0104] (1) Temperature stress

[0105] Temperature is the primary factor causing material stress and performance degradation in microdevices. For microelectronic devices, rising temperatures typically accelerate the diffusion of materials within the device and can lead to the rapid evolution of microstructural defects. The Arrhenius model, commonly used in lifetime modeling, describes the temperature stress acceleration mechanism. The relationship between the failure rate λ and temperature T, λ(T), can generally be expressed as:

[0106]

[0107] Among them, λ0 is a constant related to material properties, E a is the activation energy, k B is the Boltzmann constant, and T is the absolute temperature. For accelerated stress test, when the temperature stress dominates the failure, the acceleration factor AF T It can be expressed as:

[0108]

[0109] Among them, T acc and T norm They represent the absolute temperatures of accelerated test conditions and normal test conditions respectively.

[0110] (2) Humidity stress

[0111] Air humidity significantly affects the insulation properties, material corrosion, and interface contact characteristics of the dielectric inside and outside the microsensor. For microsensors using nanomaterials or composite materials as sensitive elements, changes in humidity can also lead to differences in adsorption / desorption on the material surface, affecting electrical and mechanical properties.

[0112] When the failure mechanism induced by humidity is mainly an activation process, such as a chemical reaction or diffusion process induced by humidity, and the failure rate increases exponentially with humidity, the relationship between the failure rate λ and humidity RH, λ(RH), can usually be written as:

[0113] λ(RH)=λ0exp(βRH) (3)

[0114] Where λ0 is a constant related to material properties, β is the humidity sensitivity coefficient, and RH is the relative humidity.

[0115] At this time, the acceleration factor can be expressed as:

[0116]

[0117] Among them, RH acc and RH norm Represent the relative humidity under accelerated test conditions and normal test conditions respectively.

[0118] When the humidity-induced failure process is closely related to the power relationship of humidity, such as mechanical stress caused by hygroscopic expansion and nonlinear changes in interface contact resistance:

[0119]

[0120] Where m is the humidity acceleration index, RH1 represents the actual ambient relative humidity, and RH2 represents the reference relative humidity.

[0121] At this time, the acceleration factor can be expressed as:

[0122]

[0123] (3) Mechanical stress

[0124] Mechanical stresses such as vibration and shock can cause fatigue and cracks in sensor structures or packaging interfaces. Due to differences in miniaturized design and material properties, microsensors are prone to stress concentration under mechanical stress, leading to a high rate of damage accumulation. The Coffin-Manson model can be used to describe the lifespan of mechanical stress:

[0125] N f =C mech Δε -p (7)

[0126] Among them, N f is the number of cycles to failure, Δε is the strain amplitude, p is a constant related to the material, C mech is a constant related to the material and structure. In the accelerated test, higher intensity or higher frequency vibration is used as the stress source, and the acceleration of mechanical failure is achieved by expanding parameters such as Δε. The acceleration factor can be written as:

[0127]

[0128] Among them, Δε acc Indicates accelerated test conditions; Δε norm Indicates normal operating conditions.

[0129] (4) Electromagnetic interference stress

[0130] Electromagnetic interference can affect both the signal processing of microsensors and the dielectric materials of sensitive components. Device aging can be accelerated by increasing electromagnetic field strength, increasing voltage amplitude, or introducing higher-frequency electromagnetic waveforms. Assuming that device failure primarily occurs through mechanisms such as electrical breakdown, insulation degradation, or charge accumulation, a power model can be used to describe this:

[0131] N EM =C em E -m (9)

[0132] Among them, N EM is the life under electromagnetic interference, C em and m are constants related to materials and structures, and E represents the electric field strength or environmental stress. The acceleration factor can be expressed as:

[0133]

[0134] Among them, E acc and E norm represent the electromagnetic stress under accelerated conditions and normal conditions, respectively, and m is an index related to the material aging mechanism.

[0135] (5) Coupling acceleration factor

[0136] In complex power systems, environmental stresses do not act independently. The lifespan of power microsensors is affected by a variety of environmental stress factors, including temperature, humidity, mechanical stress, and electromagnetic interference. These environmental stress factors do not exist in isolation but rather couple with each other, contributing to the sensor's aging process. To accurately assess sensor lifespan, this paper presents the principle and calculation method of a multi-factor coupled acceleration factor, which characterizes the combined impact of different environmental stresses on sensor failure.

[0137] Through theoretical analysis, failure analysis and preliminary accelerated tests, the main impact paths of the main interference sources on failure and whether there are obvious interactive effects between them are determined.

[0138] If some stresses are “weakly coupled” or “approximately independent”, they can be simplified to a product model. Under the assumption of mutual independence, the total acceleration factor AF can be multi Written as:

[0139] AF multi =AF T AF H AF M AF EM (11)

[0140] In many cases, the contributions of various stresses to failure mechanisms are not linearly separable. For example, rising temperature can change the hygroscopic properties of a material, thereby amplifying the effect of humidity. Mechanical vibration can also affect device interfaces, making them more sensitive to temperature or humidity. Certain destructive mechanisms of electromagnetic interference may be more closely coupled to temperature. To more accurately reflect the coupling effect, a coupling factor can be added to the model. The total acceleration factor can then be written as:

[0141]

[0142] Among them, β T , β H , βM and β EM are the weights or power exponents of the acceleration factors of each factor, Ω(β ij ,…) is the “coupling correction function” obtained based on experimental data or mechanism analysis, which is used to describe the interaction between two or multiple stresses (i, j refers to different stress types).

[0143] Through the above-mentioned multi-factor coupling model (the calculation formula of the acceleration factor of each environmental factor), we can not only more accurately estimate the lifespan impact caused by a single factor or a few factors, but also capture the nonlinear effects caused by the interaction of multiple factors to a certain extent.

[0144] In this embodiment, the process of determining multiple accelerated test stages and acceleration factor combinations corresponding to each accelerated test stage based on the acquired historical degradation state information of the sensor in step 1 includes:

[0145] Setting up multiple accelerated test stages based on the acquired historical degradation status information of the sensor;

[0146] Based on the historical degradation state information of the sensor, the acceleration factor dynamic function is used to configure the initial acceleration factor combination corresponding to multiple acceleration test stages;

[0147] An objective function is constructed with the goal of minimizing the accelerated test time, and the objective function is solved using a preset model solving method to obtain configuration information of the acceleration factor combination corresponding to each accelerated test stage;

[0148] The acceleration factor combination corresponding to each accelerated test stage is generated based on the configuration information of the acceleration factor combination corresponding to each accelerated test stage and the initial acceleration factor combination.

