Automatic testing and screening method for electronic components

Abstract

The invention discloses an automatic testing and screening method for electronic components, which relates to the technical field of electronic engineering and microelectronics and comprises the following steps: generating a multi-dimensional stress field through a temperature control module, a humidity control module and an electromagnetic interference simulation module; calling a preset extreme working condition combination sequence based on the target application scene of the component; under a dynamic coupling environment, dominant parameters and recessive parameters including a dielectric relaxation spectrum, a carrier mobility transient response and a thermoacoustic emission signal are synchronously collected; inputting the hidden parameter time sequence data into physical failure models including an electrochemical corrosion model and a thermal mechanical fatigue model; dynamically adjusting the passing threshold of the test item according to the model output; the step of applying accelerated degradation excitation to the screened component comprises the following steps of: circulating a set temperature for a set number of times; applying a bias voltage-temperature stress; and repeating the step S2 to monitor the implicit parameter drift amount, and eliminating the device with out-of-tolerance drift. The device guarantee is provided for the fields of new energy automobiles, aerospace electronics and the like.

Description

Technical Field

[0001] This invention relates to the fields of electronic engineering and microelectronics, specifically to an automated testing and screening method for electronic components. Background Technology

[0002] Electronic components are the "cornerstone" of electronic devices, and their quality directly determines the reliability and performance of the final product. With the development of semiconductor technology, components are evolving towards "miniaturization, integration, and multi-functionality," with the number of parameters increasing from a dozen to hundreds, making it difficult for manual testing to cover all characteristics.

[0003] With the surge in demand for reliability in various application scenarios, fields such as automotive electronics, aerospace (where components need to be radiation-resistant and vibration-resistant), and medical equipment (where sensor accuracy needs to be within 0.1%) have an urgent need for "zero-defect" screening.

[0004] Mass production scales in industries such as consumer electronics and new energy can reach millions per batch. Traditional manual testing (taking several minutes per chip) is extremely inefficient, and manual operation is prone to errors (such as reading deviations and poor contact), resulting in poor screening consistency. Automated testing can reduce labor costs, replace more than 50% of manual labor, and reduce the flow of defective products to downstream industries through rapid screening. For example, if a chip fault is found after soldering a mobile phone motherboard, the rework cost increases by more than 10 times.

[0005] Industry 4.0 and "smart manufacturing" require the testing process to be integrated into the production line closed loop: data is collected in real time through automated testing systems and fed back to the production end to optimize processes (such as adjusting wafer cutting parameters through chip test data), so as to achieve digital linkage of "testing-production-improvement".

[0006] However, there is still room for improvement in existing technologies, mainly in terms of insufficient test coverage and scenario simulation. For example, the performance of complex components is affected by multiple dimensions such as temperature, humidity, and electromagnetic interference. Existing automated systems mostly focus on testing "normal temperature static parameters," making it difficult to simulate extreme conditions such as RF power stability at -55°C, leading to missed detection of "potential defects." Test parameters emphasize explicit parameters while neglecting implicit ones, prioritizing the testing of easily quantifiable explicit parameters but ignoring implicit characteristics, resulting in selected products being "short-term qualified but long-term unsuccessful." Summary of the Invention

[0007] To address the aforementioned technical problems, an automated testing and screening method for electronic components is provided. This technical solution solves the problems of insufficient implicit feature extraction capability and difficulty in multi-source data fusion.

[0008] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0009] An automated testing and screening method for electronic components includes:

[0010] S1. Generate a multi-dimensional stress field through the temperature control module, humidity control module, and electromagnetic interference simulation module; based on the target application scenario of the components, call the preset extreme working condition combination sequence;

[0011] S2. In a dynamic coupling environment, dominant and latent parameters, including dielectric relaxation spectrum, transient response of carrier mobility, and thermoacoustic emission signal, are collected synchronously.

[0012] S3. Input the implicit parameter time series data into the physical failure model, including the electrochemical corrosion model and the thermomechanical fatigue model; dynamically adjust the pass threshold of the test items based on the model output;

[0013] S4. Apply accelerated degradation excitation to the selected components, including: cycling at a predetermined temperature for a predetermined number of times; applying bias-temperature stress; repeating step S2 to monitor the amount of latent parameter drift, and eliminating components with excessive drift.

[0014] Preferably, S1 specifically includes:

[0015] Temperature control module: a three-stage temperature control architecture using liquid nitrogen injection cooling, semiconductor Peltier stacking, and instantaneous heating of silicon carbide film;

[0016] Transient cooling of the device surface is achieved through a high-pressure liquid nitrogen micro-spray system; rapid heating is achieved in a short time using a SiC ceramic heating film, combined with closed-loop control using an infrared temperature sensor; a microchannel turbulence system is built into the test chamber, and the airflow distribution is optimized through computational fluid dynamics to ensure that the temperature gradient on the device surface is less than or equal to a predetermined temperature.

