Prediction of IGBT junction temperature using improved cuckoo search-based extreme learning machine

doi:10.1016/j.microrel.2021.114267

Microelectronics Reliability

Volume 124, September 2021, 114267

https://doi.org/10.1016/j.microrel.2021.114267 Get rights and content

Highlights

•
The junction temperature prediction using ICS-ELM model is insensitive to the measured-location.
•
The relationship between the junction temperature and VCE(sat), Ic and the aging cycles number can be formulated.
•
The determination coefficient (R2) by ICS-ELM model can achieve the optimal value.

Abstract

The insulated-gate bipolar transistor (IGBT) is one of the most widely used power transistors in switching and industrial control systems. Its actual junction temperature plays a critical factor in determining the dynamic performance, reliability and life-time of the device. Although some noninvasive measurement methods such as optical and physical contact methods may be used to estimate the junction temperature, the measurement accuracy is very sensitive to the measured position. Therefore, the prediction using cuckoo search-based extreme learning machine for junction temperature is developed to reach a high-accuracy solution without measured-location sensitivity. Firstly, the accelerated aging and single pulse tests in IGBT are implemented to collect the IGBT failure related parameters, e.g. collector-emitter saturation voltage (V_CE(sat)), junction temperature, collector current (I_c) and the aging cycles number. With the curved surface fitting for the collected data, the relationship between the junction temperature and the other parameters can be formed. Based on the extreme learning machine optimized by the improved Cuckoo Search method, called ICS-ELM, V_CE(sat), I_c and the aging cycles number are taken as input, and the output is the predicted junction temperature. The performance results reveal that the determination coefficient (R²) by ICS-ELM model achieves the optimal value, i.e. 0.9975, which is superior to the curved surface fitting method, Cuckoo search optimizing extreme learning machine, support vector machine and extreme learning machine.

Introduction

IGBT is now widely applied in power systems such as electric vehicles and high-speed traction, etc. [1]. However, the failure rate caused by power semiconductor devices and the photovoltaic inverter with IGBT accounts for about 34% and 37% of the total failures, respectively [2], [3]. Accordingly, the IGBT reliability should be accurately evaluated for IGBT-based device operation.

It is known that some crucial factors causing IGBT failures include collector-emitter saturation voltage (V_CE(sat)), collector current (I_c), junction temperature (T_j), thermal resistance and case temperature (T_c), etc. [4]. Among them, the failure of power electronic systems related to the temperature rising accounts for 55% of the total failures [5], [6], [7]. Literature [8], [9] indicated that junction temperature is a key factor to seriously affect IGBT life and reliability.

At present, the direct junction temperature measurement based on optical or contact devices is used to measure the IGBT junction temperature. However, it may suffer from some disadvantages like high cost, inaccuracy or physical device damage [10]. Alternatively, there are three major indirect junction temperature measurement methods, e.g. physical model [11], [12], probability statistics [13], [14] and machine learning [15], [16]. The physical model is established on the basis of the internal physical structure, material properties and mechanical properties of devices [17]. The probability statistics method is applied to form a temperature distribution function from experimental data [18]. The machine learning models [19] are developed to find the relationship between the independent variables and junction temperature in the IGBT.

The content structure of this article is as follows: The second section reviews existing methods such as direct methods, indirect methods, and artificial intelligence methods for measuring the junction temperature. The third section obtains the experimental data of IGBT through IGBT experiment, processes the experimental data, and proposes a multi-parameter IGBT junction temperature prediction model. In the fourth section, junction temperature prediction model is established. Section 5 analyzes and discusses the junction temperature prediction results of each model. The sixth section is the conclusion of this article.