[0149] In complex power systems, power microsensors are exposed to the combined effects of multiple environmental stress factors, and the impact of these stress factors on sensor life is time-dependent and nonlinear. Traditional accelerated tests often fix stress levels such as temperature, humidity, or vibration at the beginning of the test, resulting in a lack of feedback on the real-time degradation status of the sensor in subsequent tests. Once the selected stress is too high, atypical failure modes may be triggered; if the stress is too low, the acceleration effect is not obvious and the test cycle is too long. Moreover, as the test time progresses, the aging process of the sensor will show a non-uniform distribution: in the early stages, certain material defects or interface problems may be more significant; in the later stages, thermal-mechanical fatigue or dielectric damage may gradually dominate.

[0150] In response to this staged aging mechanism, the present invention proposes a dynamic acceleration ratio optimization strategy. According to the characteristics of the sensor and the different attenuation characteristics of various environmental factors, the ratio of the acceleration factor is dynamically adjusted in different test stages to ensure sufficient stimulation of each failure mode.

[0151] Among them, the dynamic acceleration ratio refers to the process of dynamically adjusting the acceleration factor of each environmental stress according to the real-time degradation state of the sensor during the acceleration test to adapt to the failure mechanism requirements of different test stages. By monitoring the degradation state of the sensor or based on a priori mechanism analysis, the acceleration test is divided into several stages (such as early, middle, and late stages), and different acceleration factor combinations are set for each stage to accurately stimulate the main failure modes of each stage, such as Figure 3 shown.

[0152] In a specific embodiment, in the kth test phase, the acceleration factors applied to temperature, humidity, mechanical stress, and electromagnetic interference can be selected as DAFO uses a dynamic function to assign different acceleration factor combinations to each stage:

[0153]

[0154] Where k∈{1,2,3,4,…,K}, K is the total number of test stages, and Φ(k) can be determined through test data feedback (real-time monitoring) or prior knowledge (staged analysis of device failure mechanisms). The optimization objective function can be determined by the following multi-factor comprehensive indicators and can be dynamically adjusted during the test. The optimization goal can be described as:

[0155]

[0156] Among them, t k is the duration of the k-th stage experiment, P modlei represents the probability or characterization index of the i-th failure mode being stimulated to a certain confidence level under this test design, and δ is the pre-set threshold requirement. This means that within the limited test time, multiple failure modes can be fully examined, ultimately obtaining a more realistic comprehensive life evaluation result.

[0157] Different aging processes usually dominate in different time periods. For example, in the early stage, there may be material defects or interface problems, in the middle stage, it may be dominated by chemical diffusion or corrosion, and in the later stage, it may be dominated by fatigue, stress cracking, etc. If the product is pressed to failure only under "a very high stress", it may trigger atypical or even "new" failure mechanisms, making it difficult to map the test results to the lifespan in the actual use environment. Targeted changes in stress levels at different stages (such as high temperature and high humidity first, and then increased mechanical vibration and electromagnetic interference) can more effectively stimulate different failure mechanisms. This not only avoids excessive stress, but also allows the device to experience the main failure processes within a reasonable time, obtaining more comprehensive and accurate lifespan data.

[0158] In this embodiment, in step 2, the process of performing an accelerated test on the sensor based on a combination of acceleration factors corresponding to a plurality of accelerated test stages to obtain an accelerated test data set includes:

[0159] adjusting an accelerated test environment corresponding to each accelerated test stage based on acceleration factors in different acceleration factor combinations corresponding to the multiple accelerated test stages, and performing an accelerated test on the sensor based on the adjusted accelerated test environment in each accelerated test stage to obtain degradation status information of the sensor in each accelerated test stage;

[0160] An accelerated test data set is generated based on the acquired environmental factor information, acceleration factor combination, and sensor degradation status information corresponding to each accelerated test stage.

[0161] In a specific embodiment, the accelerated test can be divided into three stages, each of which forms a corresponding acceleration factor:

[0162] (1) Initial stage (k = 1)

[0163] Intensifying temperature and humidity stresses promotes the rapid evolution of material defects. Focusing on accelerating failure mechanisms most sensitive to temperature and humidity (such as material defects, interface oxidation, and moisture absorption reactions) to observe whether early failures are triggered during this phase. Therefore, high temperature and high humidity are the primary stresses, while mechanical stress and electromagnetic interference are relatively normal or slightly amplified.

[0164] Ω=1

[0165] When Phase 1 reaches the preset test time, or it is detected that the main failure mechanism has changed (interface defect saturation or moisture absorption problem tends to be stable), or the failure rate of some devices increases significantly, it enters a new aging stage.

[0166] The environmental parameters under normal working conditions are:

[0167] Temperature: T norm =25℃=298K

[0168] Humidity: RH norm =50%RH

[0169] Mechanical stress: Δε norm =0.01

[0170] Electromagnetic interference: E norm =1V / m

[0171] The environmental parameters under the initial accelerated test conditions are:

[0172] Temperature: T norm =55℃=328K

[0173] Humidity: RH norm =80%RH

[0174] Mechanical stress: Δε norm =0.02

[0175] Electromagnetic interference: E norm =2V / m

[0176] According to the above formulas (1) to (10), calculate the acceleration factors: (E a =0.7eV=0.7×1.602×10 -19 J, k B =1.381×10 -23 J / K);

[0177] In the initial stage (k = 1), there is no coupling correction factor (Ω = 1), and each stress acts independently. Then, according to formula (11), AF multi =2.58.

[0178] (2) Middle stage (k = 2)

[0179] Weaken temperature and humidity, and instead strengthen mechanical stress and electromagnetic interference to stimulate mid-term failure mechanisms such as "chemical diffusion, corrosion, material fatigue, and insulation performance degradation", gradually reduce the acceleration factors of temperature and humidity, and increase the acceleration factors of mechanical stress and electromagnetic interference. Coupling corrections can also be added to reflect the comprehensive amplification effect of environmental stress.

[0180] Ω≥1

[0181] When Phase 2 reaches the preset test time, or it is detected that failure modes such as chemical corrosion / diffusion are saturated, or a new dominant mechanism (fatigue cracks, insulation degradation, etc.) appears, Phase 2 ends and enters Phase 3.

[0182] In the middle stage (k=2), the temperature acceleration factor is reduced by 80%, Humidity acceleration factor is reduced by 80%, The mechanical stress acceleration factor and electromagnetic interference acceleration factor remain unchanged, and the coupling correction factor Ω=1.2. multi =324.03.