[0017] Preferably, S1 specifically includes:

[0018] Humidity control module: based on dynamic injection of saturated water vapor using a dual-pressure method and precise control of condensation dew point;

[0019] Wide humidity coverage: A predetermined range of relative humidity is generated through a graded pressurized saturator, and the predetermined relative humidity can still be maintained at a predetermined high temperature; a piezoelectric micro-pump array is used to precisely control the water vapor injection volume to achieve rapid humidity switching; anti-condensation design: a semiconductor condenser plate is integrated at the bottom of the device carrier stage to monitor the surface temperature and dew point difference in real time.

[0020] Preferably, S1 specifically includes:

[0021] Electromagnetic Interference Simulation Module: Constructs a programmable composite spectrum interference field, including wideband coverage, and synthesizes arbitrary waveform interference through a vector signal generator array, supporting 5G millimeter wave and high-frequency scenarios for automotive radar; spatial field strength control, using a waveguide slot antenna matrix, and optimizing the layout with electromagnetic simulation software; synchronous coupling mechanism, sharing timestamps with the temperature and humidity module, superimposing electromagnetic pulse groups at the inflection point of temperature cycles to simulate the multi-stress synergistic effect of real environment.

[0022] Preferably, S2 specifically includes:

[0023] High-precision acquisition of explicit parameters:

[0024] Voltage / current parameters are measured accurately in the nanoampere level and microvolt level across the entire temperature range using a quantum tunneling current sensor and a Josephson junction voltage reference.

[0025] The rise / fall time and eye opening of high-speed signals are captured by a sampling oscilloscope, and the EMI interference immunity is tested simultaneously by superimposing the data; time-interleaved sampling technology is applied to compress the single measurement time.

[0026] Preferably, S2 specifically includes:

[0027] Deep perception of latent parameters:

[0028] Dielectric relaxation spectrum material degradation monitoring: A sinusoidal sweep voltage is applied, and the real and imaginary parts of the dielectric are extracted through a lock-in amplifier and a three-electrode system to calculate the loss factor.

[0029] Transient response of carrier mobility was obtained by applying a nanosecond-level high-voltage pulse and using a 4D carrier imaging system to plot the mobility distribution.

[0030] The degradation model correlation shows that if the temperature coefficient of mobility is greater than a predetermined threshold, it conforms to the Arrhenius model and is judged as normal thermal degradation; if the temperature coefficient of mobility is less than the predetermined threshold, an electromigration risk warning is triggered.

[0031] Thermoacoustic emission signal crack detection employs a multi-modal sensor array, including a piezoelectric sensor to capture the acoustic emission energy of microcrack propagation; and an infrared thermal imager to simultaneously record the transient temperature rise in the crack region.

[0032] The feature extraction algorithm identifies crack initiation if the wavelet packet energy entropy of the acoustic emission signal is greater than a predetermined threshold; and locates the poor weld area of ​​the solder joint if the thermal diffusion rate changes abruptly.

[0033] Preferably, S3 specifically includes:

[0034] The electrochemical corrosion model takes as input parameters real-time relative humidity from the humidity module, loss factor in the low-frequency band of dielectric relaxation spectrum, and ion migration current monitored by the quantum tunneling current sensor. A corrosion rate equation is established, and the threshold is dynamically adjusted. When the corrosion rate exceeds the predetermined threshold, it is judged as a high risk of electrochemical migration, and the upper limit of the allowable leakage current is automatically tightened from the standard value and the coating thickness is re-inspected.

[0035] The thermomechanical fatigue model takes as input parameters the number of temperature cycles, heating and cooling rates, cumulative thermoacoustic emission energy, and measured values ​​of the thermal expansion coefficient of the packaging material. Lifetime prediction is based on the Coffin-Manson correction formula with dynamic threshold adjustment. If the predicted lifetime is a predetermined number of cycles, three additional extreme temperature shock tests are forcibly added, and the thermoacoustic emission energy threshold is adjusted to the predetermined threshold.

[0036] Preferably, S3 specifically includes:

[0037] The dynamic threshold adjustment strategy uses a multi-parameter weighted decision tree to construct an implicit parameter weight matrix and dynamically adjust the weight ratio according to the application scenario. A comprehensive score is calculated, and failure is determined when the comprehensive score is greater than a predetermined threshold, and feedback is given to the test item threshold in real time.

[0038] The machine learning online correction uses a Bayesian optimization algorithm to update the model parameters based on historical data after a predetermined number of tests. Abnormal cases feed back into the model by extracting features from misjudged / missed samples to generate adversarial training data.

[0039] Preferably, S4 specifically includes:

[0040] Extreme temperature cycling is achieved by alternating liquid nitrogen injection and silicon carbide infrared heating to reach the limit of a predetermined temperature; based on the number of intelligent matching cycles according to the target device lifespan, it is divided into consumer electronics, automotive electronics, and aerospace devices; failure excitation utilizes the CTE mismatch effect to forcibly expose solder joint microcracks and through-silicon via fractures.

[0041] Bias-temperature stress is applied using an electrothermal coupling loading mode. The static bias is applied as a predetermined multiple of the rated voltage, while the ambient temperature is simultaneously raised to a predetermined temperature range. Dynamic disturbance is applied by superimposing square wave switching stress to simulate the actual switching losses of power devices. The sudden increase in gate oxide tunneling current is captured in real time by an in-situ quantum tunneling sensor, and the hot spot area is located by combining infrared thermal imaging.

