Method and system for evaluating service life of inductor

Abstract

The invention provides a method and a system for evaluating the service life of an inductor, and relates to the technical field of electrical variable measurement, and the method comprises the steps: capturing and compensating a measurement system common-mode error introduced by environmental stress in real time in a differential measurement mode through introducing a self-calibration test clamp of an integrated benchmark reference path; the problem of measurement errors caused by self-aging of a test system is solved, so that clamp degradation and real performance attenuation of the inductor to be tested are effectively separated, and high fidelity of the obtained first electrical performance data set is ensured; by calculating a physical electrical residual vector between the high-precision measured data and a predicted electrical performance data set, a diagnosis indication signal which truly reflects an unmodeled failure mechanism can be generated, and then a closed-loop feedback optimization module is driven to execute accurate adaptive adjustment operation, and finally, the whole evaluation method gets rid of system error interference, and the evaluation accuracy is improved. And outputting a residual service life prediction result with high scientificity and engineering application credibility.

Description

Technical Field

[0001] This invention relates to the field of electrical variable measurement technology, specifically to a method and system for evaluating the lifespan of an inductor. Background Technology

[0002] Inductor life assessment technology originated from the growing demand for reliability in electronic components. Early on, inductors were often considered highly reliable passive devices, and their assessment primarily focused on initial electrical performance parameters, with lifespan issues not receiving sufficient attention. As power electronics technology advances towards higher frequencies, higher power densities, and miniaturization, the degradation mechanism of key materials under multiple electrical, thermal, and mechanical stresses is studied through the establishment of thermally accelerated life tests based on the Arrhenius model, finite element simulation analysis, and multiphysics coupling models. This allows for the accurate prediction of the reliable operating time of inductors under specific operating conditions, providing crucial scientific basis for the design and preventative maintenance of high-reliability systems.

[0003] In the prior art, publication number CN120028605A, entitled "A High-Frequency Inductor Evaluation Device and Method," a high-frequency inductor that meets RF link requirements can be obtained by setting a minimum test circuit for the high-frequency inductor under test and evaluating the S-parameters of the minimum test circuit in the high-frequency band using a vector network analyzer. This enables rapid screening and evaluation of high-frequency inductors, avoiding reliance on batch testing on production lines and indirect judgment of high-frequency inductor performance based on product performance indicators and pass-through rates. The aforementioned high-frequency inductor evaluation device can quickly and accurately evaluate inductor performance under high-frequency conditions, improving the flexibility and applicability of testing.

[0004] However, the above technical solution has the following technical drawbacks: The core of inductor life assessment lies in simulating its long-term degradation process under harsh conditions such as high temperature and high current. However, the above-mentioned technical solution is essentially a performance measurement system under a normal temperature and small signal environment. It cannot actively apply any thermal or electrical stress to the inductor under test. Therefore, it can only compare the performance "snapshot" before and after the aging test, and cannot be used as an assessment platform to perform the aging process or conduct online monitoring. It fundamentally does not meet the necessary conditions for life assessment and lacks the ability to apply accelerated aging stress. Secondly, the aforementioned technical solutions cannot compensate for measurement errors caused by the aging of the test fixture itself, which is closely related to the previous drawback. Since the evaluation is based on the S-parameters of the entire "minimum test circuit" including the inductor, once it is subjected to aging tests in high-temperature or other stress environments, the test circuit board, connectors, and other components themselves will also experience performance degradation. Therefore, the final measured S-parameter changes are the result of the combined effects of inductor degradation and test fixture degradation; the errors of both cannot be separated, making it impossible to accurately quantify the true performance degradation of the inductor, thus compromising the scientific validity and credibility of the evaluation results. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for evaluating the lifespan of inductors, so as to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: A method for evaluating the lifespan of an inductor, comprising the following steps: Step 1: While applying a preset electrical excitation to the inductor under stress, simultaneously acquire: The first dataset characterizes the actual electrical performance of the inductor under test, and the second dataset characterizes the internal physical state of the inductor under test; Step 2: Based on a pre-defined multiphysics coupling model that maps internal physical states to electrical performance, generate a predicted electrical performance dataset for the inductor under test using the second dataset. Step 3: Calculate the residual between the first dataset and the predicted electrical performance dataset to quantify the deviation between the actual electrical performance and the predicted electrical performance of the inductor under test; Step 4: Based on the characteristics of the residual, identify the current dominant failure mechanism of the inductor under test and generate a diagnostic indication signal characterizing the dominant failure mechanism; Step 5: In response to the diagnostic indication signal, perform adaptive adjustment operations, which include: Adjust the type of stress applied to the inductor under test and update the prediction model used to predict the remaining lifespan of the inductor under test.

[0007] Step one specifically includes: While applying a preset electrical excitation to the inductor under stress, differential measurement is performed through a self-calibrating test fixture that integrates a reference path parallel to the test path of the inductor under test to calibrate in real time and obtain the first dataset used to characterize the actual electrical performance of the inductor under test. Simultaneously, a non-contact physical field imaging method is used to acquire physical characteristic parameters that reflect the health of the internal structure of the inductor under test, which are used as the second dataset; A passive device with stable performance under stress is deployed in the reference path to capture and compensate for the common-mode error of the measurement system introduced by environmental stress in real time; the non-contact physical field imaging method includes: using a high-speed infrared thermal imager to capture the transient thermal response of the inductor under test when responding to power pulse excitation, and using a three-dimensional magnetic field probe array to draw a spatial distribution map of the near-field magnetic field.