[0183] (3) Late stage (k=3)

[0184] Further weaken the influence of temperature and humidity, greatly enhance mechanical stress and electromagnetic interference, and add stronger coupling corrections. The main focus is on late failure mechanisms such as "thermal-mechanical fatigue, material cracking, dielectric breakdown / aging". Temperature and humidity basically return to normal or acceleration is canceled, while mechanical stress and electromagnetic interference are increased to a higher level. There may be significant coupling terms, and the acceleration factors of each stress are further adjusted to highlight thermal-mechanical fatigue and dielectric damage.

[0185] Ω≥1

[0186] The termination conditions of the later stage are the occurrence of final failure (complete loss of main performance indicators) or the reaching of the maximum test time.

[0187] In the later stage (k=3), the temperature acceleration factor does not accelerate, Humidity acceleration factor does not accelerate, The mechanical stress acceleration factor increases by 20%, The electromagnetic interference acceleration factor increases by 20%, Coupling correction factor Ω=1.5. At this time, AF multi =409.88.

[0188] Table 1 Acceleration factors at each stage

[0189]

[0190] The dynamic acceleration ratio optimization strategy adjusts the acceleration factor for each environmental stress in stages, rationally highlighting different failure mechanisms at different stages of the test. This avoids atypical failures caused by excessive stress, improves acceleration efficiency, and obtains more representative lifespan data. Compared to traditional fixed-stress accelerated testing, this method significantly improves both scientificity and innovation, providing a more efficient and feasible solution for lifespan assessment of high-reliability devices such as power microsensors.

[0191] In this embodiment, before performing sensor life prediction based on the life prediction model in step 3, the life prediction model needs to be trained. The training process of the life prediction model can be as follows:

[0192] The environmental factor information, sensor degradation status information, and acceleration factor of each accelerated test stage in the accelerated test data set are combined into a feature matrix as input, and the environmental factor information of each accelerated test stage in the accelerated test data and the sensor degradation status information corresponding to the acceleration factor are converted into remaining life information as output. Random forest is used to train the life prediction model to generate a trained life prediction model.

[0193] Under the framework of the aforementioned multi-factor coupled acceleration test and dynamic acceleration ratio optimization strategy (DAFO), a large amount of degradation data and environmental stress data are collected at each stage. These data often show the characteristics of high dimension (containing multiple dependent variables such as temperature, humidity, mechanical stress, electromagnetic interference, etc.), nonlinearity and stage (stage-by-stage acceleration). Therefore, the present invention adopts the integrated learning method of random forest to integrate the multi-source information obtained from the acceleration test in a data-driven manner, thereby establishing a high-accuracy life prediction. Random forest has outstanding advantages in resisting overfitting, handling nonlinear characteristics and scalability, and can better cope with the needs of modeling the degradation laws of micro sensors in complex multi-stress environments.

[0194] Specifically, a random forest consists of multiple decision trees. It uses bootstrap sampling and random feature selection to sample the training data multiple times with replacement. Multiple decision trees are then trained independently, and a comprehensive prediction result is ultimately obtained through voting or averaging. Compared to a single decision tree, a random forest significantly reduces the risk of overfitting the training data through its ensemble learning mechanism. It is also more robust to noisy or redundant input features.

[0195] In a specific embodiment, the input features can be degradation indicators collected by the sensor at different stages and the corresponding stress conditions. The environmental stress (T, RH, Δε, E, etc.) collected at each stage is combined with the acceleration factor (AF T , AF RH , AF M , AF EM ) and coupling correction (Ω) are combined into the feature matrix X. The target output can be the remaining life of the sensor or the life span under normal operating conditions. The regression target can be constructed from the degradation / failure time observed in the staged accelerated test (and its conversion to the normal operating life). The corresponding failure time (or remaining life, performance degradation level) is used as the target vector Y.

[0196] Then, the collected multi-stage data are integrated and labeled (actual failure time or stage life indicators), and the training set data is used for multiple training. Each tree will replace some samples from the training set and randomly select some features for node division. During training, the random forest will sample the data multiple times to generate several decision trees, and finally output the life prediction value by voting or averaging. Cross-validation or leave-out validation can be used to evaluate the model's fit and generalization ability, such as Figure 4 shown.

[0197] Random Forest Regression Model:

[0198]

[0199] in, is the predicted output of input feature X, M is the number of decision trees, h m (X) is the predicted value of X by the mth decision tree.

[0200] Random forests improve model performance through two types of randomness: 1) For each decision tree, a subset of the original training set is randomly sampled with replacement as training data; 2) At each node split, a random subset of features (typically the square root of the total number of features) is evaluated, and the optimal feature is selected for the split. This approach reduces correlation between features and improves the model's generalization ability.

[0201] For regression tasks, the goal of random forest is to minimize the mean squared error between the predicted values ​​and the actual values:

[0202]

[0203] Where N is the number of samples, is the model’s predicted value for the i-th sample, y i is the actual value of the i-th sample.

[0204] In this embodiment, after the life prediction model is trained in step 3 to obtain a trained life prediction model (i.e., the aforementioned random forest model), the remaining life information of the sensor to be tested can be predicted based on the trained life prediction model. The process includes:

[0205] Collect the degradation status information of the sensor to be tested and the current environmental factors information;

[0206] The degradation state information and the current environmental factor information are input into the trained life prediction model to obtain the remaining life prediction information of the sensor to be tested.

[0207] In a specific embodiment, see Table 2, which shows an example of a failed sample data test set, with k, T acc RH acc , Δε, E and AF multi , construct the input feature vector. Through the trained life prediction model, the prediction analysis of the sensors that have not failed can be performed:

[0208] 1) k = 1, T acc =55℃、RH acc =80%, Δε=0.01, E=1V / m and AF multi =2.58, the model predicts t norm =280 hours.

[0209] 2) k = 1, T acc=25℃、RH acc =50%, Δε=0.025, E=3V / m and AF multi =409.88, the model predicts t norm =5100 hours.

[0210] Table 2 Data test set

[0211] Serial number stage <![CDATA[T acc ]]> <![CDATA[RH acc ]]> Δε E <![CDATA[t acc ]]> <![CDATA[AF multi ]]> <![CDATA[t norm ]]> 1 1 55℃ 80％ 0.01 1v / m 120 2.58 309.6 2 1 55℃ 80％ 0.01 1v / m 95 2.58 245.1 3 2 45℃ 70％ 0.02 2v / m 72 324.03 23305 4 2 45℃ 70％ 0.02 2v / m 50 324.03 16201.5 5 2 45℃ 70％ 0.02 2V / m 80 324.03 25922.4 6 3 25℃ 50％ 0.025 3V / m 20 409.88 8197.6 7 3 25℃ 50％ 0.025 3V / m 26 409.88 10637.9 8 3 25℃ 50％ 0.025 3V / m 18 409.88 7377.8

[0212] In another specific embodiment, using a micro-temperature sensor in power equipment as an example, the data collection, model input, lifespan prediction, and result application are described in detail. Assume that the sensor is installed in a high-voltage switchgear cabinet in a substation, monitoring the temperature of the busbar connection point. In this operating environment, the sensor is subjected to long-term periodic temperature fluctuations and humidity changes, as well as exposure to strong electromagnetic fields.

