A rectifier bridge aging life prediction method and early warning system

Abstract

The application belongs to the technical field of power electronics, and provides a rectifier bridge aging life prediction method and early warning system, which comprises the following steps: collecting the actual operating current waveform of the input end of the rectifier bridge in the target application scene, extracting characteristic parameters and generating a dynamic current loading sequence to simulate two types of thermal stress, namely transient thermal shock and cumulative temperature rise under intermittent working conditions; synchronously collecting the forward voltage drop and junction-to-case thermal resistance during the dynamic loading process to form degradation trajectories respectively to separate and track two failure modes, namely bondwire thermal fatigue and solder layer creep fatigue; comparing the pulse cycle numbers required for both to reach the failure threshold, automatically identifying the failure site that occurs first, extracting the corresponding degradation data, combining the transient junction temperature and thermal runaway risk markers to form a standardized training data set, selecting a matched life evaluation model and inputting the target working condition parameters to output the remaining useful life, and triggering a graded warning according to the calculation result to effectively improve the prediction accuracy.

Description

Technical Field

[0001] This invention belongs to the field of power electronics technology, specifically based on a method for predicting the aging life of a rectifier bridge and an early warning system. Background Technology

[0002] Currently, rectifier bridge life prediction tests typically employ a constant current loading method, which involves applying a constant DC current throughout the test and estimating the lifespan by monitoring the degradation trend of parameters such as forward voltage drop over time. However, in actual use, rectifier bridges operate in an intermittent state of repeated switching on and off, rather than continuous conduction. Under intermittent operating conditions, there are two independent types of thermal stress—the peak current generated during each conduction generates transient thermal shock, and the cumulative effect of multiple conductions over a long period generates an average temperature rise. Constant current loading can only generate steady-state thermal stress and cannot reflect the above two types of thermal effects, leading to a disconnect between the test conditions and actual operating conditions.

[0003] Furthermore, the rectifier bridge contains two key interconnecting components: bonding wires and solder layers. Two types of thermal stress act on different components—transient thermal shock mainly causes thermal fatigue of the bonding wires, manifested as a gradual increase in forward voltage drop; average temperature rise mainly causes creep fatigue of the solder layer, manifested as a gradual increase in junction-shell thermal resistance. Existing lifetime prediction methods typically collect all monitoring data in an aliased manner, making it impossible to distinguish the degradation processes of the two types of damage, and difficult to determine which failure mode reaches the critical point first and thus the final end of the device's lifespan. Moreover, the dominant failure modes differ under different application conditions—bonding wires often fail first under conditions of high peak value and low duty cycle, while solder layers often fail first under conditions of low peak value and high duty cycle. Existing methods lack an automatic identification mechanism and usually model a single failure mode. When the preset mode does not match the actual situation, the prediction results are severely biased. At the same time, bonding wire fatigue is described by the Coffin-Manson model, while solder layer fatigue is described by the Darveaux model. The physical mechanisms of the two types of failure are different, and a single model cannot accurately describe the degradation processes of the two different mechanisms.

[0004] Therefore, the present invention provides a method for predicting the aging life of a rectifier bridge and an early warning system. Summary of the Invention

[0005] In order to overcome the shortcomings of the prior art, at least one technical problem raised in the background art is solved.

[0006] The technical solution adopted by this invention to solve its technical problem is: a method for predicting the aging life of a rectifier bridge, comprising the following steps:

[0007] S1: Collect the actual operating current waveform at the input of the rectifier bridge in the target application scenario, extract three characteristic parameters: peak current, duty cycle and conduction frequency, and generate a dynamic current loading sequence.

[0008] S2: Apply a dynamic current loading sequence to the AC input terminal of the rectifier bridge under test. During the loading process, collect the forward voltage drop, junction-to-case thermal resistance, reverse leakage current and chip transient junction temperature to form the degradation trajectory of the forward voltage drop and junction-to-case thermal resistance as a function of the number of pulse cycles.

[0009] S3: Based on the degradation trajectory, by comparing the number of pulse cycles required to reach failure by the forward voltage drop and the junction-shell thermal resistance, the failure site that occurs first under loading conditions is determined, including bond wire thermal fatigue and solder layer creep fatigue; when the reverse leakage current shows a sudden acceleration, it is marked as a risk of thermal runaway and the corresponding number of pulse cycles is recorded.

[0010] S4: Based on the first failure location, extract complete degradation data from the initial stage to the failure and the number of pulse cycles from the corresponding degradation trajectory, supplemented by transient junction temperature data and thermal runaway risk markers, to form a standardized training dataset;

[0011] S5: Select a matching lifetime assessment model based on the feature parameter type of the training dataset, input the feature parameters under the target working condition into the model, and output the number of pulse cycles required to reach the failure threshold under the target working condition.

[0012] As a further aspect of the present invention: in step S1, the target application scenario is the equipment or system in which the rectifier bridge is finally installed and used;

[0013] The process of collecting the actual operating current waveform at the input terminal of the rectifier bridge in the target application scenario is as follows: the data is collected on the AC input side of the rectifier bridge, so that the device under test is running normally in the target application scenario, covering a complete working cycle, and the complete waveform data of the current flowing through the input terminal of the rectifier bridge changing with time is continuously recorded during the working cycle period.

[0014] As a further aspect of the present invention: In step S1, the process of extracting the three characteristic parameters—peak current, duty cycle, and conduction frequency—and generating a dynamic current loading sequence is as follows:

[0015] The acquired continuous current waveform is segmented according to the conduction stage, with each conduction stage being a pulse. The maximum current value reached during the pulse is read as the peak current. The proportion of conduction time to total time within a complete working cycle is calculated as the duty cycle. The number of conduction pulses per unit time is calculated as the conduction frequency. The extracted three parameter values ​​are input into the programmable test power supply, which outputs the corresponding current waveform according to the set peak current, duty cycle, and conduction frequency, thus obtaining the dynamic current loading sequence.

