Estimation of the ageing of metallic layers in power semiconductor modules using the eddy current method and artificial neural networks

Tien A. Nguyen, Pierre Yves Joubert, Stéphane Lefebvre

Research output: Contribution to journalArticle

Abstract

In high power operations, the ageing of power semiconductor modules has been often observed by several failures due to high temperature cycling. The main failures may be metallization reconstruction, solder delaminations, bond wire lift-offs or bond wire heel crackings, conchoidal breaking of ceramics. The paper focuses on the non-contact monitoring of the ageing of the aluminum metallization top layer and of the solder bottom layer of a power die, using the eddy current method. The ageing is assumed to induce a decrease of these layers conductivity. The evaluation of both layers conductivity changes are estimated using artificial neural networks starting from eddy current data provided by finite element computations carried out in the case of several aged die configurations. The error of estimation is less than a few percent in the considered cases and it demonstrates the relevance of the eddy current method to monitor the ageing state of power modules. The proposed approach provides relevant results which will be validated on experimental data in future works.

Original languageEnglish
Pages (from-to)129-141
Number of pages13
JournalProgress In Electromagnetics Research M
Volume40
DOIs
Publication statusPublished - 1 Jan 2014
Externally publishedYes

    Fingerprint

ASJC Scopus subject areas

  • Electronic, Optical and Magnetic Materials
  • Condensed Matter Physics

Cite this