[0213] 1) Data Collection: The micro-temperature sensor operates normally, continuously measuring the temperature of the busbar joints. The sensor uses this temperature value as part of the degradation status information. Simultaneously, the accompanying environmental sensor collects information on current environmental factors, such as the actual temperature and humidity around the sensor, and the current load level within the switchgear, which can affect sensor lifespan.

[0214] 2) Model Input: This data is uploaded in real time via on-site edge acquisition devices. A data acquisition unit installed in the switchgear packages and transmits temperature sensor readings and environmental parameters within the switchgear. The monitoring terminal or backend server uses the acquired sensor degradation information and current environmental factors as input data into a pre-trained life prediction model.

[0215] It should be noted that this lifespan prediction model, constructed based on the method proposed in this invention, fully considers the effects of temperature, humidity, mechanical stress, and electromagnetic interference on sensor aging, and is trained using accelerated test data. In this embodiment, the model inputs include a series of recent sensor temperature measurements and their corresponding timestamps, ambient humidity levels, and historical accumulated operating time.

[0216] 3) Lifespan Prediction: Based on the collected data, the model outputs the remaining lifespan of the micro-temperature sensor. Assume that the model calculations indicate that the sensor is expected to operate stably for approximately 300 days. Lifespan prediction can be performed periodically or triggered immediately upon detecting an abnormal performance trend, enabling timely assessment.

[0217] 4) Application of results: The lifespan prediction results can be used for operation and maintenance decision-making and safety assurance. For example, a threshold warning mechanism can be set in the system to automatically trigger an alarm when the predicted remaining lifespan falls below a preset threshold.

[0218] In this embodiment, as the sensor continues to operate, the model may issue a warning after several months that its lifespan is less than 90 days, indicating that the temperature sensor is nearing its lifespan limit and needs to be replaced during the next overhaul. Based on this, operations and maintenance personnel can prepare spare parts and develop a maintenance plan in advance to replace the sensor before it fails, thus avoiding sensor failure. After replacing the sensor, the model can be recalibrated or retrained to adapt to the characteristics of the new sensor, and its lifespan can continue to be monitored and evaluated. This practical application case shows that the lifespan prediction model provides real-time assessment and prediction of the health status of power microsensors, improving the reliability and intelligence of power equipment monitoring systems.

[0219] In another specific embodiment, the life prediction model of the present invention can be used in conjunction with a system including a sensor acquisition module, an edge computing processing module, a cloud model prediction module, and a user interaction and early warning module to form a complete closed-loop system from on-site data collection, remote life calculation to information feedback and early warning.

[0220] 1) Acquisition module: Responsible for acquiring front-end data, including real-time acquisition units for the power microsensors to be tested and related environmental parameters. This module is typically composed of multiple microsensors and data acquisition devices installed on-site at the power equipment. Taking a temperature sensor as an example, the sensor acquisition module acquires the output signal of the sensor itself on the one hand, and information about factors related to the sensor's environment (such as ambient temperature, humidity, vibration acceleration, electromagnetic field strength, etc.) on the other hand. The acquisition module may include the sensor body, signal conditioning circuitry, and a primary A / D converter or digital interface circuit to convert analog signals into digital signals. For digital microsensors, their measured values ​​are read directly through the communication interface. This module typically sends data to the edge computing module on-site via wired or wireless means.

[0221] 2) Edge computing module: Located on edge devices near the sensor deployment site (such as communication gateways or industrial control computers within substations). Its primary function is to perform preliminary processing and local storage of raw data transmitted by the acquisition module, and to undertake certain real-time analysis tasks. The edge computing module preprocesses sensor data, including data cleaning, filtering and noise reduction, outlier removal, and simple feature extraction, improving data quality while reducing invalid data transmission. For example, if a sensor temporarily experiences an abnormal spike in readings, the edge-side algorithm can promptly filter it to avoid interference with the lifespan model's judgment. Secondly, the edge computing module can execute portions of the lifespan assessment calculations or health monitoring algorithms locally as needed, achieving rapid response. The edge module packages the organized data and sends it to the cloud-based model prediction module via a secure and reliable network connection. The communication protocol used by the edge module can be industrial Ethernet, 4G / 5G cellular networks, or dedicated IoT protocols.

[0222] 3) Cloud-based model prediction module: Deployed on a remote server or cloud computing platform, it carries out the operation of the life prediction model. This module receives sensor degradation status data and environmental factor data uploaded by the edge computing module, and calls the pre-trained power microsensor life prediction model for calculation and processing. The cloud has storage and computing capabilities, and can centrally analyze large-scale data from multiple field devices. After obtaining new data, the model prediction module will integrate historical data to evaluate the current health status of the sensor and output the remaining life estimation result or health indicators (such as health factor value, failure probability, etc.). Due to the abundant cloud computing resources, the module can run complex machine learning models (such as random forests, neural networks, etc.) to improve prediction accuracy, and can dynamically adjust model parameters for different stress factor combinations to simulate sensor aging behavior consistent with the field environment. In addition, the cloud-based module has established a database to store the historical operation data, prediction results and model version information of all sensors, so as to track the long-term evolution of sensor performance and continuously improve the model.

[0223] 4) User Interaction and Early Warning Module: This module provides a human-machine interface and automatic alarm functionality for power operation and maintenance personnel and system managers, transforming the lifespan assessment results obtained by the cloud-based model prediction module into intuitive information output and safety measures. This module can be implemented in various formats, such as a web monitoring platform, mobile terminal application, or local monitoring terminal, to display the real-time status and lifespan prediction information of each sensor. Through a graphical user interface, users can easily view the sensor's current remaining lifespan, historical health curve, and trends in key environmental parameters that affect lifespan. Once the cloud-based model prediction module determines that a sensor's lifespan is nearing end, the early warning submodule automatically generates an alarm signal. This user interaction and early warning module forms a closed-loop lifespan assessment system: from data collection and cloud-based analysis to result presentation and manual intervention, the health status of power microsensors remains monitorable, manageable, and under control throughout their lifespan, significantly improving the efficiency and safety of microsensor operation and maintenance in power systems.