[0016] As a further aspect of the present invention: In step S2, the process of forming the degradation trajectory of the positive voltage drop and junction-shell thermal resistance as a function of the number of pulse cycles is as follows:

[0017] Using the number of pulse cycles as the horizontal axis, record the measured forward voltage drop value and the corresponding number of pulse cycles for each pulse conduction, and mark the data points on the coordinate plane.

[0018] For each junction-shell thermal resistance measurement, record the measured junction-shell thermal resistance value and the corresponding number of pulse cycles, and mark the data points on the coordinate plane;

[0019] Connecting the data points of the forward voltage drop sequentially forms the degradation trajectory of the forward voltage drop, and connecting the data points of the junction-shell thermal resistance sequentially forms the degradation trajectory of the junction-shell thermal resistance.

[0020] As a further aspect of the present invention: in step S2, the forward voltage drop is obtained when the rectifier bridge is in the forward conduction state, the junction-to-case thermal resistance is obtained during the pulse turn-off interval, and the transient junction temperature of the chip is obtained by the thermistor method at the end of each pulse conduction period.

[0021] As a further aspect of the present invention: In step S3, the process of determining the failure location that first occurs under loading conditions is as follows:

[0022] Using the rate of change of the initial value of the forward voltage drop reaching a preset threshold as the voltage drop failure criterion, the pulse cycle number corresponding to the first time the forward voltage drop reaches the voltage drop failure criterion is found in the degradation trajectory of the forward voltage drop and marked as the voltage drop failure cycle number.

[0023] The rate of change of the initial value of junction-shell thermal resistance reaching a preset threshold is used as the criterion for thermal resistance failure. The number of pulse cycles corresponding to the first time the junction-shell thermal resistance reaches the thermal resistance failure criterion is found in the degradation trajectory of junction-shell thermal resistance and marked as the number of thermal resistance failure cycles.

[0024] By comparing the number of failure cycles of pressure drop with the number of failure cycles of thermal resistance, the failure site corresponding to the smaller one is the failure site that occurs first under loading conditions.

[0025] As a further aspect of the present invention: in step S3, the process of marking the risk of thermal runaway is as follows:

[0026] Based on the reverse leakage current data, a curve of the reverse leakage current changing with the number of pulse cycles is constructed, a measurement window is set, and the total increment of the reverse leakage current within the current measurement window is calculated.

[0027] If the total increase in reverse leakage current within the current measurement window exceeds a preset multiple of the total increase in reverse leakage current within the previous measurement window, it is determined to be a sudden acceleration, marked as a thermal runaway risk node, and the corresponding number of pulse cycles is recorded.

[0028] As a further aspect of the present invention: in step S4, the process of forming a standardized training dataset is as follows:

[0029] Based on the judgment result of S3, the corresponding degradation trajectory is selected as the data extraction object; if a thermal runaway risk node is marked, the number of pulse cycles corresponding to the thermal runaway risk node is used as the cutoff position, and the degradation data before the cutoff position is extracted and included in the dataset; the number of pulse cycles and feature parameter values ​​corresponding to each data point are extracted and arranged in ascending order of the number of cycles to form a complete degradation sequence; the degradation sequence is arranged in a four-column format of pulse cycle number, feature parameter value, transient junction temperature, and thermal runaway risk mark to form a standardized training dataset.

[0030] As a further aspect of the present invention: in step S5, the process of outputting the remaining service life is as follows:

[0031] The failure location was determined based on the feature parameter type of the training dataset. The Coffin-Manson model was used for thermal fatigue of the bond wire, and the Darveaux model was used for creep fatigue of the solder layer.

[0032] The degradation trajectory from initial to failure is divided into multiple stages. For each stage, the corresponding temperature parameters and pulse cycle increments are measured. The data sets from multiple stages are substituted into the selected model for parameter calibration. A data set from a stage that was not calibrated is used as a validation set for deviation verification. After successful verification, the temperature parameters under the target operating condition are input into the calibrated model, and the number of pulse cycles required to reach the failure threshold under the target operating condition is output. A graded warning is triggered based on the deviation between the calculated result and the expected value.

[0033] A rectifier bridge aging life early warning system includes the following modules:

[0034] Dynamic loading module: Collect the actual operating current waveform at the input of the rectifier bridge in the target application scenario, extract three characteristic parameters: peak current, duty cycle and conduction frequency, and generate a dynamic current loading sequence;

[0035] Degradation trajectory acquisition module: Apply a dynamic current loading sequence to the AC input terminal of the rectifier bridge under test. During the loading process, the forward voltage drop, junction-to-case thermal resistance, reverse leakage current and chip transient junction temperature are acquired to form degradation trajectories of the forward voltage drop and junction-to-case thermal resistance as a function of the number of pulse cycles.

[0036] Failure Location Determination Module: Based on the degradation trajectory, the module determines the failure location that occurs first under loading conditions by comparing the number of pulse cycles required to reach failure by comparing the forward voltage drop and the junction-shell thermal resistance. This includes bond wire thermal fatigue and solder layer creep fatigue. When the reverse leakage current suddenly accelerates, it is marked as a risk of thermal runaway and the corresponding number of pulse cycles is recorded.