[0008] Step two specifically includes: Using the physical feature parameters in the second dataset as input to a pre-established parameterized reduced-order multiphysics coupling model, the multiphysics coupling model maps the physical feature parameters to the corresponding electrical performance parameters, thereby generating a predicted electrical performance dataset for the inductor under test. The multiphysics coupling model is obtained by performing high-fidelity finite element simulation on the inductor under test in the offline stage, scanning different physical damage levels and combinations of operating parameters in the simulation, recording their corresponding electrical performance responses, and finally fitting the analytical function for real-time calculation through model order reduction technology trained by neural network.

[0009] Step two also includes: The internal physical state parameters input to the model include the interfacial thermal resistance extracted from the transient thermal response and the magnetic field distribution entropy extracted from the near-field magnetic distribution; the predicted electrical performance dataset output by the multiphysics coupling model includes the predicted inductance value, and the specific calculation process is broken down as follows: Using the nominal inductance value as a reference, the inductance attenuation caused by thermal degradation and magnetic degradation is subtracted sequentially to obtain the corrected prediction result, i.e., the predicted inductance value.

[0010] Step three specifically includes: A multi-dimensional physical-electrical residual vector is calculated to simultaneously quantify the prediction accuracy of the multiphysics coupling model across multiple electrical performance dimensions. Each component of the physical-electrical residual vector is defined as the difference between the actual electrical performance parameters in the first dataset and the theoretical predicted performance in the predicted electrical performance dataset. The step of calculating the physical electrical residual vector also includes recording and storing historical trajectories, specifically the historical trajectory of the evolution of the physical electrical residual vector over time, so as to provide a data basis for identifying the time evolution trend of the failure mechanism in subsequent steps.

[0011] Step four specifically includes: The physical electrical residual vector is mapped to a preset two-dimensional failure feature space composed of mutually orthogonal failure mode basis vectors. By calculating the projection components of the physical electrical residual vector onto each failure mode basis vector, the contribution of different failure modes to the current performance deviation is determined, and a diagnostic indication signal that can quantify the current dominant failure mechanism is generated accordingly.

[0012] Step four also includes: The failure mode basis vectors are determined in advance by conducting accelerated aging tests and failure physics analysis on inductors of the same type as the inductor under test. One basis vector points to the direction of solder joint fatigue or terminal oxidation failure mode, which is mainly characterized by the increase of equivalent series resistance, while the other basis vector points to the direction of core material aging or structural cracking failure mode, which is mainly characterized by the decrease of inductance. The diagnostic indication signal is a failure mode bias index, and the calculation process of the failure mode bias index specifically includes: The absolute value of the inductor residual is divided by a preset inductor failure threshold to obtain the normalized inductor residual; the absolute value of the resistance residual contribution is divided by a preset resistance failure threshold to obtain the normalized resistance residual; the normalized resistance residual is defined as the resistance residual contribution, characterizing the severity of solder joint fatigue failure modes; the normalized inductor residual is defined as the inductor residual contribution, characterizing the severity of core degradation failure modes; the difference between the resistance residual contribution and the inductor residual contribution is divided by the sum of the resistance residual contribution and the inductor residual contribution to obtain the failure mode bias index.

[0013] Step five specifically includes: In response to the diagnostic indication signal, an adaptive adjustment operation is performed; the adaptive adjustment operation maps the value of the diagnostic indication signal to a preset decision matrix of associated failure mechanisms and test strategies. The decision matrix pre-sets the optimal stress combination corresponding to different numerical ranges of the diagnostic indication signal, including increasing the amplitude and frequency of temperature cycling when the diagnostic indication signal indicates that the solder joint fatigue is the main factor, and increasing the magnitude of the DC bias current when the indication core aging is the main factor.

[0014] Step five also includes: The decision matrix selectively performs at least one of the following: 1. Output a stress adjustment command to dynamically change the type or intensity of stress applied to the inductor under test, thereby accelerating aging in response to the current dominant failure mechanism; 2. From a model library containing multiple failure mode-specific degradation models, select a degradation model that matches the dominant failure mechanism characterized by the diagnostic indication signal to update the prediction model used to predict its remaining useful life.

[0015] An inductor lifespan assessment system, comprising: The multimodal data synchronization module is used to simultaneously acquire data while applying a preset electrical excitation to the inductor under stress. The first dataset characterizes the actual electrical performance of the inductor under test, and the second dataset characterizes the internal physical state of the inductor under test; The physical model derivation module, based on a pre-defined multi-physics coupling model that maps internal physical states to electrical performance, uses a second dataset to generate a predicted electrical performance dataset for the inductor under test. The residual quantization analysis module is used to calculate the residual between the first dataset and the predicted electrical performance dataset to quantify the deviation between the actual electrical performance and the predicted electrical performance of the inductor under test. The failure mechanism diagnosis module identifies the dominant failure mechanism of the inductor under test based on the characteristics of the residual and generates a diagnostic indication signal characterizing the dominant failure mechanism. The closed-loop feedback optimization module is used to respond to diagnostic indication signals and perform adaptive adjustment operations, which include: Adjust the type of stress applied to the inductor under test and update the prediction model used to predict the remaining lifespan of the inductor under test.