[0224] The present invention combines a dynamic acceleration ratio with a multi-stress staged strategy. Traditional accelerated testing methods are usually conducted under a single stress condition, resulting in single test data that is difficult to fully reflect the failure mechanism of sensors in complex environments. However, the present invention divides the accelerated test into several stages, applies different stress combinations in each stage, and dynamically adjusts the acceleration ratio, so that the data covers a wider range of failure mechanisms. This multi-stage, multi-stress strategy not only enhances data diversity, but also ensures that different failure mechanisms are fully stimulated in the corresponding stages.

[0225] For example, the testing process can be divided into initial, mid-term, and final stages, with different stress combinations applied in each stage to target different failure mechanisms. For example, in the initial stage, temperature and humidity stresses are intensified to accelerate the evolution of material defects; in the mid-term, mechanical stress and electromagnetic interference are enhanced to stimulate chemical diffusion and material fatigue; and in the final stage, mechanical stress and electromagnetic interference are further intensified to examine later failure mechanisms such as thermo-mechanical fatigue and dielectric breakdown. During the test, the acceleration ratio of each environmental stress is dynamically adjusted based on the sensor's real-time degradation status. This dynamic adjustment ensures that the test can adapt to the different failure mechanisms of the sensor at different stages, avoiding atypical failures that may be caused by a single, constant high stress.

[0226] The above-mentioned staged multi-stress dynamic strategy significantly enhances the diversity of experimental data. Due to the singleness of stress conditions, traditional methods often cannot cover a variety of failure mechanism intervals, resulting in limited model training samples and difficulty in accurately capturing complex failure patterns. However, the present invention, through a multi-stage, multi-stress combination design, covers a variety of failure mechanisms from material defects and interface problems to thermal-mechanical fatigue and dielectric breakdown, greatly enriching the training samples and providing richer and more diverse data support for the random forest model, which is conducive to the model learning complex failure patterns.

[0227] At the same time, while introducing the random forest model, the present invention still retains physical models such as the Arrhenius model and the Coffin-Manson model as prior information. This method fully utilizes the respective advantages of data-driven and physical models, realizes the complementarity of the two, and ensures the accuracy and interpretability of life prediction. Embedding the prior knowledge of the physical model in the data-driven model enables the model to not only learn the patterns in the data, but also make reasonable predictions based on the laws of physics. This avoids the blindness that may occur in pure data-driven models when faced with complex failure mechanisms. The physical model provides a theoretical basis for the failure mechanism, so that the prediction results of the random forest model have a physical explanation, which improves the credibility of the model and the trust of users.

[0228] This paper addresses the critical challenge of evaluating the lifespan of power microsensors in complex environments by proposing an innovative method that combines multi-field coupled dynamic acceleration testing with a dynamic acceleration ratio optimization strategy. Through systematic, multi-dimensional accelerated aging and data acquisition, combined with advanced data analysis and lifespan prediction models, this paper achieves a high-fidelity assessment of potential failure processes that sensors may experience throughout their lifespan.

[0229] 1. The accelerated test method proposed in the present invention comprehensively considers multiple environmental factors such as temperature, humidity, mechanical stress, and electromagnetic interference, and can more realistically simulate the complex working conditions of power microsensors in actual working environments. Traditional accelerated tests often only focus on a single or a few stress factors, which makes it difficult to fully reflect the aging behavior of sensors in real environments. The present invention comprehensively covers the main environmental factors that affect the life of sensors through a multi-factor coupling model (i.e., the calculation formula of the acceleration factor of each environmental factor), ensuring that the test conditions are highly consistent with the actual working conditions, thereby improving the scientific nature and accuracy of life evaluation. The use of a multi-factor coupling model can effectively characterize the comprehensive impact of different environmental stresses on the life of sensors. The interactions between the various factors are complex and nonlinear, and a single factor model is difficult to accurately describe the failure mechanism. The present invention takes into account the interactions between factors such as temperature, humidity, mechanical stress, and electromagnetic interference by introducing a coupling acceleration factor, ensuring that the model can predict the life of sensors in multi-stress environments with high precision. This multi-factor coupling processing method significantly improves the accuracy and reliability of the model.

[0230] 2. Traditional accelerated tests usually fix the stress level at the beginning of the test and cannot adapt to the changes in the failure mechanism of the sensor at different aging stages. The dynamic acceleration ratio optimization strategy (DAFO) proposed in the present invention can dynamically adjust the acceleration factor of each environmental stress according to the real-time degradation state of the sensor. This staged dynamic adjustment method enables the accelerated test to accurately stimulate the corresponding failure mode at different stages, avoiding atypical failures caused by excessive stress, while ensuring that the main failure modes are fully stimulated within a reasonable time, significantly improving the scientific nature and effectiveness of the test. The present invention divides the accelerated test into several stages, and each stage applies different combinations of acceleration factors for different failure mechanisms. The initial stage mainly strengthens temperature and humidity stresses to accelerate the evolution of material defects and interface problems; the mid-stage turns to enhance mechanical stress and electromagnetic interference to stimulate failure mechanisms such as chemical diffusion, corrosion and material fatigue; the late stage further strengthens mechanical stress and electromagnetic interference to examine late failure mechanisms such as thermal-mechanical fatigue, material cracking and dielectric breakdown. This phased test design ensures that each failure mechanism can be fully stimulated at the corresponding stage, avoiding interference and mutual influence between different failure mechanisms, and improving the accuracy and comprehensiveness of life assessment.

[0231] 3. The present invention adopts the random forest, an integrated learning method, which can effectively process high-dimensional and nonlinear data characteristics. The degradation data and environmental stress data collected by the sensor at different stages often have high-dimensional, nonlinear and stage-specific characteristics. It is difficult for the traditional single model to accurately characterize its complex degradation laws. By constructing multiple decision trees and integrating the prediction results of each decision tree through an integrated learning mechanism, the random forest significantly improves the model's anti-overfitting ability and generalization performance, and can more accurately predict the life of the sensor. Traditional life prediction methods are highly dependent on theoretical models, which often have difficulties when faced with complex, multi-factor coupled failure mechanisms. The present invention uses a data-driven random forest model and uses a large amount of accelerated test data for training. It can autonomously learn the degradation laws of sensors in multiple stress environments, reducing dependence on a single theoretical model. This method not only improves the accuracy of life prediction, but also enhances the adaptability and flexibility of the model, and can better cope with complex situations in practical applications.