[0042] Preferably, S4 specifically includes:

[0043] The model quantifies the amount of hidden drift and the dielectric relaxation spectrum drift. It calculates the rate of change of the loss factor before and after accelerated aging in the low-frequency band. When the rate of change before and after accelerated aging is greater than a predetermined proportion, the polymer encapsulation is determined to have failed hydrolysis. When the shift of the characteristic relaxation peak is greater than a predetermined threshold, the concentration of ionic impurities is indicated to be excessive.

[0044] Carrier mobility decay, extract mobility degradation rate under pulse test, thermoacoustic emission energy accumulation, and statistically analyze the total energy of acoustic emission signal during acceleration.

[0045] Dynamic elimination criteria, multi-parameter fusion decision tree, parameter types include dielectric loss variation, mobility decay, thermoacoustic energy, and quantum tunneling current.

[0046] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0047] This invention proposes a method to dynamically invoke a pre-set sequence of extreme operating conditions based on the target application scenario. The temperature control module achieves extreme temperature jumps from -196℃ to 300℃ through liquid nitrogen injection and silicon carbide infrared heating; the humidity module maintains 95% RH saturated water vapor at 85℃; and the electromagnetic interference module injects full-band noise with a field strength of 200V / m to solve the distortion problem of traditional single-stress accelerated testing.

[0048] By simultaneously capturing three latent parameters, dielectric relaxation spectroscopy is used to detect epoxy resin molecular chain breakage, transient response of carrier mobility is used to locate lattice defects, and thermoacoustic emission signals are used to capture microcrack propagation. This enables molecular-level early diagnosis of latent failures such as encapsulation hydrolysis, ion migration, and fatigue cracks, reducing the false negative rate by more than 90% compared to traditional methods.

[0049] An adaptive criterion system is constructed based on a physical failure model. The electrochemical corrosion model is linked to humidity, ion current, and dielectric loss parameters to dynamically tighten the leakage current threshold. The thermomechanical fatigue model is based on the temperature change rate and acoustic emission cumulative energy to correct the lifetime equation in real time. Through Bayesian optimization algorithm, the model parameters are automatically updated after every thousand tests, so that the misjudgment rate continues to converge to <0.5%, which is significantly better than the coarse screening with a fixed threshold. Attached Figure Description

[0050] Figure 1 This is a flowchart of an automated testing and screening method for electronic components. Detailed Implementation

[0051] The following description is intended to disclose the invention and enable those skilled in the art to implement it. The preferred embodiments described below are merely examples, and other obvious variations will occur to those skilled in the art.

[0052] Reference Figure 1 As shown, an automated testing and screening method for electronic components includes:

[0053] S1. Generate a multi-dimensional stress field through the temperature control module, humidity control module, and electromagnetic interference simulation module; based on the target application scenario of the components, call the preset extreme working condition combination sequence;

[0054] S2. In a dynamic coupling environment, dominant and latent parameters, including dielectric relaxation spectrum, transient response of carrier mobility, and thermoacoustic emission signal, are collected synchronously.

[0055] S3. Input the implicit parameter time series data into the physical failure model, including the electrochemical corrosion model and the thermomechanical fatigue model; dynamically adjust the pass threshold of the test items based on the model output;

[0056] S4. Apply accelerated degradation excitation to the selected components, including: cycling at a predetermined temperature for a predetermined number of times; applying bias-temperature stress; repeating step S2 to monitor the amount of latent parameter drift, and eliminating components with excessive drift.

[0057] It should be noted that the environmental simulation layer (S1) precisely triggers the failure mechanism:

[0058] By using spatiotemporal synchronous coupling of temperature, humidity, and EMI to approximate real working conditions, the stress decoupling distortion caused by traditional step loading is solved.

[0059] Data Awareness Layer (S2) → Deep Capture of Latent Defects:

[0060] Dielectric relaxation spectrum (10) - 3H z Low frequency) → Molecular polarization level monitoring (1000 times more sensitive than IR testing);

[0061] Carrier transient response (100ns pulse) → Lattice defect imaging (locating dislocation clusters in SiCMOSFETs);

[0062] Thermoacoustic signal (40MHz sensor) → Microcrack energy quantization (detection of 10μm-level tin whiskers);

[0063] Intelligent Decision Layer (S3) → Dynamic Threshold Reconstruction:

[0064] Physical model-driven: Electrochemical corrosion model, correlated with humidity, ion current, and dielectric loss (Rcorr > 0.75 trigger threshold tightening); Thermomechanical fatigue model, integrating CTE, temperature change rate, and acoustic emission energy (predicted lifetime Nf < 10). 4 (an additional mandatory test);

[0065] Lifetime verification layer (S4) → Residual reliability screening:

[0066] Accelerated stress: Temperature cycling is -65℃~175℃ / 50℃ / min (exceeding JEDEC standard);

[0067] Bias stress: 12V@200℃ + dynamic disturbance (excites gate oxide trap);

[0068] S1 Extreme Environment Simulation → Multi-stress Coupling, S2 Implicit Parameter Awareness → Molecular-Level Data, S3 Dynamic Threshold Decision → Model Output, S4 Accelerated Lifetime Verification → Drift Data Feedback → Device Elimination, High-Reliability Delivery.