[0016] Compared with the prior art, the beneficial effects of the present invention are: This invention solves the technical problem of measurement error caused by the aging of the test system itself by introducing a self-calibrating test fixture that integrates a reference path parallel to the test path of the inductor under test. The solution deploys stable passive devices in the reference path and captures and compensates for the common-mode error of the measurement system introduced by environmental stress in real time through differential measurement. This effectively separates the performance degradation of the test fixture itself from the actual performance degradation of the inductor under test, ensuring the purity and accuracy of the first dataset characterizing the actual electrical performance of the inductor under test. Based on the aforementioned high-fidelity first dataset, the present invention further enhances the scientific validity and reliability of all subsequent analysis steps. By calculating the physical electrical residual vector between the high-precision measured data and the predicted electrical performance dataset, a diagnostic indication signal that truly reflects the unmodeled failure mechanism can be generated. This signal then drives the closed-loop feedback optimization module to perform precise adaptive adjustment operations, ultimately freeing the entire evaluation method from the interference of system errors and ensuring that the final output of the remaining service life prediction result has high scientific validity and engineering application reliability. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the overall application process of the present invention; Figure 2 This is a schematic diagram of the overall method steps of the present invention; Figure 3 This is a schematic diagram of the overall system framework structure of the present invention. Detailed Implementation

[0018] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0019] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0020] Example 1: Please see Figures 1 to 2 This invention provides a technical solution: a method for evaluating the lifespan of an inductor, the specific steps of which include: Step 1: While applying a preset electrical excitation to the inductor under stress, simultaneously acquire: The first dataset characterizes the actual electrical performance of the inductor under test, and the second dataset characterizes the internal physical state of the inductor under test; Step 2: Based on a pre-defined multiphysics coupling model that maps internal physical states to electrical performance, generate a predicted electrical performance dataset for the inductor under test using the second dataset. Step 3: Calculate the residual between the first dataset and the predicted electrical performance dataset to quantify the deviation between the actual electrical performance and the predicted electrical performance of the inductor under test; Step 4: Based on the characteristics of the residual, identify the current dominant failure mechanism of the inductor under test and generate a diagnostic indication signal characterizing the dominant failure mechanism; Step 5: In response to the diagnostic indication signal, perform adaptive adjustment operations, which include: Adjust the type of stress applied to the inductor under test and update the prediction model used to predict the remaining lifespan of the inductor under test.

[0021] In this embodiment, Figure 1 The steps are as follows: Step 1: synchronous data acquisition; Step 2: prediction performance generation; Step 3: residual calculation and deviation quantification; Step 4: failure mechanism identification and diagnosis; Step 5: adaptive adjustment and model update.

[0022] Example 2: Step one specifically includes: While applying a preset electrical excitation to the inductor under stress, differential measurement is performed through a self-calibrating test fixture that integrates a reference path parallel to the test path of the inductor under test to calibrate in real time and obtain the first dataset used to characterize the actual electrical performance of the inductor under test. Simultaneously, a non-contact physical field imaging method is used to acquire physical characteristic parameters that reflect the health of the internal structure of the inductor under test, which are used as the second dataset; A passive device with stable performance under stress is deployed in the reference path to capture and compensate for the common-mode error of the measurement system introduced by environmental stress in real time; the non-contact physical field imaging method includes: using a high-speed infrared thermal imager to capture the transient thermal response of the inductor under test when responding to power pulse excitation, and using a three-dimensional magnetic field probe array to draw a spatial distribution map of the near-field magnetic field.

[0023] In this embodiment, the self-calibrating test fixture is a precision test device that integrates two parallel electrical paths: one for connecting and testing the inductor under test (DUT), i.e., the DUT path; and the other for deploying a reference passive device with known and highly stable electrical performance parameters under a wide range of stress environments, i.e., the reference path. During use, the preset electrical excitation is applied synchronously to the two electrical paths, and the measurement system simultaneously acquires response signals from both electrical paths. Since any drift or noise introduced by environmental stress to the measurement circuit and instrument interface will act indiscriminately on the two tightly coupled electrical paths, forming a common-mode error, the degree to which the measurement result of the reference path deviates from its theoretical true value at this moment accurately quantifies the system measurement error at that instant. Finally, by performing differential processing, that is, subtracting the common-mode error captured by the reference path from the original measurement data of the inductor under test in real time, the self-calibrating test fixture outputs a high-fidelity actual electrical performance data of the inductor under test that has been dynamically self-calibrated, thereby ensuring the accuracy and reliability of the monitoring results under harsh stress conditions for a long time. Furthermore, by integrating raw data from different physical dimensions acquired through non-contact physical field imaging into a single parameter for quantifying the health status of the inductor's internal physical structure, namely the Structural Health Index (hereinafter referred to as Shealth), the calculation logic of Shealth is as follows: The normalized structural function thermal resistance peak drift (hereinafter referred to as Rpeak) is multiplied by the thermal degradation weight (hereinafter referred to as wth), and then added to the product of the normalized leakage magnetic field spatial distribution entropy (hereinafter referred to as Hmag) and the magnetic degradation weight (hereinafter referred to as wmag). This linear superposition structure ensures that there is a clear positive proportional relationship between the physical degradation as input (i.e., Rpeak and Hmag) and the output Shealth, that is, any increase in internal structural damage will monotonically increase this exponential value. The allocation of wth and wmag follows the SMART principle and is preset based on the failure physics analysis of a specific inductor in a specific application scenario. In this embodiment, the following settings are made: In applications with high risk of thermal failure, wth=0.7 is set so that the evaluation model can adaptively focus on the most likely dominant failure path; through standardization, the output value range of Shealth is limited to the interval [0, 1]. Furthermore, when Shealth approaches 0, it indicates that the heat conduction path and magnetic field distribution inside the inductor are close to their initial intact state, and the structure is very healthy; conversely, when Shealth approaches 1, it quantitatively indicates that one or more physical degradations have reached the failure threshold, and the structure is severely deteriorated. Rpeak was obtained by analyzing transient thermal response data captured by a high-speed infrared thermal imager, while Hmag was calculated by plotting the near-field magnetic field distribution map using a three-dimensional magnetic field probe array.

