Comparative evaluation of reliability assessment methods of power modules in motor drive inverter

https://doi.org/10.1016/j.microrel.2020.113730Get rights and content

Abstract

The complete percentile lifetime models are essentially required for a mission profile based reliability evaluation of power modules. However, the information on the complete lifetime models or lifetime data of power modules is rarely fully provided by manufacturers. The Monte Carlo method is a good solution to obtain the time-dependent reliability function when only limited lifetime information such as a certain Bx lifetime is given. However, there is a lack of research on the accuracy of the reliability function obtained by the Monte Carlo analysis. In this paper, a time-dependent reliability function of the IGBT module which is obtained by Monte Carlo method from a certain Bx lifetime is compared with that estimated by the complete percentile lifetime model in the case study of motor drive inverter. The complete percentile lifetime model is developed by the use of power cycling test results of 30 IGBT modules.

Introduction

A mission profile based reliability evaluation is the state-of-the-art reliability assessment method of power modules, where the temperature stress is one of the main causes of the failure [1]. For the mission profile based reliability evaluation, a lifetime model of the target power module regarding the temperature stress is essentially required.

To build the lifetime model, the power cycling tests under different temperature stress conditions are performed [2]. Typically, the lifetime model is built considering three main temperature stress factors, which are the junction temperature swing (△Tj), mean junction temperature (Tjm) and heating time (ton) [2,3]. Several tests are required for the statistical analysis of power cycling test results, which is essential in reliability assessments. At least six tests under each test condition are recommended [4]. Finally, the percentile lifetime models are developed based on the test results and acceleration models [5].

It is worth mentioning that the percentile lifetime is a time by which a certain percentage of the items might have failed [4]. For example, B10 lifetime means the time by which 10% of the power modules have failed or the reliability is 0.9.

In general, the information on complete lifetime data of the power modules is rarely fully provided by manufacturers. Even though a lifetime model is given, that lifetime is a certain percentile lifetime model or a lifetime model without statistical analysis as provided in [6,7]. Therefore, the time-dependent reliability function is not available, which is important for the both component-level and system-level reliability evaluations.

To overcome that, in [8], the Monte Carlo method has been applied by considering the variations in the lifetime model and parameters of the power module. From this approach, the time-dependent reliability function of the power module can be obtained. However, there is a lack of research in the prior-art work on the accuracy of the reliability function obtained by the Monte Carlo method compared with the reliability function based on complete percentile lifetime models.

In this paper, the reliability assessment methods are comparatively evaluated. The lifetime model is developed based on the power cycling tests of 30 IGBT modules. Then, the time-dependent reliability function of the IGBT in motor drive system is obtained by performing the mission profile based lifetime analysis with complete percentile lifetime models. The time-dependent reliability function of the IGBT is also built by the Monte Carlo method based on B10 lifetime model with and without lifetime model information. Finally, the reliability estimation results are comparatively evaluated.

Section snippets

Power cycling test setup and target IGBT module

Fig. 1 shows the power cycling test system used for the lifetime model development. This system consists of 6 individual power cycling test systems so that 6 tests can be performed at the same time for reducing the test time. This test system enables to carry out the power cycling test with realistic operating conditions of power converters as similar as possible. Furthermore, the real-time condition monitoring of the tested power modules is possible. More detailed information on it can be

Power cycling test results

The 5% increase of VCE_ON is considered as its end-of-life criterion during the power cycling test. In order to investigate the effect of the Tjm on the lifetime of the IGBT module, the power cycling test under the Conditions 1, 2 and 3 are performed (See Table 1), where the △Tj is fixed to 60 °C but the Tjm are 70 °C, 80 °C and 100 °C, respectively. In the case of Condition 1, the failure in the first IGBT module occurs in TLU at 1003717 cycles as it is shown in the Fig. 3.

It can also be

Reliability evaluation based on lifetime model

Fig. 6 shows a configuration of 3-phase motor drive system with Permanent Magnet Synchronous Motor (PMSM) for the case study. It consists of a single-phase rectifier and boost power factor correction circuit on the grid side and a 2-level inverter on the machine side. The related parameters are listed in Table 3. The reliability assessments are carried out with the simplified mission profiles of the motor drive system composed of speed and torque mission profiles as shown in Fig. 7.

The lower

Conclusion

In this paper, a mission profile based reliability assessment methods of the power devices have been studied. The time-dependent reliability functions of the IGBT in the motor drive system obtained by two reliability assessment methods are comparatively evaluated. Furthermore, the effect of the lifetime model information on the accuracy of the reliability function, obtained by the Monte Carlo method has been investigated. It can be concluded that using the complete lifetime model to obtain the

Declaration of competing interest

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

References (10)

There are more references available in the full text version of this article.

Cited by (8)

View all citing articles on Scopus
View full text
© 2020 Elsevier Ltd. All rights reserved.