[0069] S1 specifically includes:

[0070] Temperature control module: a three-stage temperature control architecture using liquid nitrogen injection cooling, semiconductor Peltier stacking, and instantaneous heating of silicon carbide film;

[0071] Transient cooling of the device surface is achieved through a high-pressure liquid nitrogen micro-spray system; rapid heating is achieved in a short time using a SiC ceramic heating film, combined with closed-loop control using an infrared temperature sensor; a microchannel turbulence system is built into the test chamber, and the airflow distribution is optimized through computational fluid dynamics to ensure that the temperature gradient on the device surface is less than or equal to a predetermined temperature.

[0072] Humidity control module: based on dynamic injection of saturated water vapor using a dual-pressure method and precise control of condensation dew point;

[0073] Wide humidity coverage: A predetermined range of relative humidity is generated through a graded pressurized saturator, and the predetermined relative humidity can still be maintained at a predetermined high temperature; a piezoelectric micro-pump array is used to precisely control the water vapor injection volume to achieve rapid humidity switching; anti-condensation design: a semiconductor condenser plate is integrated at the bottom of the device carrier stage to monitor the surface temperature and dew point difference in real time.

[0074] Electromagnetic Interference Simulation Module: Constructs a programmable composite spectrum interference field, including wideband coverage, and synthesizes arbitrary waveform interference through a vector signal generator array, supporting 5G millimeter wave and high-frequency scenarios for automotive radar; spatial field strength control, using a waveguide slot antenna matrix, and optimizing the layout with electromagnetic simulation software; synchronous coupling mechanism, sharing timestamps with the temperature and humidity module, superimposing electromagnetic pulse groups at the inflection point of temperature cycles to simulate the multi-stress synergistic effect of real environment.

[0075] It should be noted that the temperature control module:

[0076] Liquid nitrogen injection refrigeration utilizes the Joule-Thomson effect (the adiabatic expansion and heat absorption of high-pressure liquid nitrogen) through micro-nozzles. Achieving molecular-level cold energy transfer breaks through the efficiency cliff (COP<0.3) of traditional compressor refrigeration below -65℃;

[0077] The transient response of the SiC heating film, based on the wide bandgap characteristics of silicon carbide (3.2eV), enables rapid heat generation through electron transition, with a thermal response time of <50ms, thus preventing thermal breakdown of the device caused by temperature overshoot.

[0078] Temperature field uniformity control: Microchannel turbulence design, through CFD simulation to optimize the channel inclination angle (45°±5°) and pore size distribution (gradient density 0.5~3mm), still maintains surface temperature gradient ≤±1.5℃ at a high temperature of 150℃;

[0079] The thermal expansion compensation algorithm, with a built-in material CTE database, adjusts the position of the support platform in real time to counteract contact failure caused by thermal deformation.

[0080] Humidity control module:

[0081] The staged pressure saturator (0.1~2.5MPa) precisely controls the saturated vapor pressure according to the Clausius-Clapeyron equation, maintaining 98% RH at a high temperature of 85℃;

[0082] The piezoelectric micropump array controls the flow, with each micropump independently controlling a flow rate of 0.1 μL / min, achieving a switch from 10% RH to 95% RH within 20 seconds;

[0083] Critical control for condensation prevention, dynamic dew point tracking, and integration of a Pt100 thin-film temperature sensor (±0.1℃) and a capacitive dew point meter to calculate the temperature-dew point difference ΔT in real time.

[0084] ΔT ≥ 3℃ is considered a safe zone;

[0085] A temperature of 1℃≤ΔT<3℃ indicates a warning state (reducing the humidification rate);

[0086] ΔT<1℃ indicates forced activation of the semiconductor condenser plate (cooling rate 5℃ / s).

[0087] Electromagnetic interference simulation module:

[0088] The implementation path for wideband coverage involves a vector signal generator array that supports continuous frequency sweep from 0.1 to 40 GHz (1 Hz resolution) and synthesizes arbitrary waveforms through IQ modulation.

[0089] 5G FR2 band: 24.25~52.6GHz (supports 256QAM modulation error <1.5%);

[0090] 77 / 79GHz automotive radar: linear frequency modulation slope ±300MHz / μs;

[0091] Guarantee of uniformity of spatial field strength:

[0092]

[0093] S2 specifically includes:

[0094] High-precision acquisition of explicit parameters:

[0095] Voltage / current parameters are measured accurately in the nanoampere level and microvolt level across the entire temperature range using a quantum tunneling current sensor and a Josephson junction voltage reference.

[0096] The rise / fall time and eye opening of high-speed signals are captured by a sampling oscilloscope, and the EMI interference immunity is tested simultaneously by superimposing the data; time-interleaved sampling technology is applied to compress the single measurement time.