[0024] Step two specifically includes: Using the physical feature parameters in the second dataset as input to a pre-established parameterized reduced-order multiphysics coupling model, the multiphysics coupling model maps the physical feature parameters to the corresponding electrical performance parameters, thereby generating a predicted electrical performance dataset for the inductor under test. The multiphysics coupling model is obtained by performing high-fidelity finite element simulation on the inductor under test in the offline stage, scanning different physical damage levels and combinations of operating parameters in the simulation, recording their corresponding electrical performance responses, and finally fitting the analytical function for real-time calculation through model order reduction technology trained by neural network.

[0025] Furthermore, to construct a multiphysics coupling model, a high-fidelity digital twin model highly consistent with the physical characteristics of the inductor under test is created using finite element simulation software in the offline stage. Potential physical damages, including core cracks and material aging, are parameterized along with key operating parameters. Automated scripts are used to perform large-scale parametric scanning simulations on this model, systematically simulating thousands of combinations of damage levels and operating parameters. The corresponding inductance value and equivalent series resistance, among other electrical performance responses, are recorded for each combination, generating a comprehensive input-to-output mapping dataset. This mapping dataset is used to train a deep neural network. Through model order reduction techniques, the network learns and accurately fits the complex physical laws inherent in the simulation data. This trained and lightweight neural network is then solidified and deployed as an analytical function capable of real-time computation. In practical applications, it can predict the electrical performance of the inductor in milliseconds based on the input damage and operating condition information, thus replacing the previously time-consuming finite element simulation.

[0026] Step two also includes: The internal physical state parameters input to the model include the interfacial thermal resistance extracted from the transient thermal response and the magnetic field distribution entropy extracted from the near-field magnetic distribution; the predicted electrical performance dataset output by the multiphysics coupling model includes the predicted inductance value, and the specific calculation process is broken down as follows: Using the nominal inductance value as a reference, the inductance attenuation caused by thermal degradation and magnetic degradation is subtracted sequentially to obtain the corrected prediction result, i.e., the predicted inductance value.

[0027] Furthermore, step two involves calculating the predicted inductance value using a parameterized multiphysics coupling model derived from electromagnetic field theory and simplified by model order reduction techniques, hereinafter referred to as Lpred; The calculation logic of Lpred begins with the ideal benchmark of the nominal inductance value (hereinafter referred to as Lnom), and then subtracts the inductance decay caused by thermal degradation and magnetic degradation, respectively, namely the thermal inductance decay (hereinafter referred to as ΔLth) and the magnetic inductance decay (hereinafter referred to as ΔLmag). This linear subtraction structure clearly demonstrates in the parameter positions that the negative impacts of the two physical degradation paths on the total inductance value are decoupled and additive. Each inductance decay is driven by its corresponding internal physical state parameter through a linear mapping relationship. ΔLth is the product of the increment of the interfacial thermal resistance (hereinafter referred to as Rthint) and the thermal degradation sensitivity coefficient (hereinafter referred to as kth), while ΔLmag is the product of the increment of the spatial distribution entropy of the leakage magnetic field (hereinafter referred to as Hmag) and the magnetic degradation sensitivity coefficient (hereinafter referred to as kmag). To further quantify the technical effect, the predicted electrical degradation degree (hereinafter referred to as Dpred) is introduced. The predicted inductance attenuation is normalized, and the Dpred value range is limited to the interval [0, 1). As the output value of Dpred approaches 0, the predicted inductance decay is zero, making Lpred exactly equal to Lnom. According to the inherent logic of the analytical function of the multiphysics coupling model, this result necessarily requires that Rthint and Hmag, which are inputs to the multiphysics coupling model, are both in their initial healthy state, i.e., the increment is zero. This situation accurately maps the intact internal physical state to the completely healthy external electrical performance. When the output value of Dpred approaches the middle value, this embodiment sets the middle value at 0.5, indicating that the inductor has suffered moderate physical damage and has caused measurable electrical performance degradation. At this time, the multiphysics coupling model predicts that Lpred has degraded based on the increment of the input Rthint or Hmag, but has not yet reached the critical point of failure. When the output value of Dpred approaches 1, the predicted total inductance decay is approaching the failure criterion used for normalization, indicating that the predicted electrical performance is close to its end of life. This is because internal physical damage, including the increase of Rthint or Hmag, through the amplification effect of their respective kth and kmag, mathematically leads to Lpred approaching its failure threshold. To make the results of Dpred more diagnostically valuable, its value range is divided into intervals: 1) According to inductor application specifications, a 20% decrease in inductance value is considered a failure; statistical analysis of historical simulation and experimental data shows that Dpred is a critical node for state transition around 0.05 and 0.15. 2) The preliminary standard is set as follows: [0, 0.05] is "normal predictive performance"; (0.05, 0.15) is "predictive performance warning"; (0.15, 1.0) is "predictive performance dangerous"; by substituting historical physical damage data into the model for backtracking verification, different health levels are distinguished, and the setting of the "predictive performance warning" interval is to provide sufficient advance notice. 3) The verification results show that the standard is reasonable and effective. Therefore, the above three-level classification standard is finally determined for subsequent health status assessment and residual analysis. Step three specifically includes: A multi-dimensional physical-electrical residual vector is calculated to simultaneously quantify the prediction accuracy of the multiphysics coupling model across multiple electrical performance dimensions. Each component of the physical-electrical residual vector is defined as the difference between the actual electrical performance parameters in the first dataset and the theoretical predicted performance in the predicted electrical performance dataset. The step of calculating the physical electrical residual vector also includes recording and storing historical trajectories, specifically the historical trajectory of the evolution of the physical electrical residual vector over time, so as to provide a data basis for identifying the time evolution trend of the failure mechanism in subsequent steps.