Section snippets

Literature review

The reliability of the IGBT can be evaluated via directly or indirectly measuring the junction temperature. Optical measurement method and physical contact method [20], [21] are classified as direct methods. On the other hand, physical models, probability statistics and machine learning are indirect methods, which have attracted more attention currently [22], [23], [24], [25]. For example, a physics-based transient electrothermal model proposed in literature [23] considered the

Set-up of the data collection experiment

In order to obtain the experimental data for V_CE(sat), junction temperature, I_c and number of cycles of the IGBT, this study designs the accelerated aging and the single pulse tests. The steps of the data collection experiment are illustrated as follows. Firstly, perform the accelerated aging test. After every 1000 times test is completed, the IGBT module is put into the thermostat. Then, adjust the thermostat temperature by experimental setting at the range of [0 °C, 100 °C] until it reaches

The principle of ELM

ELM is an algorithm based on the single hidden layer feedforward neural network with fast learning speed and high predication accuracy, being widely used in various fields [44].

Firstly, N sets data is selected as the input sample X, where x_j = [x_j1, x_j2, ⋯, x_jn]^⊤; Y is the output sample, y_j = [y_j1, y_j2, ⋯, y_jm]^⊤; V is taken as the number of hidden layer nodes.

Sigmoidal function is chosen as the activation function of the hidden layer, which is shown in Eq. (1). $g (x) = \frac{1}{1 + e^{- x}}$

The relationship between the

Performance analysis

In the study, total 386 groups of data obtained from IGBT tests are used for T_j prediction. In the model, V_CE(sat), I_c and the aging cycle number are taken as the input variables, while T_j is taken as the output variable. The first 270 groups are used for training process, and the last 116 groups are used for testing process. For comparison, ICS-ELM, CS-ELM, ELM and Support Vector Machine (SVM) are involved in the prediction of IGBT junction temperature. In addition, Mean Absolute Error (MAE),

Conclusions

It is well known that the junction temperature of IGBT is regarded as the most crucial factor to determine its working efficiency as well as failure situation. With increasing power electronics applications in industry, more attention has been drawn on the reliability, safety and stability in the device with IGBT. In the proposed ICS-ELM model, the IGBT junction temperature was firstly selected as the main failure precursor. Through accelerated aging and single pulse tests in IGBT, the

CRediT authorship contribution statement

Conceptualization, Boying Liu, Guolong Chen; Methodology, Boying Liu and Guolong Chen.; Software, Boying Liu., Guolong Chen and Hsiung-Cheng Lin; Formal analysis, Boying Liu and Hsiung-Cheng Lin.; Data curation, Boying Liu, Guolong Chen and Weipeng Zhang; Writing—original draft preparation, Boying Liu, Guolong Chen, Hsiung-Cheng Lin, Weipeng Zhang and JiaQi Liu.; Writing—review and editing, Hsiung-Cheng Lin, Weipeng Zhang and JiaQi Liu. All authors have read and agreed to the published version

Declaration of competing interest

The authors declare that we have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

References (51)

L.W Zhou et al.
Junction temperature management of IGBT module in power electronic converters
Microelectron. Reliab.
(2014)
J.Y. Xu et al.
"A unified MMC reliability evaluation based on physics-of-failure and SM lifetime correlation," (in English)
Int. J. Electr. Power Energy Syst.
(Mar 2019)
L.L. Li et al.
Extreme learning machine optimized by whale optimization algorithm using insulated gate bipolar transistor module aging degree evaluation
Expert Syst. Appl.
(2019)
B. Gao et al.
A temperature spectrum density distribution based condition evaluation method and application in IGBT
Appl. Therm. Eng.
(2016)
G.B. Huang et al.
Extreme learning machine: theory and applications
Neurocomputing
(Dec 2006)
N. Sun et al.
An adaptive dynamic short-term wind speed forecasting model using secondary decomposition and an improved regularized extreme learning machine
Energy
(2018)
A.R. Lima et al.
Variable complexity online sequential extreme learning machine, with applications to streamflow prediction
J. Hydrol.
(Dec 2017)
Z.Y. Zhou et al.
Regularization incremental extreme learning machine with random reduced kernel for regression
Neurocomputing
(2018)
Z.L. Sun et al.
Sales forecasting using extreme learning machine with applications in fashion retailing
Decis. Support. Syst.
(Dec 2008)
E. Valian et al.
Improved cuckoo search for reliability optimization problems
Comput. Ind. Eng.
(Jan 2013)

H. Zhang et al.

Optimal heat exchanger network synthesis based on improved cuckoo search via Lévy flights

Chem. Eng. Res. Des.