[0037] Training data generation module: Based on the first failure location, extract complete degradation data from the initial stage to failure and the number of pulse cycles from the corresponding degradation trajectory, supplemented by transient junction temperature data and thermal runaway risk markers, to form a standardized training dataset;

[0038] The remaining life estimation module selects a matching life assessment model based on the feature parameter types of the training dataset, inputs the feature parameters under the target operating condition into the model, and outputs the number of pulse cycles required to reach the failure threshold under the target operating condition.

[0039] The beneficial effects of this invention are as follows: By acquiring actual operating current waveforms and generating dynamic current loading sequences with independently adjustable peak values ​​and duty cycles, the test loading conditions are no longer constant current, but rather a composite waveform capable of simulating transient thermal shock and cumulative temperature rise under intermittent operating conditions. This solves the problem of the disconnect between traditional constant current loading and real intermittent operating conditions, making the test conditions closer to actual conditions and the lifetime prediction results more reliable. By simultaneously acquiring forward voltage drop and junction-shell thermal resistance and forming their respective degradation trajectories, the two failure modes—bond wire thermal fatigue and solder layer creep fatigue—have their own independent quantitative indicators, solving the problem of the inability to distinguish between the two types of damage data in traditional methods, and providing a data basis for accurately determining the location of the first failure. By comparing the number of pulse cycles required for the forward voltage drop and junction-shell thermal resistance to reach the failure threshold, different operating conditions can be automatically identified. The method identifies the first failure location, solving the problem of inaccurate predictions caused by presupposing a single failure mode in traditional methods. This ensures that subsequent model selection is consistent with the actual failure mechanism, avoiding prediction deviations caused by misjudgment of failure modes. Based on the identified failure location, a life assessment model matching its physical mechanism is selected—the Coffin-Manson model based on the thermal cycling plastic strain accumulation mechanism is used for bond line fatigue, and the Darveaux model based on the high-temperature creep deformation accumulation mechanism is used for solder layer fatigue. This solves the problem that a single model cannot accurately describe the degradation process of two different physical mechanisms, improving the accuracy of life prediction and the model's generalization ability. After calculating the remaining service life, a graded early warning is triggered based on the risk level of the remaining service life, providing an intuitive basis for engineering maintenance decisions and realizing a complete closed loop from life prediction to engineering early warning. Attached Figure Description

[0040] The invention will now be further described with reference to the accompanying drawings.

[0041] Figure 1 This is a flowchart of the steps in a rectifier bridge aging life prediction method according to Embodiment 1 of the present invention;

[0042] Figure 2 This is a schematic diagram of the dynamic current loading sequence generated in the rectifier bridge aging life prediction method of Embodiment 1 of the present invention;

[0043] Figure 3 This is a schematic diagram of the positive voltage drop degradation trajectory generated in the rectifier bridge aging life prediction method of Embodiment 1 of the present invention;

[0044] Figure 4 This is a schematic diagram of the junction-shell thermal resistance degradation trajectory generated in a rectifier bridge aging life prediction method according to Embodiment 1 of the present invention.

[0045] Figure 5 This is a flowchart of a rectifier bridge aging life early warning system according to Embodiment 2 of the present invention. Detailed Implementation

[0046] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below in conjunction with specific embodiments.

[0047] Example 1: Please refer to Figure 1-4 As shown in the figure, the aging life prediction method for a rectifier bridge according to an embodiment of the present invention includes the following steps:

[0048] S1: Collect the actual operating current waveform at the input of the rectifier bridge in the target application scenario, extract three characteristic parameters: peak current, duty cycle and conduction frequency, and generate a dynamic current loading sequence.

[0049] In S1, the target application scenario refers to the specific equipment or system that the rectifier bridge is ultimately installed and used in, such as electric vehicle drive motor controllers, photovoltaic inverters, industrial frequency converters, switching power supplies, charging piles, etc.

[0050] In S1, the process of collecting the actual operating current waveform at the input terminal of the rectifier bridge in the target application scenario is as follows:

[0051] Data acquisition is performed on the AC input side of the rectifier bridge in the target scenario. A current probe is connected in series in the AC input line of the rectifier bridge or a current clamp is used. The probe is connected to an oscilloscope or data acquisition card (sampling rate of 10,000 points per second). The device under test is allowed to operate normally in the target application scenario, covering the working cycle (such as a standard driving cycle of 1800 seconds for an electric vehicle or a complete working cycle for a frequency converter). The complete waveform data of the current flowing through the input of the rectifier bridge over time during the working cycle period is continuously recorded.

[0052] In S1, the process of extracting three characteristic parameters—peak current, duty cycle, and on-frequency—to generate a dynamic current loading sequence is as follows:

[0053] The acquired continuous current waveform is segmented according to the conduction phase, with each conduction phase being a pulse; the pulse is analyzed, and the maximum current value reached during the pulse is read as the peak current; the proportion of conduction time to total time in a complete working cycle is calculated as the duty cycle; the number of conduction pulses per unit time is calculated as the conduction frequency.

[0054] After extracting the above three parameters from the waveform data, input the values ​​of these three parameters into the programmable test power supply or electronic load device. The device outputs the corresponding current waveform according to the set peak current, duty cycle and conduction frequency to obtain the dynamic current loading sequence, which is used as the electrical stress input for subsequent aging tests.

[0055] Understandably, the purpose of the analysis in step S1 is to: by collecting the real current waveform in the actual application scenario of the rectifier bridge and extracting three characteristic parameters, namely peak current, duty cycle and conduction frequency, to generate a dynamic current loading sequence, so that the test loading condition is no longer a constant current but a composite waveform that can simulate two types of thermal stress under discontinuous conditions—transient thermal shock (determined by peak current) and cumulative temperature rise (determined by duty cycle). This solves the problem of the disconnect between traditional constant current loading and real discontinuous conditions, and creates input conditions for separating and tracking the two types of damage in subsequent tests, thereby improving the degree of fit between the test conditions and the actual situation and making the life prediction results more reliable.