[0028] Furthermore, step three shifts the focus of monitoring from "whether performance has degraded" to "whether performance has degraded as expected" by calculating the residual between the actual electrical performance and the theoretically predicted performance based on the physical model. This quantification of the deviation captures and amplifies hidden or emerging failure mechanisms that the physical model fails to cover. The core algorithm of this step lies in calculating and quantifying the physical electrical residual vector (hereinafter referred to as ccWD). The initial source of this algorithm is the model baseline fault diagnosis method in modern control theory. The specific calculation logic is as follows: calculate each component of the residual vector, that is, the inductance residual component (hereinafter referred to as ccL) is equal to the measured inductance value (hereinafter referred to as Lmeas) minus Lpred, and the resistance residual contribution component (hereinafter referred to as ccR) is equal to the measured equivalent series resistance value (hereinafter referred to as Resrmeas) minus the predicted equivalent series resistance value (hereinafter referred to as Resrpred); then, in order to obtain a comprehensive scalar index, calculate the normalized residual norm (hereinafter referred to as Enorm), which is equal to the Euclidean norm of the inductance residual component (hereinafter referred to as ccL) and the resistance residual contribution component (hereinafter referred to as ccR) after normalizing them with their corresponding failure thresholds; In this algorithm, Lmeas and Resrmeas are paired inputs with Lpred and Resrpred; they are placed in the subtraction operation, which is the fundamental operation of residual analysis, designed to extract "unexpected" information; ccL and ccR are intermediate results, forming a point ccWD in the multidimensional diagnostic space; the final output Enorm is a measure of the distance from this point to the origin, which compresses the multidimensional deviation information into a single, easily judged scalar; this structure from multidimensional input to a single scalar output optimally achieves the technical goal of extracting key fault severity information from complex deviation phenomena. Through normalization, the output range of the normalized Enorm is mapped to the interval [0, 1]. As the Enorm output approaches 0, it means that Lmeas = Lpred and Resrmeas = Resrpred; this indicates that all the electrical performance degradation of the current inductor can be fully explained by the multiphysics coupling model based on the internal physical state in step two. When the Enorm output approaches 1, it means that the absolute value of at least one of the parameters cCL and cCR approaches 1. Specifically, if an inter-turn short circuit occurs, it will cause Lmeas to decrease. At this time, the physical state parameters Rthint and Hmag have not changed, so Lpred remains at a high level. This will result in cCL being -1, but the absolute value of cCL is 1, which in turn makes Enorm approach 1.

[0029] Step four specifically includes: The physical electrical residual vector is mapped to a preset two-dimensional failure feature space composed of mutually orthogonal failure mode basis vectors. By calculating the projection components of the physical electrical residual vector onto each failure mode basis vector, the contribution of different failure modes to the current performance deviation is determined, and a diagnostic indication signal that can quantify the current dominant failure mechanism is generated accordingly.

[0030] Step four also includes: The failure mode basis vectors are determined in advance by conducting accelerated aging tests and failure physics analysis on inductors of the same type as the inductor under test. One basis vector points to the direction of solder joint fatigue or terminal oxidation failure mode, which is mainly characterized by the increase of equivalent series resistance, while the other basis vector points to the direction of core material aging or structural cracking failure mode, which is mainly characterized by the decrease of inductance. The diagnostic indication signal is a failure mode bias index, and the calculation process of the failure mode bias index specifically includes: The absolute value of the inductor residual is divided by a preset inductor failure threshold to obtain the normalized inductor residual; the absolute value of the resistance residual contribution is divided by a preset resistance failure threshold to obtain the normalized resistance residual; the normalized resistance residual is defined as the resistance residual contribution, characterizing the severity of solder joint fatigue failure modes; the normalized inductor residual is defined as the inductor residual contribution, characterizing the severity of core degradation failure modes; the difference between the resistance residual contribution and the inductor residual contribution is divided by the sum of the resistance residual contribution and the inductor residual contribution to obtain the failure mode bias index.