[0097] Deep perception of latent parameters:

[0098] Dielectric relaxation spectrum material degradation monitoring: A sinusoidal sweep voltage is applied, and the real and imaginary parts of the dielectric are extracted through a lock-in amplifier and a three-electrode system to calculate the loss factor.

[0099] Transient response of carrier mobility was obtained by applying a nanosecond-level high-voltage pulse and using a 4D carrier imaging system to plot the mobility distribution.

[0100] The degradation model correlation shows that if the temperature coefficient of mobility is greater than a predetermined threshold, it conforms to the Arrhenius model and is judged as normal thermal degradation; if the temperature coefficient of mobility is less than the predetermined threshold, an electromigration risk warning is triggered.

[0101] Thermoacoustic emission signal crack detection employs a multi-modal sensor array, including a piezoelectric sensor to capture the acoustic emission energy of microcrack propagation; and an infrared thermal imager to simultaneously record the transient temperature rise in the crack region.

[0102] The feature extraction algorithm identifies crack initiation if the wavelet packet energy entropy of the acoustic emission signal is greater than a predetermined threshold; and locates the poor weld area of ​​the solder joint if the thermal diffusion rate changes abruptly.

[0103] It should be noted that explicit parameter collection:

[0104] Full-temperature-range nanoampere-level leakage current monitoring: Quantum tunneling sensor, based on the quantum effect of electron wave nature passing through the potential barrier (barrier width ≤1nm), achieved in the range of -70℃ to 200℃, with a resolution of 0.1fA. Low-temperature challenge is achieved by using a graphene-boron nitride heterojunction to counteract the carrier freezing effect (drift <0.05fA at -70℃); High-temperature stability is guaranteed by heat dissipation through a diamond substrate, achieving thermal noise ≤0.2fA at 150℃.

[0105] The Josephson junction voltage reference is traced back to its origins by establishing a quantized voltage reference with an accuracy of 0.02 μV using superconducting quantum interference (SQUID).

[0106]

[0107] In the formula, f is the microwave frequency; h is Planck's constant; and e is the electron charge.

[0108] Latent parameter awareness:

[0109] Dielectric relaxation spectroscopy (DRS) material degradation early warning:

[0110] Three electrodes: The working electrode is a platinum nanopillar array (increasing the effective area by 10 times); the reference electrode is an Ag / AgCl solid electrode (with a stable potential of ±1mV across the entire temperature range); the counter electrode is a porous carbon dispersion that maintains a current of 10 at 85℃ / 85%RH. -4 Stability of low-frequency scanning at Hz.

[0111] Failure criterion mapping relationship (δ value is dielectric loss factor):

[0112] frequency band δ value anomaly characteristics Failure Mechanism Early warning lead time <![CDATA[10 - 3Hz]]> Δδ>0.01 / 24h Epoxy resin hydrolysis >3000 hours 100Hz β-relaxation peak temperature drift >10℃ Plasticizer precipitation >5000 hours 10kHz α relaxation strength decreased by 30%. Crosslinking loss >2000 hours

[0113] Transient Response of Carrier Mobility

[0114] 4D carrier imaging: time dimension femtosecond laser pump-detector (resolution 100 ps);

[0115] Spatial dimension is achieved through confocal microscopy scanning (accuracy 1 μm);

[0116] The energy dimension is the synchrotron X-ray energy spectrum (resolution 0.1 eV);

[0117] The density dimension is reconstructed from the Thomas-Fermi shielding model;

[0118] Mobility temperature coefficient early warning model:

[0119] In the formula, α is the temperature coefficient of mobility; μ is the carrier mobility; and T is the temperature.

[0120] If α > -1.5% / ℃ → normal thermal decay (consistent with the Arrhenius model);

[0121] If α < -2.0% / ℃ → risk of electromigration (Black equation lifetime prediction);

[0122] Thermo-acoustic-electric multimodal crack detection:

[0123] Piezoelectric sensor array:

[0124]

[0125] Transient capture by infrared thermal imager: Crack friction heat generation model.

[0126]

[0127] In the formula, ΔT represents the temperature change; E f Frictional energy; ρ is the material density; c p ρ represents specific heat capacity; V represents volume; a resolution of 0.03℃ can detect localized temperature rises >5℃ (corresponding to 10μm-level cracks).

[0128] Wavelet packet energy entropy diagnostic algorithm:

[0129] Number of decomposition layers: 8;

[0130] Characteristic frequency band: 3.5~10MHz (crack propagation frequency);

[0131] Criterion: Energy entropy > 3.5 → Crack initiation (confidence level 99.2%).

[0132] S3 specifically includes:

[0133] The electrochemical corrosion model takes as input parameters real-time relative humidity from the humidity module, loss factor in the low-frequency band of dielectric relaxation spectrum, and ion migration current monitored by the quantum tunneling current sensor. A corrosion rate equation is established, and the threshold is dynamically adjusted. When the corrosion rate exceeds the predetermined threshold, it is judged as a high risk of electrochemical migration, and the upper limit of the allowable leakage current is automatically tightened from the standard value and the coating thickness is re-inspected.