[0056] S2: Apply a dynamic current loading sequence to the AC input terminal of the rectifier bridge under test. During the loading process, collect the forward voltage drop, junction-to-case thermal resistance, reverse leakage current and chip transient junction temperature to form the degradation trajectory of the forward voltage drop and junction-to-case thermal resistance as a function of the number of pulse cycles.

[0057] The specific implementation process of S2 is as follows:

[0058] Using the number of pulse cycles as the horizontal axis, for each pulse turn-on, record the measured forward voltage drop value and the corresponding number of pulse cycles, and mark the data points on the coordinate plane;

[0059] For each junction-to-shell thermal resistance measurement, the measured junction-to-shell thermal resistance value and the corresponding number of pulse cycles are recorded, and the data points are marked on the coordinate plane. As the loading continues, the number of data points increases continuously. Connecting the data points of the forward voltage drop in sequence forms a curve that gradually rises with the number of pulse cycles, which is the degradation trajectory of the forward voltage drop. Connecting the data points of the junction-to-shell thermal resistance in sequence forms a curve that gradually rises with the number of pulse cycles, which is the degradation trajectory of the junction-to-shell thermal resistance. The two degradation trajectories reflect the aging process of the bonding line and the solder layer, respectively.

[0060] It should be noted that the acquisition process for forward voltage drop, junction-to-case thermal resistance, reverse leakage current, and transient junction temperature of the chip can be as follows: When the rectifier bridge is in forward conduction, measure the voltage between its anode and cathode, and record the forward voltage drop during each pulse conduction period; during the pulse turn-off interval, apply a small current (such as 1mA to 25mA) that does not produce a significant self-heating effect to the rectifier bridge, and calculate the thermal resistance value by monitoring the junction temperature change and combining it with power loss; at the end of each pulse conduction period, use the thermistor method—by pre-calibrating the linear relationship between forward voltage drop and temperature—to convert the measured forward voltage drop into the chip junction temperature;

[0061] Understandably, the purpose of the analysis in step S2 is to simultaneously collect four parameters—forward voltage drop, junction-to-case thermal resistance, reverse leakage current, and transient junction temperature—after applying the dynamic current loading sequence to the rectifier bridge. The degradation trajectory of the forward voltage drop is used to track the thermal fatigue process of the bond wires, the degradation trajectory of the junction-to-case thermal resistance is used to track the creep fatigue process of the solder layer, the reverse leakage current serves as a monitoring indicator for the precursor of thermal runaway, and the transient junction temperature provides temperature input parameters for the subsequent lifetime model. This solves the problem of data aliasing between the two types of failure modes, realizes independent monitoring and data separation of the two types of failure modes, and provides a data foundation for the subsequent automatic identification of the first failure location.

[0062] S3: Based on the degradation trajectory, by comparing the number of pulse cycles required to reach failure by the forward voltage drop and the junction-shell thermal resistance, the failure site that occurs first under loading conditions is determined, including bond wire thermal fatigue and solder layer creep fatigue; when the reverse leakage current shows a sudden acceleration, it is marked as a risk of thermal runaway and the corresponding number of pulse cycles is recorded.

[0063] In S3, based on the degradation trajectory, the process of determining the initial failure location under loading conditions by comparing the number of pulse cycles required to reach failure by the forward voltage drop and the junction-shell thermal resistance is as follows:

[0064] Using the degradation trajectory of the positive voltage drop already formed in S2, the first positive voltage drop measured before the start of the test is taken as the initial value of the voltage drop. 105% of the initial value of the voltage drop is taken as the voltage drop failure threshold. In the degradation trajectory of the positive voltage drop, the number of pulse cycles corresponding to the first time the positive voltage drop reaches the voltage drop failure threshold is found and marked as the voltage drop failure cycle number.

[0065] Using the degradation trajectory of the junction-shell thermal resistance already formed in S2, the junction-shell thermal resistance measured for the first time before the start of the test is taken as the initial value of the thermal resistance. 120% of the initial value of the thermal resistance is taken as the thermal resistance failure threshold. In the degradation trajectory of the junction-shell thermal resistance, the number of pulse cycles corresponding to the first time the junction-shell thermal resistance reaches the thermal resistance failure threshold is found and marked as the thermal resistance failure cycle number.

[0066] It should be noted that the voltage drop failure threshold is set based on the following criteria: Reference power cycling test (IEC 60749-34 / IEC 60747-9): if the on-state voltage drop exceeds 5% of the initial value, the device is considered to have failed; or, according to the AQG324 automotive-grade standard, if the rate of change of the on-state voltage drop and saturation voltage drop Von exceeds 5%, the device is considered to have failed. The thermal resistance failure threshold is set based on the following criteria: reference power cycling test: if the thermal resistance exceeds 20% of the initial value, the device is considered to have failed; or, according to the AQG324 standard: if the rate of change of the junction-to-case thermal resistance exceeds 20%, the device is considered to have failed.

[0067] Compare the number of cycles for voltage drop failure with the number of cycles for thermal resistance failure;

[0068] If the number of cycles for failure due to pressure drop is less than the number of cycles for failure due to thermal resistance, then it is determined that thermal fatigue of the bonded wire occurs first under loading conditions.

[0069] If the number of failure cycles due to voltage drop is greater than the number of failure cycles due to thermal resistance, then it is determined that creep fatigue of the solder layer occurs first under loading conditions.

