CN-122019961-A - Transient thermal resistance curve fitting method for IGBT module

CN122019961ACN 122019961 ACN122019961 ACN 122019961ACN-122019961-A

Abstract

The invention provides a transient thermal resistance curve fitting method of an IGBT module, which relates to the technical field of semiconductor device testing and comprises the steps of obtaining transient thermal resistance data of the IGBT module through actual measurement; and constructing a fourth-order polynomial function to fit the smoothed data to obtain a transient thermal resistance curve fitting result. The method has the advantages that based on the transient thermal resistance data of the IGBT module obtained through actual measurement, the Kalman filtering is introduced to conduct forward filtering and backward smoothing on the data, external interference noise existing in the actual measurement is effectively restrained, real characteristics of the curve are reserved, the nonlinear least square and fourth-order polynomial function fitting is combined, complex nonlinear dynamics of the thermal resistance curve are accurately captured, a complete data processing and model optimizing flow is constructed, smoothness, accuracy and robustness of the transient thermal resistance fitting curve are remarkably improved, and more reliable technical support is provided for thermal characteristic analysis, state monitoring and service life assessment of the IGBT module.

Inventors

SHEN JIATAO
GU ZHAOQI
Shen Renfeng
LIU WEIDONG

Assignees

嘉兴斯达微电子有限公司

Dates

Publication Date: 20260512
Application Date: 20251208

Claims (9)

1. The transient thermal resistance curve fitting method for the IGBT module is characterized by comprising the following steps of: Step S1, obtaining transient thermal resistance data of an IGBT module through actual measurement; s2, carrying out Kalman filtering smoothing on the transient thermal resistance data to obtain smoothed data; step S3, a fourth-order polynomial function is constructed to fit the smoothed data to obtain a transient thermal resistance curve fitting result of the IGBT module, and the fourth-order polynomial function has the following expression: Wherein r 1 ～r 4 ,τ 1 ～τ 4 is a fitting coefficient larger than zero obtained by fitting by using a nonlinear least square method, and r 1 +r 2 +r 3 +r 4 ＝Z th_jc ,Z th_jc is a steady-state value of thermal resistance.
2. The method for fitting the transient thermal resistance curve of the IGBT module according to claim 1, wherein the step S1 includes: Step S11, applying a step electric power to the IGBT module by using a power circulation test bench, controlling the temperature of a substrate of the IGBT module by a water cooling method, and measuring the real-time temperature of the substrate by using a thermocouple method; Step S12, recording the chip real-time junction temperature corresponding to the IGBT module built-in chip in the process from the step electric power application to the heat balance; And step S13, calculating to obtain a plurality of discrete transient thermal resistance data according to the chip real-time temperature and the substrate real-time temperature.
3. The method for fitting the transient thermal resistance curve of the IGBT module according to claim 2, wherein in step S11, the real-time temperature of the substrate is obtained by measuring a calibration curve of the saturation voltage and the temperature of the chip built in the IGBT module.
4. The method for fitting the transient thermal resistance curve of the IGBT module according to claim 2, wherein in step S3, the transient thermal resistance data is calculated according to the following formula: Wherein Z th_jc (T) is the transient thermal resistance data at time T, T j (T) is the chip real-time junction temperature at time T, T c (T) is the substrate real-time temperature at time T, and P loss is the power loss of the applied step electric power.
5. The IGBT module transient thermal resistance curve fitting method according to claim 1, further comprising preprocessing the transient thermal resistance data before executing the step S2, the preprocessing process comprising: A1, arranging the transient thermal resistance data according to time sequence to form a two-dimensional data table; Step A2, eliminating mutant abnormal points and drift abnormal points in the two-dimensional data table to obtain an eliminated data table; and step A3, intercepting the data of the effective interval of transient thermal resistance change from the data table after elimination to obtain preprocessed data.
6. The method for fitting transient thermal resistance curve of IGBT module according to claim 1, wherein said transient thermal resistance data is a plurality of discrete thermal resistance data sequences arranged in time sequence, said step S2 comprises: Step S21, for each transient thermal resistance data, recursively calculating a state estimation value and covariance of each time point by using the current time point and observation data before the current time point, and further performing forward filtering on the thermal resistance data sequence to obtain a filtered data sequence; Step S22, for the filtered data sequence, reverse recursion is started from the last time point by utilizing the observation information of all time sequences, and the state estimation value of each time point is corrected and optimized to realize backward smoothing of the filtered data sequence, and finally the smoothed data is obtained.
7. The method according to claim 1, wherein the fitting coefficient greater than zero obtained by the nonlinear least square fitting in the step S3 includes performing initial parameter estimation and setting a parameter boundary, and then optimizing and searching for an optimal parameter as the fitting parameter by using the nonlinear least square method.
8. The method of fitting a transient thermal resistance curve of an IGBT module of claim 7, wherein the optimizing the search for the optimal parameters using a nonlinear least squares method comprises: and carrying out iterative updating on the initial parameters by minimizing the sum of squares of residual errors between the model predicted value and the smoothed data, and stopping iteration when the iteration times reach the preset optimized iteration times or the variation of the initial parameters is smaller than a preset parameter variation limiting value and the changed parameter values are within the parameter boundaries, so as to obtain the fitting parameters.
9. The method for fitting a transient thermal resistance curve of an IGBT module according to claim 1, further comprising generating a process data analysis map for visual presentation during the execution of steps S1 to S3, and visually presenting the transient thermal resistance curve fitting result after the execution of step S3 is completed.