[0134] The thermomechanical fatigue model takes as input parameters the number of temperature cycles, heating and cooling rates, cumulative thermoacoustic emission energy, and measured values ​​of the thermal expansion coefficient of the packaging material. Lifetime prediction is based on the Coffin-Manson correction formula with dynamic threshold adjustment. If the predicted lifetime is a predetermined number of cycles, three additional extreme temperature shock tests are forcibly added, and the thermoacoustic emission energy threshold is adjusted to the predetermined threshold.

[0135] The dynamic threshold adjustment strategy uses a multi-parameter weighted decision tree to construct an implicit parameter weight matrix and dynamically adjust the weight ratio according to the application scenario. A comprehensive score is calculated, and failure is determined when the comprehensive score is greater than a predetermined threshold, and feedback is given to the test item threshold in real time.

[0136] The machine learning online correction uses a Bayesian optimization algorithm to update the model parameters based on historical data after a predetermined number of tests. Abnormal cases feed back into the model by extracting features from misjudged / missed samples to generate adversarial training data.

[0137] It should be noted that the electrochemical corrosion model (ECM) is as follows:

[0138] Quantum sensing of ion migration current (IonIion) using the tunneling effect at the graphene-electrolyte interface to monitor Cl- / Cu +Ion concentration (resolution 0.1 fA), breaking through the signal-to-noise ratio limitation of traditional electrochemical impedance spectroscopy (EIS) at high temperatures (100-fold improvement in accuracy at 85℃): I ion =q·A·n·v d

[0139] In the formula, q is the charge; A is the area; v d This refers to the drift rate;

[0140] Molecular correlations of dielectric relaxation spectra (δ), low frequency range (10) - The loss factor δ (3Hz) directly reflects the polymer's moisture absorption and expansion rate;

[0141] Establish the corrosion rate equation:

[0142] In the formula, R corr E is the corrosion rate; k is the rate constant, which depends on the material and environmental conditions; a The activation energy is kJ / k, corresponding to the Cl- ion diffusion activation energy, and its unit is electron volt (eV); B is Boltzmann's constant, approximately 1.38 × 10⁻²³ J / K; T is absolute temperature, in Kelvin (K). I represents the rate of change of the loss factor δ over time. ion I is the ion current; ref Reference current;

[0143] Activation energy barrier: E a = 0.65 eV (diffusion barrier of Cl- in Cu2O film);

[0144] Dynamic response logic:

[0145]

[0146] Thermomechanical fatigue model (TMF):

[0147] The lifetime prediction equation is based on the Coffin-Manson modified formula:

[0148]

[0149] In the formula, N f Fatigue life; C is a constant that depends on the material and environmental conditions; Δ∈ eff For effective strain; β is a material parameter, ranging from an empirical value of 0.5 to a measured value of 0.3 for molding compound; γ represents the rate of temperature change; γ is the sensitivity coefficient of the temperature change rate measured by high-speed infrared thermal imaging (γ = 0.15 for SnAgCu solder).

[0150] Effective response:

[0151] In the formula, Δ∈ eff For effective strain; CTE is the coefficient of thermal expansion; ΔT is the temperature change; α is the acoustic energy-strain conversion factor, α = 2.5 × 10⁻⁶. -3 E TAE Thermoacoustic emission energy;

[0152] Breakthrough in the experimental measurement of material parameters:

[0153] β value: The molding compound value changed from an empirical value of 0.5 to a measured value of 0.3 (using laser scattering method for a certain FCBGA package);

[0154] γ value: Sensitivity coefficient of temperature change rate measured by high-speed infrared thermal imaging (γ = 0.15 for SnAgCu solder).

[0155] S4 specifically includes:

[0156] Extreme temperature cycling is achieved by alternating liquid nitrogen injection and silicon carbide infrared heating to reach the limit of a predetermined temperature; based on the number of intelligent matching cycles according to the target device lifespan, it is divided into consumer electronics, automotive electronics, and aerospace devices; failure excitation utilizes the CTE mismatch effect to forcibly expose solder joint microcracks and through-silicon via fractures.

[0157] Bias-temperature stress is applied using an electrothermal coupling loading mode. The static bias is applied as a predetermined multiple of the rated voltage, while the ambient temperature is simultaneously raised to a predetermined temperature range. Dynamic disturbance is applied by superimposing square wave switching stress to simulate the actual switching losses of power devices. The sudden increase in gate oxide tunneling current is captured in real time by an in-situ quantum tunneling sensor, and the hot spot area is located by combining infrared thermal imaging.

[0158] The model quantifies the amount of hidden drift and the dielectric relaxation spectrum drift. It calculates the rate of change of the loss factor before and after accelerated aging in the low-frequency band. When the rate of change before and after accelerated aging is greater than a predetermined proportion, the polymer encapsulation is determined to have failed hydrolysis. When the shift of the characteristic relaxation peak is greater than a predetermined threshold, the concentration of ionic impurities is indicated to be excessive.

[0159] Carrier mobility decay, extract mobility degradation rate under pulse test, thermoacoustic emission energy accumulation, and statistically analyze the total energy of acoustic emission signal during acceleration.