[0232] In addition, the accelerated test method and life prediction model of the present invention have wide applicability and are applicable to various types of power micro sensors. Regardless of whether it is a sensor based on different materials, structures or functions, the method of the present invention can accurately evaluate its life by adjusting the corresponding acceleration factors and model parameters. This wide applicability gives the present invention broad market potential and application prospects in the field of power sensors. In addition to power micro sensors, the method and model of the present invention can also be extended to other micro devices that require high reliability evaluation, such as communication equipment, medical equipment, aerospace electronic components, etc. By adjusting environmental stress factors and optimizing test design, the method of the present invention can adapt to the needs of high reliability device life evaluation in different fields, and has a high degree of versatility and extensibility.

[0233] Example 2:

[0234] The present invention based on the same inventive concept also provides a system for constructing a life prediction model for a power micro sensor. Figure 5 Shown, including:

[0235] A determination module, configured to determine a plurality of accelerated test stages and a combination of acceleration factors corresponding to each accelerated test stage based on the acquired historical degradation state information of the sensor;

[0236] an acceleration module, configured to perform an acceleration test on the sensor based on a combination of acceleration factors corresponding to a plurality of acceleration test stages to obtain an acceleration test data set;

[0237] The training module is used to train the life prediction model using the accelerated test data set as input and the remaining life of the sensor as output to obtain a trained life prediction model, wherein the remaining life of the sensor is calculated based on the accelerated test data set.

[0238] Preferably, the determining module is further configured to:

[0239] Setting up multiple accelerated test stages based on the acquired historical degradation status information of the sensor;

[0240] Based on the historical degradation state information of the sensor, the acceleration factor dynamic function is used to configure the initial acceleration factor combination corresponding to multiple acceleration test stages;

[0241] An objective function is constructed with the goal of minimizing the accelerated test time, and the objective function is solved using a preset model solving method to obtain configuration information of the acceleration factor combination corresponding to each accelerated test stage;

[0242] The acceleration factor combination corresponding to each accelerated test stage is generated based on the configuration information of the acceleration factor combination corresponding to each accelerated test stage and the initial acceleration factor combination.

[0243] Preferably, the calculation formula of the acceleration factor dynamic function in the determination module is as follows:

[0244]

[0245] Among them, Φ(k) represents the dynamic function of the acceleration factor of stage k, Indicates the acceleration factor corresponding to the temperature factor of the kth stage, Indicates the acceleration factor corresponding to the humidity factor of the kth stage, represents the acceleration factor corresponding to the mechanical stress factor of stage k, Indicates the acceleration factor corresponding to the electromagnetic interference factor in the kth stage.

[0246] Preferably, the calculation formula of the acceleration factor corresponding to the temperature factor in the determination module is as follows:

[0247]

[0248] Among them, E a is the activation energy, k B is the Boltzmann constant, T is the absolute temperature, T acc Indicates the absolute temperature of the accelerated test conditions, T norm Indicates the absolute temperature under normal test conditions, AF T Indicates the acceleration factor of temperature factor.

[0249] Preferably, the calculation formula of the acceleration factor corresponding to the humidity factor in the determination module is as follows:

[0250] AF RH =exp[β(RH acc -RHnorm )]

[0251] Among them, β is the humidity sensitivity coefficient, RH is the relative humidity, RH acc Indicates the relative humidity of the accelerated test conditions, RH norm Indicates the relative humidity under normal test conditions, AF RH Indicates the acceleration factor of humidity.

[0252] Preferably, the calculation formula of the acceleration factor corresponding to the mechanical stress factor in the determination module is as follows:

[0253]

[0254] Among them, Δε acc Indicates accelerated test conditions, Δε norm represents normal working conditions, p represents a constant related to the material, AF M Indicates the acceleration factor of the mechanical stress factor.

[0255] Preferably, the calculation formula of the acceleration factor corresponding to the electromagnetic interference factor in the determination module is as follows:

[0256]

[0257] Among them, E acc represents the electromagnetic stress under accelerated conditions, E norm represents the electromagnetic stress under normal conditions, m is an index related to the material aging mechanism, AF EM Indicates the acceleration factor corresponding to the electromagnetic interference stress factor.

[0258] Preferably, the calculation formula of the objective function in the determination module is as follows:

[0259]

[0260] Among them, t k is the duration of the k-th stage experiment, P modlei It represents the probability that the i-th failure mode is stimulated to a certain confidence level under the test design, δ is the pre-set threshold requirement, K is the total number of test stages, and Φ(k) represents the dynamic function of the acceleration factor of stage k.

[0261] Preferably, the acceleration module is further configured to:

[0262] adjusting an accelerated test environment corresponding to each accelerated test stage based on acceleration factors in different acceleration factor combinations corresponding to the multiple accelerated test stages, and performing an accelerated test on the sensor based on the adjusted accelerated test environment in each accelerated test stage to obtain degradation status information of the sensor in each accelerated test stage;

[0263] An accelerated test data set is generated based on the acquired environmental factor information, acceleration factor combination, and sensor degradation status information corresponding to each accelerated test stage.

[0264] Preferably, the training process of the life prediction model in the training module includes:

[0265] The environmental factor information, sensor degradation status information, and acceleration factor of each accelerated test stage in the accelerated test data set are combined into a feature matrix as input, and the environmental factor information of each accelerated test stage in the accelerated test data and the sensor degradation status information corresponding to the acceleration factor are converted into remaining life information as output. Random forest is used to train the life prediction model to generate a trained life prediction model.

[0266] Preferably, the system further comprises:

[0267] Model application module, used to collect degradation status information of the sensor to be tested and current environmental factors information;

[0268] The degradation state information and the current environmental factor information are input into the trained life prediction model to obtain the remaining life prediction information of the sensor to be tested.

[0269] The present invention is equipped with a real-time acquisition module and a main controller to continuously monitor and control the sensors during the test. The real-time acquisition module includes a multi-channel data acquisition card or an embedded A / D conversion unit, which connects to the microsensors to be tested and environmental parameter sensors, and can simultaneously acquire sensor output signals and environmental stress parameters at a high sampling rate. The main controller, which can be an industrial control computer or a high-performance single-chip microcomputer, is connected to the environmental control box, loading device, and acquisition module via an interface. It sends control instructions to the environmental control box and loading device to achieve closed-loop control of the temperature field, humidity, and mechanical and electrical stress application process; it also reads the status data of each sensor in real time from the acquisition module. Communication can be wired or wireless, depending on reliability requirements. Wired communication with strong anti-interference capabilities is preferred to connect each sensor acquisition unit to the main controller, ensuring stable and reliable data transmission in high-temperature and electromagnetic interference environments. The main controller can communicate with the host computer via Ethernet or USB interfaces, transmitting collected data and receiving control instructions at high speed. If necessary, wireless communication can also be used between the sensor nodes and the controller.