[0070] It should be noted that if the number of voltage drop failure cycles equals the number of thermal resistance failure cycles, and it is determined that the bonding wire thermal fatigue and solder layer creep fatigue occur simultaneously under the loading condition (which is an occasional situation), then the sample size is increased and the determination is made again: under the same loading condition, at least 10 rectifier bridges from the same batch are tested independently, and the number of voltage drop failure cycles and the number of thermal resistance failure cycles for each device are recorded. The bonding wire failure cycle and the solder layer failure cycle are compared and determined. After all the sample tests are completed, the number of devices with bonding wire failure and the number of devices with solder layer failure are counted. The failure location corresponding to the one with more failures is taken as the final failure location. If they are still the same, the sample size is increased (e.g., 30 devices) and the determination is made again until the failure location that occurs first can be distinguished.

[0071] In S3, when the reverse leakage current suddenly accelerates, the process of marking it as a risk of thermal runaway and recording the corresponding number of pulse cycles is as follows:

[0072] Based on the reverse leakage current data collected in S2, a curve of reverse leakage current changing with the number of pulse cycles is constructed, a measurement window is set (e.g., 100 cycles), and the total increment of reverse leakage current within the current measurement window is calculated.

[0073] The total increment of reverse leakage current is the difference between the reverse leakage current at the end of the measurement window and the reverse leakage current at the beginning of the measurement window.

[0074] It should be noted that the curve of reverse leakage current changing with the number of pulse cycles shows a slow and uniform upward trend during the normal aging stage of the device.

[0075] If the total increase in reverse leakage current in the current measurement window exceeds 5 times the total increase in reverse leakage current in the previous measurement window, it is determined to be a sudden acceleration climb.

[0076] When a sudden acceleration is detected, it is marked as a thermal runaway risk node and the corresponding number of pulse cycles is recorded.

[0077] Understandably, the purpose of the analysis in step S3 is to automatically identify which failure mode occurs first in the bonding wire thermal fatigue and solder layer creep fatigue under the loading condition by comparing the number of pulse cycles required for the forward voltage drop and junction-shell thermal resistance to reach their respective failure thresholds, and to determine the end of the device's lifespan. This solves the problem that traditional methods preset a single mode, leading to inaccurate predictions, because the dominant failure modes differ under different operating conditions. At the same time, by monitoring the sudden acceleration of the reverse leakage current, thermal runaway risk nodes are marked in a timely manner and a basis is provided for subsequent data truncation, which improves the accuracy of failure location judgment and the relevance of model selection, and effectively avoids lifespan prediction deviations caused by misjudgment of failure modes.

[0078] S4: Based on the first failure location, extract complete degradation data from the initial stage to the failure and the number of pulse cycles from the corresponding degradation trajectory, supplemented by transient junction temperature data and thermal runaway risk markers, to form a standardized training dataset;

[0079] The specific implementation process of S4 is as follows:

[0080] Based on the judgment result of S3, if the bonding wire thermal fatigue is determined to occur first, the degradation trajectory of the positive voltage drop is selected as the data extraction object; if the solder layer creep fatigue is determined to occur first, the degradation trajectory of the junction-shell thermal resistance is selected as the data extraction object.

[0081] If thermal runaway risk nodes are marked in S3 and the corresponding pulse cycle counts are recorded, then the pulse cycle counts corresponding to the thermal runaway risk nodes are used as the cutoff position, and only the degradation data before the thermal runaway risk nodes are extracted from the degradation trajectory and included in the dataset.

[0082] It should be noted that data after the thermal runaway risk point is not included in the training dataset. This is because after thermal runaway is triggered, the device has entered the abnormal failure stage. The reverse leakage current increases sharply, forming positive feedback. The forward voltage drop or junction-to-case thermal resistance changes drastically in a very short time. Its degradation rate and change pattern are completely different from the normal aging stage. If data after the thermal runaway risk point is mixed into the training set, it will skew the model's fitting results, making the model describe abnormal degradation behavior rather than the actual aging process. This will lead to an underestimation of the remaining lifetime. Moreover, once thermal runaway occurs, the device usually fails completely in a very short time. There is no time window for intervention in engineering. Therefore, only the normal aging data before thermal runaway occurs is used as the basis for modeling.

[0083] Data points are extracted from the degradation trajectory, specifically including: horizontal axis: the number of pulse cycles corresponding to each data point, from the first pulse cycle to the failure threshold; vertical axis: the characteristic parameter value (forward voltage drop or junction-shell thermal resistance) corresponding to each data point, from the initial value to the failure threshold.

[0084] Each extracted data set is arranged in ascending order of the number of iterations to form a complete degradation sequence.

[0085] Based on the obtained degradation sequence, combined with transient junction temperature data and thermal runaway risk markers, a standardized training dataset is formed by arranging the data items according to the data format, specifically: First column: number of pulse cycles (number of cycles from the 1st to the failure threshold); Second column: characteristic parameter values ​​(forward voltage drop or junction-to-case thermal resistance, corresponding to the failure location); Third column: transient junction temperature data (chip junction temperature corresponding to each measurement); Fourth column: thermal runaway risk marker (marked as 0 if the current cycle number is before the risk node; marked as 1 if it is after the risk node; marked as 0 if there is no risk event). For example, the standardized training dataset can be seen in Table 1 below.

[0086] Table 1: Standardized training dataset (taking forward pressure drop as an example);

[0087] Understandably, the purpose of the analysis in step S4 is to extract complete degradation data from the initial stage to the failure threshold and the number of pulse cycles from the corresponding degradation trajectory based on the initial failure location determined in S3, and supplement it with transient junction temperature data and thermal runaway risk markers to form a standardized training dataset. This solves the problems of mismatch between modeling data and failure location and abnormal data contaminating the training set after thermal runaway, ensuring that the input data in the S5 modeling stage corresponds to the failure location and does not contain abnormal data after thermal runaway, thereby improving the accuracy of subsequent model parameter calibration and the effectiveness of model training.