(2018)

Z.-F. Liu et al.

Short-term photovoltaic power prediction on modal reconstruction: a novel hybrid model approach

Sustain. Energy Technol. Assessments

(2021)

K. Gorecki et al.

Nonlinear compact thermal model of the IGBT dedicated to SPICE

IEEE Trans. Power Electron.

(Dec 2020)

M.Y. Ma et al.

In-situ health monitoring of IGBT modules of an on-line medium-voltage inverter system using industrial internet of things

CSEE J. Power Energy Syst.

(Sep 2020)

S. Shao et al.

Tunnel magnetoresistance-based short-circuit and over-current protection for IGBT module

IEEE Trans. Power Electron.

(Oct. 2020)

A. Ismail et al.

Remaining useful life estimation for thermally aged power insulated gate bipolar transistors based on a modified maximum likelihood estimator

Int. Trans. Electr. Energy Syst.

(Jun 2020)

S.Y. Yang et al.

An industry-based survey of reliability in power electronic converters

IEEE Trans. Ind. Appl.

(May-Jun 2011)

A. Bryant

Investigation into IGBT dV/dt during turn-off and its temperature dependence

IEEE Trans. Power Electron.

(Oct 2011)

Y.L. Huang et al.

Failure mechanism of die-attach solder joints in IGBT modules under pulse high-current power cycling

IEEE J. Emerg. Sel. Top. Power Electron.

(Mar 2019)

U.M. Choi et al.

Validation of lifetime prediction of IGBT modules based on linear damage accumulation by means of superimposed power cycling tests

IEEE Trans. Ind. Electron.

(Apr 2018)

B. Ji

"In situ diagnostics and prognostics of solder fatigue in IGBT modules for electric vehicle drives," (in English)

IEEE Trans. Power Electron.

(Mar 2015)

H.P. Liu et al.

"Reliability analysis of the optimized Y-source inverter with clamping circuit," (in English)

Microelectron. Reliab.

(Sep 2019)

J.B. Zhou et al.

"Electro-thermal-mechanical multiphysics coupling failure analysis based on improved IGBT dynamic model," (in English)

IEEE Access

(2019)

G.T. Lv et al.

"Reliability analysis and design of MMC based on Mission profile for the components degradation," (in English)

IEEE Access

(2020)

S.H. Ali et al.

Lifetime estimation of discrete IGBT devices based on Gaussian process

IEEE Trans. Ind. Appl.

(Jan-Feb 2018)

Cited by (19)