[0031] Furthermore, the failure mode bias index (hereinafter referred to as Ifm) is calculated to generate a diagnostic indication signal. The specific calculation process is as follows: Ifm is equal to the difference between the resistance residual contribution (hereinafter referred to as CR) and the inductance residual contribution (hereinafter referred to as CL), divided by the sum of the two. In the above algorithm structure, the inputs CR and CL represent the relative severity of two orthogonal failure modes, namely resistive degradation and inductive degradation. They are placed in the numerator by a subtraction operation, i.e., CR-CL, which is a contrast enhancement operator. Its function is to polarize the difference between the two contributions: when the two are close, the result tends to zero, and when one is much larger than the other, the result tends to be extreme. In the denominator, they are placed in an addition operation, i.e., CR+CL, which is a normalization operator. Its function is to eliminate the influence of the overall size of the residuals, so that the exponent only reflects the relative proportion of the two failure modes, rather than the absolute severity. This "difference ratio sum" structure optimally achieves the technical goal of separating and quantifying the dominant failure mode from the mixed residual signal. The trend of the algorithm function Ifm(CR, CL) is as follows: when CR increases, Ifm tends to increase; when CL increases, Ifm tends to decrease. This opposite trend is the basis of diagnostic classification. When the contribution of a failure mode is absolutely dominant, the exponent will saturate and point to an extreme value, while when the contributions of the two are equal, the exponent will stabilize near zero, thus constructing a continuous spectrum from -1 to +1 that can clearly map the bias of failure modes. Using the mathematical structure of the normalized difference index described above, the output range of Ifm is limited to the interval [-1, 1]. When the Ifm output approaches 1, it indicates that the current dominant failure mechanism is more clearly biased towards resistive degradation, including solder joint fatigue and terminal oxidation; at the same time, CR > CL, making the numerator (CR-CL) approximately equal to the denominator (CR+CL); this physically means that the absolute value of the contribution of the resistive residual |ccR| relative to its failure threshold ThR is much greater than the absolute value of the inductor residual |ccL| relative to its threshold ThL. When Ifm equals -1, it indicates that the dominant failure mechanism is more clearly biased towards inductive degradation, including core aging and structural cracking; at the same time, CL > CR, and the value of CL is much greater than the value of CR, making the numerator (CR-CL) approximately equal to the negative value of the denominator (CR+CL); this physically means that the relative severity of the inductance residual is far greater than the contribution of the resistance residual, thus pointing the source of the fault to the physical failure affecting the integrity of the magnetic circuit structure; When Ifm approaches 0, it indicates that the current failure mechanism is a mixed mode of resistive and inductive degradation, or no obvious dominant mode; at the same time, CR=CL, and the numerator is equal to zero. This indicates that the contributions of the two failure modes are equal, or neither contributes. This result provides key information for subsequent maintenance decisions: "complex failure mode" or "fault is not yet clear". The normalization thresholds are set as follows: ThL=20µH, ThR=10mΩ; see Table 1 below for details. Table 1: Examples of failure mechanism diagnosis under different typical unknown failure scenarios ; In the table above, for "pure solder joint fatigue failure", since the contribution of the resistance residual CR is greater than the contribution of the inductance residual CL, the calculated Ifm is 0.8, close to 1, which accurately points to resistive failure; conversely, in "pure core cracking failure", since the contribution of the resistance residual CR is less than the contribution of the inductance residual CL, Ifm is -0.94, which accurately points to inductive failure. For "early inter-turn short circuit faults", it causes both a decrease in inductance and an increase in resistance, but the change in inductance is more drastic, with Ifm being -0.54, indicating a mixed mode dominated by inductive degradation; for "complex mixed aging faults", Ifm is -0.20, close to 0, indicating that the two modes contribute equally; by constructing the Ifm index, the two-dimensional residual vector information is losslessly mapped to a one-dimensional diagnostic coordinate axis with clear physical meaning, realizing the quantitative classification of the root cause of unknown faults; To transform the continuous values ​​of Ifm into discrete, definitive diagnostic conclusions, its value range [-1, 1] is divided into intervals: 1) Based on their experience in failure physics analysis, domain experts believe that an absolute index value greater than 0.7 can confirm a single dominant mode. Cluster analysis of a large amount of accelerated aging test data also shows a clear mode boundary around -0.6 and +0.6. 2) The preliminary criteria were defined as follows: [-1.0, -0.6] for "inductive failure predominates"; (-0.6, 0.6) for "mixed failure or unclear mode"; and [0.6, 1.0] for "resistive failure predominates". Cross-validation using a historical dataset containing known fault labels achieved a 96% diagnostic classification accuracy. 3) Based on the high accuracy of the verification results, the above three-level diagnostic classification criteria were finally determined and used as the core logic for generating the final diagnostic report and guiding maintenance strategies; Step five specifically includes: In response to the diagnostic indication signal, an adaptive adjustment operation is performed; the adaptive adjustment operation maps the value of the diagnostic indication signal to a preset decision matrix of associated failure mechanisms and test strategies. The decision matrix pre-sets the optimal stress combination corresponding to different numerical ranges of the diagnostic indication signal, including increasing the amplitude and frequency of temperature cycling when the diagnostic indication signal indicates that the solder joint fatigue is the main factor, and increasing the magnitude of the DC bias current when the indication core aging is the main factor.

[0032] Step five also includes: The decision matrix selectively performs at least one of the following: 1. Output a stress adjustment command to dynamically change the type or intensity of stress applied to the inductor under test, thereby accelerating aging in response to the current dominant failure mechanism; 2. From a model library containing multiple failure mode-specific degradation models, select a degradation model that matches the dominant failure mechanism characterized by the diagnostic indication signal to update the prediction model used to predict its remaining useful life.

[0033] Furthermore, based on the decision matrix mapping of Ifm, the input value of Ifm is compared with multiple preset numerical ranges in the decision matrix; once the range to which it belongs is determined, the system outputs the preset stress adjustment command (hereinafter referred to as Cmdstress) and the selected remaining service life model (hereinafter referred to as MRUL) corresponding to that range. In the above algorithm structure, Ifm is the only decision input, located at the beginning of the logic chain; Cmdstress and MRUL are parallel and independent outputs. The value of Ifm directly determines the direction of the system behavior. Its force lies in selection rather than calculation, ensuring that the system response is deterministic and interpretable. This structured arrangement optimally achieves the technical goal of transforming diagnostic insights into precise intervention. Step 5 utilizes the value of Ifm in the interval [-1, 1] to trigger different adaptive adjustment operations with clear technical objectives; When the value of Ifm is in the positive range, i.e. [0.6, 1.0], the corresponding trend is that the intervention strategy of the system focuses more on accelerating and simulating resistive degradation; an Ifm value close to 1 clearly indicates that resistive failures such as solder joint fatigue are dominant, and the decision matrix outputs instructions accordingly, including increasing the temperature cycle amplitude (hereinafter referred to as ΔTcycle); this is because temperature cycling is the principal stress that induces solder joint creep and fatigue, and at the same time, it switches to MRUL based on the thermomechanical fatigue model; When the value of Ifm is in the negative range, i.e. [-1.0, -0.6], the corresponding trend of the invention content is that the intervention strategy of the system focuses more on accelerating and simulating inductive degradation; an Ifm value close to -1 clearly indicates that inductive failures such as core aging are dominant; the decision matrix will output instructions, including increasing the DC bias current (hereinafter referred to as IDCbias), because a high IDCbias will exacerbate the magnetic saturation and thermal aging effects of the core; at the same time, switch to MRUL based on the aging kinetic model of magnetic materials; Step five, to ensure system stability and avoid overreacting to minor, uncertain diagnostic signals, further includes a processing step: The absolute value of Ifm is compared with a preset decision response threshold. The decision response threshold is set based on the analysis of the statistical distribution of diagnostic signals caused by measurement noise and model error in the system under healthy conditions. The goal is to ensure that the threshold can effectively distinguish between real failure mode signals and background noise fluctuations. If the absolute value of Ifm is less than the decision response threshold, the system will ignore the diagnostic signal in subsequent adaptive adjustment operations and maintain the current or baseline stress application strategy and RUL prediction model unchanged. To construct the decision matrix, the range of Ifm [-1, 1] needs to be divided into intervals according to the SMART principle, specifically as follows: 1) Based on physical analysis of a large number of failure samples, experts defined three modes: “dominant,” “biased,” and “mixed.” Data analysis showed that the absolute values ​​of the exponents around 0.2 and 0.6 were the key inflection points for mode differentiation. 2) The initial standard is set at five levels: [-1.0, -0.6] is "inductively dominant"; (-0.6, -0.2) is "leaning towards inductive"; [-0.2, 0.2] is "mixed / undefined"; (0.2, 0.6) is "leaning towards resistivity"; [0.6, 1.0] is "resistivity-dominant". Through backtesting of historical data, this five-level division can accurately correspond to different stress adjustment levels and model selection strategies. 3) Based on the verification results with high resolution and high matching rate, the above five-level classification criteria were finally determined and solidified in the decision matrix as the logical basis for achieving precise adaptive adjustment.