[0088] S5: Select a matching lifetime assessment model based on the feature parameter type of the training dataset, input the feature parameters under the target working condition into the model, and output the number of pulse cycles required to reach the failure threshold under the target working condition.

[0089] The specific implementation process in S5 is as follows:

[0090] In the standardized training dataset output by S4, based on the feature parameter values, if the feature parameter values ​​indicate that the corresponding failure site is bond line thermal fatigue, then the Coffin-Manson model is selected. The Coffin-Manson model is based on the physical mechanism of plastic strain accumulation in thermal cycling, including the temperature difference ΔT for each thermal cycle, and the number of pulse cycles N_f has a power function relationship with ΔT.

[0091] If the failure site is solder layer creep fatigue, the Darveaux model is selected. The Darveaux model is based on the physical mechanism of cumulative creep deformation of solder material at high temperature, and includes parameters such as average temperature and cycle time.

[0092] In the degradation trajectory, the interval from initial to failure is divided into multiple stages (e.g., 3-4 stages; taking the forward voltage drop as an example, if the failure threshold is 105% of the initial value, then the range between 1.000 times the initial value and 1.050 times the initial value is evenly divided into 4 intervals: 1.000~1.0125, 1.0125~1.025, 1.025~1.0375, 1.0375~1.050, each interval being a stage). For each stage, the average junction temperature difference, the increment of the number of pulse cycles experienced by the stage, and the average junction temperature are measured.

[0093] The pulse cycle increment is the absolute deviation between the number of pulse cycles at the start of the stage and the number of pulse cycles at the end of the stage. The average junction temperature difference is obtained by extracting the transient junction temperature data corresponding to all measurement moments between the number of pulse cycles at the start of the stage and the number of pulse cycles at the end of the stage. For each pulse cycle, the cycle temperature difference is the highest junction temperature reached by the chip when it is turned on minus the valley temperature reached after it is turned off. The cycle temperature difference values ​​of all measurement moments in the stage are added together and divided by the total number of measurements in the stage to obtain the average junction temperature difference of the stage. The average junction temperature is the average value of the transient junction temperature corresponding to all measurement moments in the stage.

[0094] If the Coffin-Manson model is selected, the average junction temperature difference and the increment of pulse cycle number in each stage are taken as data sets. The multiple data sets are plotted on a double logarithmic coordinate plane and linearly fitted to complete the parameter calibration of the Coffin-Manson model.

[0095] If the Darveaux model is selected, the average junction temperature and pulse cycle number increment of each stage are taken as the data set. The multiple data sets are then substituted into the Darveaux model for fitting to complete the Darveaux model parameter calibration.

[0096] Take a random data set from a certain period (not used for model parameter fitting calibration) as the validation set, substitute it into the calibrated model, and calculate the corresponding theoretical value of the failure cycle count.

[0097] Compare the theoretical value of the number of failure cycles with the pulse cycle deviation of the pulse cycle count increment in the data set:

[0098] If the pulse cycle deviation is within the preset deviation range (e.g., ±15%), the model is deemed to have passed.

[0099] If the pulse cycle deviation is not within the preset deviation range, it indicates that the stage division method is unreasonable or the fitting result deviates from the actual situation. The stages should be re-divided, the number of stages increased or the stage boundaries adjusted, and the data extracted again for calibration and verification until the deviation is within an acceptable range.

[0100] If the failure site is bond wire fatigue, the highest junction temperature reached by the chip during each turn-on and the lowest junction temperature after turn-off are obtained from the target operating conditions. The deviation is calculated to obtain the average junction temperature difference. The average junction temperature difference is substituted into the Coffin-Manson model to obtain the number of pulse cycles required to reach the failure threshold under the input conditions.

[0101] If the failure site is solder layer fatigue, then obtain the long-term average junction temperature of the chip under the target operating condition, substitute it into the Darveaux model, and obtain the number of pulse cycles required to reach the failure threshold under the input conditions.

[0102] The number of pulse cycles required to reach the failure threshold is the estimated remaining service life of the rectifier bridge under the target operating condition, where the target operating condition is the working condition of the equipment in which the rectifier bridge is actually installed and used.

[0103] If the remaining number of cycles calculated under the target operating condition is lower than a fixed percentage of the expected value of the normal service life under the operating condition (e.g., lower than 30% of the expected value), an early warning will be triggered, indicating that the rectifier bridge has an early failure risk in this application scenario, and it is recommended to shorten the maintenance interval or replace it in advance.

[0104] The warning levels are divided into three levels: when the remaining number of cycles is higher than 60% of the expected value, it is considered a safe state and no warning is triggered; when the remaining number of cycles is between 30% and 60% of the expected value, a Level 1 warning is triggered, indicating a moderate risk and suggesting that the monitoring frequency be increased; when the remaining number of cycles is lower than 30% of the expected value, a Level 2 warning is triggered, indicating a high risk and suggesting that the system be replaced immediately.

[0105] Understandably, the purpose of the analysis in step S5 is to select a lifetime assessment model that matches the physical mechanism of the failure site based on the characteristic parameter types of the dataset in S4—the Coffin-Manson model based on the thermal cycling plastic strain accumulation mechanism is selected for bond wire fatigue, and the Darveaux model based on the high-temperature creep deformation accumulation mechanism is selected for solder layer fatigue. The model parameters are calibrated and verified using multi-stage data in the dataset. Then, the temperature parameters under the target operating condition are input into the calibrated model, and finally, the number of pulse cycles required for the rectifier bridge to reach the failure threshold under the target operating condition is output. This solves the problem that a single model cannot accurately describe the failure process of two different physical mechanisms, improves the lifetime prediction accuracy and model generalization ability, and ensures the consistency between the prediction results and the actual failure mechanism of the device.