Hybrid domain-based fast transient thermal evaluation method for power semiconductor modules in EV motor drive
2022, Microelectronics Reliability
Citation Excerpt :
Offline thermal evaluation at the early design stage is essential to minimize the unnecessary/wasted design margin of power devices and assess the reliability of the converter [1,2].
Performance screening and reliability evaluation of electric vehicle (EV) motor drives require fast and precise thermal evaluation of power devices in both steady-state and transient-state analysis. In this paper, a hybrid-domain calculation method is proposed to fulfil the requirement. Firstly, the matrix form model is selected by characterizing the thermal response of an EV module. To quantitatively investigate the speed of different calculation methods, a generalized calculation burden evaluation is conducted based on counting the number of algebraic operations. Both the steady-state and transient-state scenarios for EV motor drives are considered. Accordingly, a hybrid calculation method for transient analysis is proposed by combining the advantages of time-domain and frequency-domain methods. Its advantage is validated by evaluating the elapsed time for a real case. Since numerous evaluation points are required for the performance screening of converters at the early design stage, the proposed method accelerates the thermal evaluation process without reducing the accuracy.
An improved extreme learning machine with self-recurrent hidden layer
2022, Advanced Engineering Informatics
Citation Excerpt :
In [19], Lu et al. developed a model of transformer fault diagnosis via improved empirical wavelet transform (IEWT) and salp swarm algorithm (SSA) optimize kernel extreme learning machine (KELM), which has higher accuracy and robustness. According to the research by Liu et al. [20], a cuckoo search-based extreme learning machine (ICS-ELM) has been designed, which can predict junction temperature without measured-location sensitivity, the results show that the method reaches a high accuracy and fitting degree. In [21], Wang et al. proposed an ELM combined with multiple-point-AdaBoost to predict the wind speed, which enhanced the wind speed prediction at the target location and provides a new idea for the field of wind speed prediction.
Extreme learning machine (ELM) is widely used in complex industrial problems, especially the online-sequential extreme learning machine (OS-ELM) plays a good role in industrial online modeling. However, OS-ELM requires batch samples to be pre-trained to obtain initial weights, which may reduce the timeliness of samples. This paper proposes a novel model for the online process regression prediction, which is called the Recurrent Extreme Learning Machine (Recurrent-ELM). The nodes between the hidden layers are connected in Recurrent-ELM, thus the input of the hidden layer receives both the information from the current input layer and the previously hidden layer. Moreover, the weights and biases of the proposed model are generated by analysis rather than random. Six regression applications are used to verify the designed Recurrent-ELM, compared with extreme learning machine (ELM), fast learning network (FLN), online sequential extreme learning machine (OS-ELM), and an ensemble of online sequential extreme learning machine (EOS-ELM), the experimental results show that the Recurrent-ELM has better generalization and stability in several samples. In addition, to further test the performance of Recurrent-ELM, we employ it in the combustion modeling of a 330 MW coal-fired boiler compared with FLN, SVR and OS-ELM. The results show that Recurrent-ELM has better accuracy and generalization ability, and the theoretical model has some potential application value in practical application.
Prediction of the void formation in no-flow underfill process using machine learning-based algorithm
2022, Microelectronics Reliability
Citation Excerpt :
The design parameters in the conventional capillary underfill process have been optimized analytically [9], but to date, there is no relevant design optimization work being reported for the no-flow underfill process. Several optimizations have been documented such as do comparison between decision trees, adaptive neuro fuzzy inference system (ANFIS) and artificial neural networks to optimize the stencil printing process for the pin-in-paste technology [10], using big data analytics in machine learning framework to predict the reliability of solder joints [11], prediction of component self-alignment in soldering using machine learning-based prediction method [12] and prediction of transistor junction temperature using cuckoo search-based extreme machine learning [13]. Even for classification techniques machine learning has been utilized by Huang et al. [14] to classify the defects in through silicon (TSV) and normal TSV.
This paper investigated the variation variables on the void formation in the no-flow underfill process. A total of 96 distinct no-flow underfill cases were numerically simulated to determine the void formations for different combinations of underfill parameters. The numerical model has been well validated with the experiment, for which the discrepancies are less than 4 %. It was found that the void formation rate increases with the chip placement speed however it decreases with the increase in bump pitch. The highest chip placement speed of 14 mm/s produces 4–6 % meanwhile, the low chip placement speed (2–5 mm/s) produces around 2–3.5 % of void formation. A supervised machine learning (ML) approach is implemented using the validated numerical results to build a prediction-based method for no-flow underfill process. Three ML methods were trained and tested. Neural networks regression was recommended compared to other methods since it has the highest R-squared value and lowest error: 0.95159 and 0.15885, respectively. Chip placement speed obtained the highest score of 1.791641 for permutation feature in ML, which indicates the most significant variable affecting the package's void formation. The obtained result will aid the industry to choose the most favourable no-flow underfill process parameters regarding the void formation issue, thus increasing the reliability of the package.
Offshore wind speed short-term forecasting based on a hybrid method: Swarm decomposition and meta-extreme learning machine
2022, Energy
As the share of global offshore wind energy in the electricity generation portfolio is rapidly increasing, the grid integration of large-scale offshore wind farms is becoming of interest. Due to the intermittency of wind, the stability of power systems is challenging. Therefore, accurate and fast offshore short-term wind speed forecasting tools play important role in maintaining reliability and safe operation of the power system. This paper proposes a novel hybrid offshore wind forecasting model based on swarm decomposition (SWD) and meta-extreme learning machine (Meta-ELM). This approach combines the advantages of SWD which has proven efficiency for non-stationary signals, with Meta-ELM which provides faster calculation with a lower computational burden. In order to enhance accuracy and stability, the signal is decomposed by implementing a swarm-prey hunting algorithm in SWD. To validate the model, a comparison against four conventional and state-of-the-art hybrid models is performed. The implemented models are tested on two real wind datasets. The results demonstrate that the proposed model outperforms the counterparts for all performance metrics considered. The proposed hybrid approach can also improve the performance of the Meta-ELM model as a well-known and robust method.
Correlation Study on Voiding in Underfill of Large Quantity Ball Grid Array Chip Using Machine Learning
2025, Journal of Electronic Packaging, Transactions of the ASME
Target adaptive extreme learning machine for transfer learning
2024, International Journal of Machine Learning and Cybernetics