[0160] Dynamic elimination criteria, multi-parameter fusion decision tree, parameter types include dielectric loss variation, mobility decay, thermoacoustic energy, and quantum tunneling current.

[0161] It should be noted that extreme temperature cycling:

[0162] Failure triggering mechanism under extreme temperature load

[0163] Liquid nitrogen injection (-196℃) → SiC infrared heating (+300℃), CTE mismatch effect is excited through thermal shock stress (copper leads 17ppm / ℃ vs ceramic substrate 7ppm / ℃);

[0164] Intelligent matching of loop counts based on specific scenarios:

[0165]

[0166] In-situ diagnosis of microcracks:

[0167] Acoustic emission sensor array to locate crack propagation events with energy >10mJ (corresponding to cracks >100μm);

[0168] X-ray in situ imaging: once every 50 cycles (resolution 0.5 μm);

[0169] Bias-temperature stress (BTS) coupling:

[0170] Third-order stress due to electrical-thermal-dynamic disturbance:

[0171]

[0172] A quantum tunneling current sensor, based on a graphene-silicon dioxide quantum dot structure (barrier thickness ≤ 1 nm), achieves a gate current resolution of 0.1 nA at a high temperature of 200 °C, enabling the localization of single-trap charge trapping events.

[0173] Latent drift quantification model:

[0174] Dielectric relaxation spectrum drift, polymer hydrolysis failure:

[0175] (Low frequency band 10-3Hz)

[0176] In the formula, Δδ represents the change in the loss factor; δ after The loss factor after filtering; δ before To accelerate the loss factor before aging;

[0177] Hydrolysis of ester bonds in epoxy resins leads to an increase in polar groups (FTIR verification showed a 30% enhancement of the C=O peak);

[0178] Ion pollution warning: characteristic relaxation peak shift >5℃ → corresponding Cl- concentration >50ppm;

[0179] Defect localization of carrier mobility decay:

[0180] 4D Imaging Degradation Rate Calculation: (Unit: % / h)

[0181] In the formula, μ0 is the initial mobility (reference value at 300K); μt Post-stress migration rate; t is time; spatial location:

[0182] Thermoacoustic emission energy accumulation model:

[0183]

[0184] In the formula, E sum To calculate the total energy of the acoustic emission signal during the acceleration process; n is the number of signals; t is time; Let f be the square of the amplitude of the i-th signal at frequency f; f is a frequency range from 1 to 40MHz.

[0185] Solder joint fatigue: E sm >80mJ→Crack length>150μm (based on Griffith fracture mechanics);

[0186] Interface layering: Energy entropy mutation > 2.0 → layering area > 5%.

[0187] 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 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 claimed invention. The scope of protection claimed by the appended claims and their equivalents is defined.

Claims

1. An automated testing and screening method of electronic components, characterized by, Comprise: S1, generate a multi-dimensional stress field through a temperature control module, a humidity control module, and an electromagnetic interference simulation module; Based on the target application scenario of the component, call the preset extreme working condition combination sequence; S2, in a dynamically coupled environment, synchronously collect explicit parameters and implicit parameters including dielectric relaxation spectrum, carrier mobility transient response, and thermal acoustic emission signal; S3, input the implicit parameter time series data into a physical failure model including an electrochemical corrosion model and a thermal mechanical fatigue model; According to the model output, dynamically adjust the passing threshold of the test item; S4, apply accelerated degradation excitation to the components that pass the screening, including: a certain number of temperature cycles; apply bias-temperature stress; repeat step S2 to monitor the drift of the implicit parameters, and eliminate the devices with excessive drift.

2. The method of claim 1, wherein the method further comprises: S3 specifically includes: The electrochemical corrosion model inputs parameters including real-time relative humidity fused with the humidity module, dielectric relaxation spectrum low-frequency loss factor, and ion migration current monitored by the quantum tunneling current sensor; a corrosion rate equation is established, and the threshold is dynamically adjusted; when the corrosion rate is greater than a certain threshold, it is determined that the electrochemical migration is high risk, the maximum leakage current is tightened from the standard value to, and the plating thickness reinspection is triggered; The thermal mechanical fatigue model inputs parameters including the number of temperature cycles collected, the temperature rise and fall rate, the thermal acoustic emission energy cumulative value, and the measured value of the thermal expansion coefficient of the packaging material; life prediction based on the Coffin-Manson correction formula, threshold dynamic adjustment, if the predicted life is a certain number of cycles, then three times of extreme temperature impact test is forcibly added, and the thermal acoustic emission energy threshold is adjusted to a certain threshold.

3. The method of claim 2, wherein the method further comprises: S3 specifically includes: Dynamic threshold adjustment strategy, multi-parameter weighted decision tree, construct implicit parameter weight matrix, dynamically adjust weight proportion according to application scenario; comprehensive score calculation, when the comprehensive score is greater than a certain threshold, it is determined to fail, and real-time feedback to the test item threshold; Machine learning online correction, using Bayesian optimization algorithm, based on historical data to update model parameters after completing a certain number of tests; abnormal case feedback model, feature extraction is performed on misjudgment / missed samples to generate adversarial training data.