[0034] Example 3: Please see Figure 3 An inductor lifespan assessment system, comprising: The multimodal data synchronization module is used to simultaneously acquire data while applying a preset electrical excitation to the inductor under stress. The first dataset characterizes the actual electrical performance of the inductor under test, and the second dataset characterizes the internal physical state of the inductor under test; The physical model derivation module, based on a pre-defined multi-physics coupling model that maps internal physical states to electrical performance, uses a second dataset to generate a predicted electrical performance dataset for the inductor under test. The residual quantization analysis module is used to calculate the residual between the first dataset and the predicted electrical performance dataset to quantify the deviation between the actual electrical performance and the predicted electrical performance of the inductor under test. The failure mechanism diagnosis module identifies the dominant failure mechanism of the inductor under test based on the characteristics of the residual and generates a diagnostic indication signal characterizing the dominant failure mechanism. The closed-loop feedback optimization module is used to respond to diagnostic indication signals and perform adaptive adjustment operations, which include: Adjust the type of stress applied to the inductor under test and update the prediction model used to predict the remaining lifespan of the inductor under test.

[0035] It should be noted that all calculation formulas in this application employ regression analysis, including but not limited to machine learning algorithms, to deeply analyze the collected parameters and identify their natural trends and interrelationships. Specialized software, such as Python's Scikit-learn library or the R language, is used to automatically generate mathematical models that match the data. Then, cross-validation and other methods are used to objectively evaluate the model performance, and continuous feedback and optimization are combined to ensure that the created formulas truly reflect the inherent laws of the data, thereby guaranteeing their effectiveness and accuracy. In all calculation formulas in this application, the parameters in each formula undergo dimensionless processing within a consistent range to ensure that different physical quantities are compared on the same scale; dimensionless processing techniques include, but are not limited to, min-max-normalization and Z-score standardization. The technical solution of this invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random-access memory (RAM), flash memory, hard disk, or optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of various embodiments of this invention.

[0036] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-including system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0037] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method and system for evaluating the lifespan of an inductor, characterized in that, The specific steps include: Step 1: While applying a preset electrical excitation to the inductor under stress, simultaneously acquire: The first dataset characterizes the actual electrical performance of the inductor under test, and the second dataset characterizes the internal physical state of the inductor under test; Step 2: Based on a pre-defined multiphysics coupling model that maps internal physical states to electrical performance, generate a predicted electrical performance dataset for the inductor under test using the second dataset. Step 3: Calculate the residual between the first dataset and the predicted electrical performance dataset to quantify the deviation between the actual electrical performance and the predicted electrical performance of the inductor under test; Step 4: Based on the characteristics of the residual, identify the current dominant failure mechanism of the inductor under test and generate a diagnostic indication signal characterizing the dominant failure mechanism; Step 5: In response to the diagnostic indication signal, perform adaptive adjustment operations, which include: Adjust the type of stress applied to the inductor under test and update the prediction model used to predict the remaining lifespan of the inductor under test.

2. The method for evaluating the service life of an inductor according to claim 1, characterized in that: Step one specifically includes: While applying a preset electrical excitation to the inductor under stress, differential measurement is performed through a self-calibrating test fixture that integrates a reference path parallel to the test path of the inductor under test to calibrate in real time and obtain the first dataset used to characterize the actual electrical performance of the inductor under test. Simultaneously, a non-contact physical field imaging method is used to acquire physical characteristic parameters that reflect the health of the internal structure of the inductor under test, which are used as the second dataset; A passive device with stable performance under stress is deployed in the reference path to capture and compensate for the common-mode error of the measurement system introduced by environmental stress in real time; the non-contact physical field imaging method includes: using a high-speed infrared thermal imager to capture the transient thermal response of the inductor under test when responding to power pulse excitation, and using a three-dimensional magnetic field probe array to draw a spatial distribution map of the near-field magnetic field.