Description

Transient thermal resistance curve fitting method for IGBT module Technical Field The invention relates to the technical field of semiconductor device testing, in particular to a transient thermal resistance curve fitting method of an IGBT module. Background With rapid development of power semiconductor technology, insulated Gate Bipolar Transistor (IGBT) modules have become core power semiconductor devices in the middle and high power fields of new energy power generation (such as photovoltaic inverters and wind power converters), smart grids (flexible direct current transmission and reactive compensation devices), industrial frequency conversion (motor driving systems), electric vehicles (whole vehicle controllers and vehicle-mounted chargers) and the like by virtue of high voltage resistance, high current, low loss and rapid switching characteristics. As a core of electric energy conversion and control, the long-term operation reliability of the IGBT module is directly related to the safety and efficiency of the whole system, and performance degradation or sudden failure of the IGBT module often causes system shutdown, equipment damage and even causes safety accidents, resulting in significant economic loss. In actual working conditions, the IGBT module generates a large amount of heat due to power loss, so that junction temperature is rapidly increased and is severely fluctuated, faults such as chip aging and bonding wire falling off can be caused, and the faults are key factors for limiting the service life of the IGBT module. Therefore, the accurate characterization and real-time monitoring of the thermal characteristics of the IGBT module become an important link for improving the reliability of the system. The transient thermal resistance curve is a core index describing the internal thermal dynamic response characteristics of the IGBT module, reflecting the temperature change from junction to shell at step power. By analyzing the curve, key thermal parameters can be extracted, an equivalent thermal model can be constructed, and the curve can be used for online junction temperature estimation and life prediction. At present, in order to obtain an accurate transient thermal resistance curve from measured data, a mathematical method is generally required to fit discrete measurement points. Common fitting methods include: 1. According to the exponential fitting method, based on a Cauer or Foster thermal network model, the thermal relaxation process with multiple time constants is represented through superposition of multiple exponential functions, and the degree of matching with a thermal transfer physical mechanism of an IGBT module is higher in theory, so that the fitting precision is relatively higher. However, the model order selection is highly dependent on engineering experience and priori knowledge, is sensitive to measurement noise, and small noise interference can cause severe fluctuation of exponential coefficients, so that the fitting result deviates from the real thermal characteristics. 2. The method divides the whole transient thermal resistance curve into a plurality of linear segments according to time scale, approximates the thermal response process by piecewise fitting a straight line, and has the characteristics of simple algorithm, high calculation efficiency and low engineering realization difficulty. However, the method has the defects in the aspects of curve smoothness and derivative continuity, obvious step distortion is easy to occur at the segment nodes, and the dynamic details of thermal behaviors are difficult to describe accurately; 3. The traditional least square fitting method is used for determining fitting parameters by minimizing the sum of squares of residual errors of model predicted values and measured data, and is a universal fitting means widely applied. However, in transient thermal resistance data processing, overall fitting deviation or local overfitting is easy to cause, and especially, robustness is lacking for abnormal fluctuation in data, and the fitting degree is still to be improved. In the comprehensive view, the conventional fitting method has common problems of sensitivity to noise and abnormal values in original data, large interference on fitting results, poor stability, multiple model parameters, complex optimization process, low calculation efficiency, inconsistent fitting precision under different temperature areas or time scales, insufficient generalization capability, and difficulty in realizing high-precision and high-robustness automatic fitting in engineering due to the fact that the flow does not integrate data processing and fitting optimization systematically. Therefore, a transient thermal resistance curve fitting method which has both data noise reduction capability and nonlinear fitting precision is continuously developed. Disclosure of Invention Aiming at the problems in the prior art, the invention provides a transient thermal resista