[0106] Example 2, please refer to Figure 5 As shown in the embodiment of the present invention, a rectifier bridge aging life early warning system includes the following modules:

[0107] Dynamic loading module: Collect the actual operating current waveform at the input of the rectifier bridge in the target application scenario, extract three characteristic parameters: peak current, duty cycle and conduction frequency, and generate a dynamic current loading sequence;

[0108] Degradation trajectory acquisition module: Apply a dynamic current loading sequence to the AC input terminal of the rectifier bridge under test. During the loading process, the forward voltage drop, junction-to-case thermal resistance, reverse leakage current and chip transient junction temperature are acquired to form degradation trajectories of the forward voltage drop and junction-to-case thermal resistance as a function of the number of pulse cycles.

[0109] Failure Location Determination Module: Based on the degradation trajectory, the module determines the failure location that occurs first under loading conditions by comparing the number of pulse cycles required to reach failure by comparing the forward voltage drop and the junction-shell thermal resistance. This includes bond wire thermal fatigue and solder layer creep fatigue. When the reverse leakage current suddenly accelerates, it is marked as a risk of thermal runaway and the corresponding number of pulse cycles is recorded.

[0110] Training data generation module: Based on the first failure location, extract complete degradation data from the initial stage to failure and the number of pulse cycles from the corresponding degradation trajectory, supplemented by transient junction temperature data and thermal runaway risk markers, to form a standardized training dataset;

[0111] The remaining life estimation module selects a matching life assessment model based on the feature parameter types of the training dataset, inputs the feature parameters under the target operating condition into the model, and outputs the number of pulse cycles required to reach the failure threshold under the target operating condition.

[0112] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for predicting the aging life of a rectifier bridge, characterized in that: Includes the following steps: S1: Collect the actual operating current waveform at the input of the rectifier bridge in the target application scenario, extract three characteristic parameters: peak current, duty cycle and conduction frequency, and generate a dynamic current loading sequence. S2: Apply a dynamic current loading sequence to the AC input terminal of the rectifier bridge under test. During the loading process, collect the forward voltage drop, junction-to-case thermal resistance, reverse leakage current and chip transient junction temperature to form the degradation trajectory of the forward voltage drop and junction-to-case thermal resistance as a function of the number of pulse cycles. S3: Based on the degradation trajectory, by comparing the number of pulse cycles required to reach failure by the forward voltage drop and the junction-shell thermal resistance, the failure site that occurs first under loading conditions is determined, including bond wire thermal fatigue and solder layer creep fatigue; when the reverse leakage current shows a sudden acceleration, it is marked as a risk of thermal runaway and the corresponding number of pulse cycles is recorded. S4: Based on the first failure location, extract complete degradation data from the initial stage to the failure and the number of pulse cycles from the corresponding degradation trajectory, supplemented by transient junction temperature data and thermal runaway risk markers, to form a standardized training dataset; S5: Select a matching lifetime assessment model based on the feature parameter type of the training dataset, input the feature parameters under the target working condition into the model, and output the number of pulse cycles required to reach the failure threshold under the target working condition. In the standardized training dataset output by S4, based on the feature parameter values, if the feature parameter values ​​indicate that the corresponding failure site is thermal fatigue of the bond line, then the Coffin-Manson model is selected. If the failure site is solder layer creep fatigue, the Darveaux model should be selected. If the failure site is bond wire fatigue, the highest junction temperature reached by the chip during each turn-on and the lowest junction temperature after turn-off are obtained from the target operating conditions. The deviation is calculated to obtain the average junction temperature difference. The average junction temperature difference is substituted into the Coffin-Manson model to obtain the number of pulse cycles required to reach the failure threshold under the input conditions. If the failure site is solder layer fatigue, then obtain the long-term average junction temperature of the chip under the target operating condition, substitute it into the Darveaux model, and obtain the number of pulse cycles required to reach the failure threshold under the input conditions.

2. The aging life prediction method for a rectifier bridge according to claim 1, characterized in that: In S1, the target application scenario is the equipment or system that the rectifier bridge is ultimately installed and used in. The process of collecting the actual operating current waveform at the input terminal of the rectifier bridge in the target application scenario is as follows: the data is collected on the AC input side of the rectifier bridge, so that the device under test is running normally in the target application scenario, covering a complete working cycle, and the complete waveform data of the current flowing through the input terminal of the rectifier bridge changing with time is continuously recorded during the working cycle period.

3. The aging life prediction method for a rectifier bridge according to claim 2, characterized in that: In S1, the process of extracting the three characteristic parameters—peak current, duty cycle, and conduction frequency—and generating the dynamic current loading sequence is as follows: The acquired continuous current waveform is segmented according to the conduction stage, with each conduction stage being a pulse. The maximum current value reached during the pulse is read as the peak current. The proportion of conduction time to total time within a complete working cycle is calculated as the duty cycle. The number of conduction pulses per unit time is calculated as the conduction frequency. The extracted three parameter values ​​are input into the programmable test power supply, which outputs the corresponding current waveform according to the set peak current, duty cycle, and conduction frequency, thus obtaining the dynamic current loading sequence.