[0270] The present invention also includes a test monitoring and data management software system. It realizes the configuration, monitoring and data processing of the entire accelerated test process. The user can set the accelerated test plan of the stress factors and their combinations in each stage, automatically coordinate the environmental chamber and each loading module to apply the corresponding stress in a time series according to the pre-set test steps, and display the key parameters and sensor outputs in real time. The data management function stores and preliminarily analyzes the large amount of sensor degradation status data and environmental stress data collected: such as drawing the attenuation curve of sensor performance over time, calculating the change rate of key performance indicators, etc. The software can integrate the life prediction model, call the life evaluation algorithm during the test, and perform life prediction analysis on the sensor being tested. In this way, the experimenter can obtain the estimated value of the remaining life of the sensor while the accelerated aging test is being carried out, and compare and verify the model prediction results with the actual degradation situation, so as to evaluate the accuracy and reliability of the model. The construction of the entire platform provides a solid experimental support for the construction and application of the life prediction model, which can simulate the degradation process of the power micro sensor throughout its life cycle in a controllable experimental environment, and provide a reference basis for subsequent practical applications.

[0271] Example 3

[0272] like Figure 6 As shown, the present invention also provides an electronic device, which may be a computer, a single-chip microcomputer, a smart mobile device, or the like. The electronic device in this embodiment may include a processor, a memory, a transceiver component, and the like. The memory, processor, and transceiver component are connected via a bus; the memory may be used to store an execution program, which may include instructions; and the processor may be used to execute the instructions stored in the memory. The memory may also be used to store data, which may be accessed and / or modified during the execution of the instructions.

[0273] The processor may be a central processing unit (CPU), or it may be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. It is the computing core and control core of the terminal, which is suitable for implementing one or more instructions, specifically suitable for loading and executing one or more instructions in the storage medium to implement the corresponding method flow or corresponding function, so as to realize the steps of a method for constructing a life prediction model of an electric microsensor in the above embodiment.

[0274] Example 4

[0275] Based on the same inventive concept, the present invention also provides a readable storage medium, specifically an electronic device readable storage medium (Memory), which is a memory device in an electronic device for storing programs and data. It can be understood that the storage medium here can include both built-in storage media in the electronic device and, of course, extended storage media supported by the electronic device. The storage medium provides a storage space, which stores the operating system of the terminal. In addition, one or more instructions suitable for being loaded and executed by the processor are also stored in the storage space. These instructions can be one or more execution programs (including program codes). It should be noted that the storage medium here can be a high-speed RAM memory or a non-volatile memory, such as at least one disk memory. The processor loads and executes one or more instructions stored in the storage medium, which can realize the steps of the method for constructing a power microsensor life prediction model in the above embodiment.

[0276] It will be understood by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0277] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts 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, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0278] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0279] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0280] The above are merely embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention are included in the scope of the claims of the present invention to be approved.

Claims

1. A method for constructing a life prediction model for a power microsensor, characterized in that: include: Determining multiple accelerated test stages and acceleration factor combinations corresponding to each accelerated test stage based on the acquired historical degradation state information of the sensor; performing an accelerated test on the sensor based on a combination of acceleration factors corresponding to a plurality of accelerated test stages to obtain an accelerated test data set; The life prediction model is trained with the accelerated test data set as input and the remaining life of the sensor as output to obtain a trained life prediction model, wherein the remaining life of the sensor is calculated based on the accelerated test data set.

2. The method according to claim 1, characterized in that The method of determining a plurality of accelerated test stages and a combination of acceleration factors corresponding to each accelerated test stage based on the acquired historical degradation state information of the sensor includes: Setting up multiple accelerated test stages based on the acquired historical degradation status information of the sensor; Based on the historical degradation state information of the sensor, the acceleration factor dynamic function is used to configure the initial acceleration factor combination corresponding to multiple acceleration test stages; An objective function is constructed with the goal of minimizing the accelerated test time, and the objective function is solved using a preset model solving method to obtain configuration information of the acceleration factor combination corresponding to each accelerated test stage; The acceleration factor combination corresponding to each accelerated test stage is generated based on the configuration information of the acceleration factor combination corresponding to each accelerated test stage and the initial acceleration factor combination.

3. The method according to claim 2, characterized in that The calculation formula of the acceleration factor dynamic function is as follows: Among them, Φ(k) represents the dynamic function of the acceleration factor of stage k, Indicates the acceleration factor corresponding to the temperature factor of the kth stage, Indicates the acceleration factor corresponding to the humidity factor of the kth stage, represents the acceleration factor corresponding to the mechanical stress factor of stage k, Indicates the acceleration factor corresponding to the electromagnetic interference factor in the kth stage.

4. The method according to claim 3, characterized in that The calculation formula of the acceleration factor corresponding to the temperature factor is as follows: Among them, E a is the activation energy, k B is the Boltzmann constant, T is the absolute temperature, T acc Indicates the absolute temperature of the accelerated test conditions, T norm Indicates the absolute temperature under normal test conditions, AF T Indicates the acceleration factor of temperature factor.

5. The method according to claim 3, characterized in that The calculation formula of the acceleration factor corresponding to the humidity factor is as follows: AF RH =exp[β(RH acc -RH norm )] Among them, β is the humidity sensitivity coefficient, RH is the relative humidity, RH acc Indicates the relative humidity of the accelerated test conditions, RH norm Indicates the relative humidity under normal test conditions, AF RH Indicates the acceleration factor of humidity.

6. The method according to claim 3, characterized in that The calculation formula of the acceleration factor corresponding to the mechanical stress factor is as follows: Among them, Δε acc Indicates accelerated test conditions, Δε norm represents normal working conditions, p represents a constant related to the material, AF M Indicates the acceleration factor of the mechanical stress factor.

7. The method according to claim 3, characterized in that The calculation formula of the acceleration factor corresponding to the electromagnetic interference factor is as follows: Among them, E acc represents the electromagnetic stress under accelerated conditions, E norm represents the electromagnetic stress under normal conditions, m is an index related to the material aging mechanism, AF EM Indicates the acceleration factor corresponding to the electromagnetic interference stress factor.

8. The method according to claim 2, characterized in that The calculation formula of the objective function is as follows: Among them, t k is the duration of the k-th stage experiment, P modlei It represents the probability that the i-th failure mode is stimulated to a certain confidence level under the test design, δ is the pre-set threshold requirement, K is the total number of test stages, and Φ(k) represents the dynamic function of the acceleration factor of stage k.