4. The method of claim 3, wherein the method further comprises: S1 specifically includes: Temperature control module: three-level temperature control architecture of liquid nitrogen jet refrigeration, semiconductor peltier stack, and silicon carbide film instantaneous heating; Through the high-pressure liquid nitrogen micro-spraying system, the device surface is instantaneously cooled; using SiC ceramic heating film, rapid heating in a short time is realized, and closed-loop control is realized with infrared temperature sensor; a micro-channel turbulent system is built in the test cavity, the airflow distribution is optimized through computational fluid dynamics, and the device surface temperature gradient is ensured to be less than or equal to a certain temperature.

5. The method of claim 4, wherein the method further comprises: S1 specifically includes: Humidity control module: based on double-pressure method saturated water vapor dynamic injection and condensation dew point precise regulation; Wide range of humidity coverage, generate a certain range of relative humidity through staged pressurized saturators, and still maintain a certain relative humidity at a certain high temperature; use piezoelectric micropump array to accurately control water vapor injection amount, realize rapid humidity switching; anti-condensation design, integrate semiconductor condensing plate at the bottom of the device bearing table, real-time monitor surface temperature and dew point difference.

6. The method of claim 5, wherein the method further comprises: S1 specifically includes: Electromagnetic interference simulation module: build programmable composite frequency spectrum interference field, including wideband coverage, synthesis arbitrary waveform interference through vector signal generator array, support 5G millimeter wave, automotive radar high frequency scene; Spatial field intensity control, adopt waveguide slot antenna matrix, cooperate with electromagnetic simulation software to optimize layout; Synchronous coupling mechanism, share timestamp with temperature and humidity module, superimpose electromagnetic pulse group at the inflection point of temperature cycle, simulate the multi-stress synergistic effect of real environment.

7. The method of claim 6, wherein the method further comprises: The S2 specifically includes: High-precision acquisition of explicit parameters: Voltage / current parameters, quantum tunneling current sensor and Josephson junction voltage reference are used to realize accurate measurement of nanampere-level leakage current and microvolt-level voltage drop in full-temperature range; Capture the rise / fall time and eye opening of high-speed signals through a sampling oscilloscope, and superimpose EMI interference test immunity synchronously; Apply time-interleaved sampling technology to compress single measurement time.

8. The method of claim 7, wherein the method further comprises: The S2 specifically includes: Deep perception of implicit parameters: Dielectric relaxation spectrum material degradation monitoring, apply sinusoidal sweep voltage, extract dielectric real part and imaginary part through lock-in amplifier and three-electrode system, calculate loss factor; Carrier mobility transient response, apply nanosecond-level high-voltage pulse, use 4D carrier imaging system to draw mobility distribution map; Degradation model correlation, if the temperature coefficient of mobility is greater than a certain threshold, it meets the Arrhenius model, and is determined to be normal thermal attenuation; If the temperature coefficient of mobility is less than a certain threshold, trigger electrical migration risk warning; Thermoacoustic emission signal crack detection, use multi-modal sensing array, including piezoelectric sensor, capture acoustic emission energy of micro-crack propagation; Infrared thermal imager, synchronously record transient temperature rise of crack area; Feature extraction algorithm, if the wavelet packet energy entropy of acoustic emission signal is greater than a certain threshold, identify crack initiation; If the thermal diffusion rate suddenly changes, locate the virtual soldering area of the solder joint.

9. The method of claim 8, wherein the method further comprises: The S4 specifically includes: Extreme temperature cycling, use liquid nitrogen spraying and silicon carbide infrared heating to realize extreme cycling of a certain temperature; According to the device life target, intelligently match the number of times, divided into consumer electronics, automotive electronics, spacecraft devices; Failure excitation, use CTE mismatch effect to force exposure of solder micro-crack and through silicon via fracture; Bias-temperature stress, use electrothermal coupling loading mode, static bias is to apply a certain multiple of rated voltage, and the environmental temperature is simultaneously raised to a certain temperature range; Dynamic disturbance, superimpose square wave switching stress, simulate the actual switching loss of power devices; Capture the sudden increase of gate oxide tunneling current in real time through in-situ quantum tunneling sensor, and locate the hot spot area combined with infrared thermal imaging.

10. The method of claim 9, wherein the method further comprises: The S4 specifically includes: Implicit drift amount quantization model, dielectric relaxation spectrum drift, calculate the change rate of loss factor in low frequency band before and after accelerated aging, when the change rate is greater than a certain proportion, determine that the polymer packaging hydrolysis failure; When the characteristic relaxation peak shift is greater than a certain threshold, identify that the ion impurity concentration exceeds the standard; Carrier mobility attenuation, extract the mobility degradation rate under pulse test, accumulate thermoacoustic emission energy, and count the total energy of acoustic emission signals during acceleration process; Dynamic elimination criterion, multi-parameter fusion decision tree, parameter types include dielectric loss change, mobility attenuation, thermoacoustic energy, quantum tunneling current.

Images