4. The aging life prediction method for a rectifier bridge according to claim 1, characterized in that: In S2, the process of forming the degradation trajectory of the positive voltage drop and junction-shell thermal resistance as a function of the number of pulse cycles is as follows: Using the number of pulse cycles as the horizontal axis, record the measured forward voltage drop value and the corresponding number of pulse cycles for each pulse conduction, and mark the data points on the coordinate plane. For each junction-shell thermal resistance measurement, record the measured junction-shell thermal resistance value and the corresponding number of pulse cycles, and mark the data points on the coordinate plane; Connecting the data points of the forward voltage drop sequentially forms the degradation trajectory of the forward voltage drop, and connecting the data points of the junction-shell thermal resistance sequentially forms the degradation trajectory of the junction-shell thermal resistance.

5. The aging life prediction method for a rectifier bridge according to claim 4, characterized in that: In S2, the forward voltage drop is obtained when the rectifier bridge is in the forward conduction state, the junction-to-case thermal resistance is obtained during the pulse turn-off interval, and the chip transient junction temperature is obtained by the thermistor method at the end of each pulse conduction period.

6. The aging life prediction method for a rectifier bridge according to claim 1, characterized in that: In S3, the process of determining the failure location that occurs first under loading conditions is as follows: Using the rate of change of the initial value of the forward voltage drop reaching a preset threshold as the voltage drop failure criterion, the pulse cycle number corresponding to the first time the forward voltage drop reaches the voltage drop failure criterion is found in the degradation trajectory of the forward voltage drop and marked as the voltage drop failure cycle number. The rate of change of the initial value of junction-shell thermal resistance reaching a preset threshold is used as the criterion for thermal resistance failure. The number of pulse cycles corresponding to the first time the junction-shell thermal resistance reaches the thermal resistance failure criterion is found in the degradation trajectory of junction-shell thermal resistance and marked as the number of thermal resistance failure cycles. By comparing the number of failure cycles of pressure drop with the number of failure cycles of thermal resistance, the failure site corresponding to the smaller one is the failure site that occurs first under loading conditions.

7. The aging life prediction method for a rectifier bridge according to claim 6, characterized in that: In S3, the process of marking the risk of thermal runaway is as follows: Based on the reverse leakage current data, a curve of the reverse leakage current changing with the number of pulse cycles is constructed, a measurement window is set, and the total increment of the reverse leakage current within the current measurement window is calculated. If the total increase in reverse leakage current within the current measurement window exceeds a preset multiple of the total increase in reverse leakage current within the previous measurement window, it is determined to be a sudden acceleration, marked as a thermal runaway risk node, and the corresponding number of pulse cycles is recorded.

8. The aging life prediction method for a rectifier bridge according to claim 7, characterized in that: In S4, the process of forming a standardized training dataset is as follows: Based on the judgment result of S3, the corresponding degradation trajectory is selected as the data extraction object; if a thermal runaway risk node is marked, the number of pulse cycles corresponding to the thermal runaway risk node is used as the cutoff position, and the degradation data before the cutoff position is extracted and included in the dataset; the number of pulse cycles and feature parameter values ​​corresponding to each data point are extracted and arranged in ascending order of the number of cycles to form a complete degradation sequence; the degradation sequence is arranged in a four-column format of pulse cycle number, feature parameter value, transient junction temperature, and thermal runaway risk mark to form a standardized training dataset.

9. The aging life prediction method for a rectifier bridge according to claim 8, characterized in that: In S5, the process of outputting the number of pulse cycles required to reach the failure threshold under the target operating condition is as follows: The failure location was determined based on the feature parameter type of the training dataset. The Coffin-Manson model was used for thermal fatigue of the bond wire, and the Darveaux model was used for creep fatigue of the solder layer. The degradation trajectory from initial to failure is divided into multiple stages. For each stage, the corresponding temperature parameters and pulse cycle increments are measured. The data sets from multiple stages are substituted into the selected model for parameter calibration. A data set from a stage that was not calibrated is used as a validation set for deviation verification. After successful verification, the temperature parameters under the target operating condition are input into the calibrated model, and the number of pulse cycles required to reach the failure threshold under the target operating condition is output. A graded warning is triggered based on the deviation between the calculated result and the expected value.

10. A rectifier bridge aging life early warning system, characterized in that: The system is used to perform the method according to any one of claims 1-9, and includes the following modules: Dynamic loading module: Collect the actual operating current waveform at the input of the rectifier bridge in the target application scenario, extract three characteristic parameters: peak current, duty cycle and conduction frequency, and generate a dynamic current loading sequence; Degradation trajectory acquisition module: Apply a dynamic current loading sequence to the AC input terminal of the rectifier bridge under test. During the loading process, the forward voltage drop, junction-to-case thermal resistance, reverse leakage current and chip transient junction temperature are acquired to form degradation trajectories of the forward voltage drop and junction-to-case thermal resistance as a function of the number of pulse cycles. Failure Location Determination Module: Based on the degradation trajectory, the module determines the failure location that occurs first under loading conditions by comparing the number of pulse cycles required to reach failure by comparing the forward voltage drop and the junction-shell thermal resistance. This includes bond wire thermal fatigue and solder layer creep fatigue. When the reverse leakage current suddenly accelerates, it is marked as a risk of thermal runaway and the corresponding number of pulse cycles is recorded. Training data generation module: Based on the first failure location, extract complete degradation data from the initial stage to failure and the number of pulse cycles from the corresponding degradation trajectory, supplemented by transient junction temperature data and thermal runaway risk markers, to form a standardized training dataset; The remaining life estimation module selects a matching life assessment model based on the feature parameter types of the training dataset, inputs the feature parameters under the target operating condition into the model, and outputs the number of pulse cycles required to reach the failure threshold under the target operating condition.

Images