3. The inductor lifespan evaluation method according to claim 2, characterized in that: Step two specifically includes: Using the physical feature parameters in the second dataset as input to a pre-established parameterized reduced-order multiphysics coupling model, the multiphysics coupling model maps the physical feature parameters to the corresponding electrical performance parameters, thereby generating a predicted electrical performance dataset for the inductor under test. The multiphysics coupling model is obtained by performing high-fidelity finite element simulation on the inductor under test in the offline stage, scanning different physical damage levels and combinations of operating parameters in the simulation, recording their corresponding electrical performance responses, and finally fitting the analytical function for real-time calculation through model order reduction technology trained by neural network.

4. The inductor lifespan assessment method according to claim 3, characterized in that: Step two also includes: The internal physical state parameters input to the multiphysics coupling model include the interfacial thermal resistance extracted from the transient thermal response and the magnetic field distribution entropy extracted from the near-field magnetic distribution; the predicted electrical performance dataset output by the multiphysics coupling model includes the predicted inductance value, and the specific calculation process is broken down as follows: Using the nominal inductance value as a reference, the inductance attenuation caused by thermal degradation and magnetic degradation is subtracted sequentially to obtain the corrected prediction result, i.e., the predicted inductance value.

5. The inductor lifespan evaluation method according to claim 4, characterized in that: Step three specifically includes: A multi-dimensional physical-electrical residual vector is calculated to simultaneously quantify the prediction accuracy of the multiphysics coupling model across multiple electrical performance dimensions. Each component of the physical-electrical residual vector is defined as the difference between the actual electrical performance parameters in the first dataset and the theoretical predicted performance in the predicted electrical performance dataset. The step of calculating the physical electrical residual vector also includes recording and storing historical trajectories, specifically the historical trajectory of the evolution of the physical electrical residual vector over time, so as to provide a data basis for identifying the time evolution trend of the failure mechanism in subsequent steps.

6. The inductor lifespan evaluation method according to claim 5, characterized in that: Step four specifically includes: The physical electrical residual vector is mapped to a preset two-dimensional failure feature space composed of mutually orthogonal failure mode basis vectors. By calculating the projection components of the physical electrical residual vector onto each failure mode basis vector, the contribution of different failure modes to the current performance deviation is determined, and a diagnostic indication signal that can quantify the current dominant failure mechanism is generated accordingly.

7. The method for evaluating the service life of an inductor according to claim 6, characterized in that: Step four also includes: The failure mode basis vectors are determined in advance by conducting accelerated aging tests and failure physics analysis on inductors of the same type as the inductor under test. One basis vector points to the direction of solder joint fatigue or terminal oxidation failure mode, which is mainly characterized by the increase of equivalent series resistance, while the other basis vector points to the direction of core material aging or structural cracking failure mode, which is mainly characterized by the decrease of inductance. The diagnostic indication signal is a failure mode bias index, and the calculation process of the failure mode bias index specifically includes: The actual inductance value of the inductor under test measured in real time by the self-calibrating test fixture in the first dataset is compared with the theoretical inductance value predicted by the multiphysics coupling model in the second dataset. The difference between the two is defined as the inductance residual. The absolute value of the inductance residual is divided by a preset inductance failure threshold to obtain the normalized inductance residual; the absolute value of the resistance residual contribution is divided by a preset resistance failure threshold to obtain the normalized resistance residual; the normalized resistance residual is defined as the resistance residual contribution, characterizing the severity of solder joint fatigue failure modes; and the normalized inductance residual is defined as the inductance residual contribution, characterizing the severity of core degradation failure modes. The failure mode bias index is obtained by subtracting the inductance residual contribution from the resistance residual contribution and dividing the difference by the sum of the resistance residual contribution and the inductance residual contribution.

8. The inductor lifespan assessment method according to claim 7, characterized in that: Step five specifically includes: In response to the diagnostic indication signal, an adaptive adjustment operation is performed; the adaptive adjustment operation maps the value of the diagnostic indication signal to a preset decision matrix of associated failure mechanisms and test strategies. The decision matrix pre-sets the optimal stress combination corresponding to different numerical ranges of the diagnostic indication signal, including increasing the amplitude and frequency of temperature cycling when the diagnostic indication signal indicates that the solder joint fatigue is the main factor, and increasing the magnitude of the DC bias current when the indication core aging is the main factor.

9. The method for evaluating the service life of an inductor according to claim 8, characterized in that: Step five also includes: The decision matrix selectively performs at least one of the following:

1. Output a stress adjustment command to dynamically change the type or intensity of stress applied to the inductor under test, thereby accelerating aging in response to the current dominant failure mechanism; 2. From a model library containing multiple failure mode-specific degradation models, select a degradation model that matches the dominant failure mechanism characterized by the diagnostic indication signal to update the prediction model used to predict its remaining useful life.

10. An inductor lifespan assessment system, characterized in that: The system is used to execute an inductor lifespan assessment method according to any one of claims 1-9, comprising: The multimodal data synchronization module is used to simultaneously acquire data while applying a preset electrical excitation to the inductor under stress. The first dataset characterizes the actual electrical performance of the inductor under test, and the second dataset characterizes the internal physical state of the inductor under test; The physical model derivation module, based on a pre-defined multi-physics coupling model that maps internal physical states to electrical performance, uses a second dataset to generate a predicted electrical performance dataset for the inductor under test. The residual quantization analysis module is used to calculate the residual between the first dataset and the predicted electrical performance dataset to quantify the deviation between the actual electrical performance and the predicted electrical performance of the inductor under test. The failure mechanism diagnosis module identifies the dominant failure mechanism of the inductor under test based on the characteristics of the residual and generates a diagnostic indication signal characterizing the dominant failure mechanism. The closed-loop feedback optimization module is used to respond to diagnostic indication signals and perform adaptive adjustment operations, which include: Adjust the type of stress applied to the inductor under test and update the prediction model used to predict the remaining lifespan of the inductor under test.

Images