View all citing articles on Scopus

View full text

Prediction of IGBT junction temperature using improved cuckoo search-based extreme learning machine

Highlights

Abstract

Introduction

Section snippets

Literature review

Set-up of the data collection experiment

The principle of ELM

Performance analysis

Conclusions

CRediT authorship contribution statement

Declaration of competing interest

Microelectron. Reliab.

Int. J. Electr. Power Energy Syst.

Expert Syst. Appl.

Appl. Therm. Eng.

Neurocomputing

Energy

J. Hydrol.

Neurocomputing

Decis. Support. Syst.

Comput. Ind. Eng.

Chem. Eng. Res. Des.

Sustain. Energy Technol. Assessments

Nonlinear compact thermal model of the IGBT dedicated to SPICE

IEEE Trans. Power Electron.

In-situ health monitoring of IGBT modules of an on-line medium-voltage inverter system using industrial internet of things

CSEE J. Power Energy Syst.

Tunnel magnetoresistance-based short-circuit and over-current protection for IGBT module

IEEE Trans. Power Electron.

Remaining useful life estimation for thermally aged power insulated gate bipolar transistors based on a modified maximum likelihood estimator

Int. Trans. Electr. Energy Syst.

An industry-based survey of reliability in power electronic converters

IEEE Trans. Ind. Appl.

Investigation into IGBT dV/dt during turn-off and its temperature dependence

IEEE Trans. Power Electron.

Failure mechanism of die-attach solder joints in IGBT modules under pulse high-current power cycling

IEEE J. Emerg. Sel. Top. Power Electron.

Validation of lifetime prediction of IGBT modules based on linear damage accumulation by means of superimposed power cycling tests

IEEE Trans. Ind. Electron.

"In situ diagnostics and prognostics of solder fatigue in IGBT modules for electric vehicle drives," (in English)

IEEE Trans. Power Electron.

"Reliability analysis of the optimized Y-source inverter with clamping circuit," (in English)

Microelectron. Reliab.

"Electro-thermal-mechanical multiphysics coupling failure analysis based on improved IGBT dynamic model," (in English)

IEEE Access

"Reliability analysis and design of MMC based on Mission profile for the components degradation," (in English)

IEEE Access

Lifetime estimation of discrete IGBT devices based on Gaussian process

IEEE Trans. Ind. Appl.