9. The method according to claim 1, characterized in that The accelerated test is performed on the sensor based on a combination of acceleration factors corresponding to a plurality of accelerated test stages to obtain an accelerated test data set, including: adjusting an accelerated test environment corresponding to each accelerated test stage based on acceleration factors in different acceleration factor combinations corresponding to the multiple accelerated test stages, and performing an accelerated test on the sensor based on the adjusted accelerated test environment in each accelerated test stage to obtain degradation status information of the sensor in each accelerated test stage; An accelerated test data set is generated based on the acquired environmental factor information, acceleration factor combination, and sensor degradation status information corresponding to each accelerated test stage.

10. The method according to claim 1, characterized in that The training process of the lifespan prediction model includes: The environmental factor information, sensor degradation status information, and acceleration factor of each accelerated test stage in the accelerated test data set are combined into a feature matrix as input, and the environmental factor information of each accelerated test stage in the accelerated test data and the sensor degradation status information corresponding to the acceleration factor are converted into remaining life information as output. Random forest is used to train the life prediction model to generate a trained life prediction model.

11. The method according to claim 1, wherein After obtaining the trained lifespan prediction model, the method further includes: Collect the degradation status information of the sensor to be tested and the current environmental factors information; The degradation state information and the current environmental factor information are input into the trained life prediction model to obtain the remaining life prediction information of the sensor to be tested.

12. A system for constructing a life prediction model for a power micro sensor, characterized in that: include: A determination module, configured to determine a plurality of accelerated test stages and a combination of acceleration factors corresponding to each accelerated test stage based on the acquired historical degradation state information of the sensor; an acceleration module, configured to perform an acceleration test on the sensor based on a combination of acceleration factors corresponding to a plurality of acceleration test stages to obtain an acceleration test data set; The training module is used to train the life prediction model using the accelerated test data set as input and the remaining life of the sensor as output to obtain a trained life prediction model, wherein the remaining life of the sensor is calculated based on the accelerated test data set.

13. The system according to claim 12, wherein: The determining module is further configured to: Setting up multiple accelerated test stages based on the acquired historical degradation status information of the sensor; Based on the historical degradation state information of the sensor, the acceleration factor dynamic function is used to configure the initial acceleration factor combination corresponding to multiple acceleration test stages; An objective function is constructed with the goal of minimizing the accelerated test time, and the objective function is solved using a preset model solving method to obtain configuration information of the acceleration factor combination corresponding to each accelerated test stage; The acceleration factor combination corresponding to each accelerated test stage is generated based on the configuration information of the acceleration factor combination corresponding to each accelerated test stage and the initial acceleration factor combination.

14. The system according to claim 13, wherein: The calculation formula of the acceleration factor dynamic function in the determination module is as follows: Among them, Φ(k) represents the dynamic function of the acceleration factor of stage k, Indicates the acceleration factor corresponding to the temperature factor of the kth stage, Indicates the acceleration factor corresponding to the humidity factor of the kth stage, represents the acceleration factor corresponding to the mechanical stress factor of stage k, Indicates the acceleration factor corresponding to the electromagnetic interference factor in the kth stage.

15. The system according to claim 14, wherein: The calculation formula of the acceleration factor corresponding to the temperature factor in the determination module is as follows: Among them, E a is the activation energy, k B is the Boltzmann constant, T is the absolute temperature, T acc Indicates the absolute temperature of the accelerated test conditions, T norm Indicates the absolute temperature under normal test conditions, AF T Indicates the acceleration factor of temperature factor.

16. The system according to claim 14, wherein: The calculation formula of the acceleration factor corresponding to the humidity factor in the determination module is as follows: AF RH =exp[β(RH acc -RH norm )] Among them, β is the humidity sensitivity coefficient, RH is the relative humidity, RH acc Indicates the relative humidity of the accelerated test conditions, RH norm Indicates the relative humidity under normal test conditions, AF RH Indicates the acceleration factor of humidity.

17. The system according to claim 14, wherein: The calculation formula of the acceleration factor corresponding to the mechanical stress factor in the determination module is as follows: Among them, Δε acc Indicates accelerated test conditions, Δε norm represents normal working conditions, p represents a constant related to the material, AF M Indicates the acceleration factor of the mechanical stress factor.

18. The system according to claim 14, wherein: The calculation formula of the acceleration factor corresponding to the electromagnetic interference factor in the determination module is as follows: Among them, E acc represents the electromagnetic stress under accelerated conditions, E norm represents the electromagnetic stress under normal conditions, m is an index related to the material aging mechanism, AF EM Indicates the acceleration factor corresponding to the electromagnetic interference stress factor.

19. The system according to claim 13, wherein: The calculation formula of the objective function in the determination module is as follows: Among them, t k is the duration of the k-th stage experiment, P modlei It represents the probability that the i-th failure mode is stimulated to a certain confidence level under the test design, δ is the pre-set threshold requirement, K is the total number of test stages, and Φ(k) represents the dynamic function of the acceleration factor of stage k.

20. The system according to claim 12, wherein: The acceleration module is further configured to: adjusting an accelerated test environment corresponding to each accelerated test stage based on acceleration factors in different acceleration factor combinations corresponding to the multiple accelerated test stages, and performing an accelerated test on the sensor based on the adjusted accelerated test environment in each accelerated test stage to obtain degradation status information of the sensor in each accelerated test stage; An accelerated test data set is generated based on the acquired environmental factor information, acceleration factor combination, and sensor degradation status information corresponding to each accelerated test stage.

21. The system according to claim 12, wherein: The training process of the life prediction model in the training module includes: The environmental factor information, sensor degradation status information, and acceleration factor of each accelerated test stage in the accelerated test data set are combined into a feature matrix as input, and the environmental factor information of each accelerated test stage in the accelerated test data and the sensor degradation status information corresponding to the acceleration factor are converted into remaining life information as output. Random forest is used to train the life prediction model to generate a trained life prediction model.

22. The system according to claim 12, wherein: The system further includes: Model application module, used to collect degradation status information of the sensor to be tested and current environmental factors information; The degradation state information and the current environmental factor information are input into the trained life prediction model to obtain the remaining life prediction information of the sensor to be tested.

23. An electronic device, characterized in that: include: at least one processor and memory; The memory and the processor are connected via a bus; The memory is used to store one or more programs; When the one or more programs are executed by the at least one processor, the method for constructing a power microsensor life prediction model according to any one of claims 1 to 11 is implemented.

24. A readable storage medium, characterized in that An execution program is stored thereon, and when the execution program is executed, the method for constructing a power microsensor life prediction model according to any one of claims 1 to 11 is implemented.

Images