Keywords: robust engineering, experiment design, performance variations, noise factors, IGBT modules, multilayer structure, shearing constraint.

Product design is often limited to find characteristics according to the specifications and then to optimize the solution. In fact, system performance is often perturbed by external parameters not taken into account during the design study such as environmental variations or fabrication variability for example [1].

A robust design goal is to optimize system performances minimizing at the same time noise factor sensitivity. Results of a good robust design are measured at the reliability, maintainability and availability level.

In our case we apply robust design principles on a transistor used in a traction application named the IGBT (Insulated Gate Bipolar Transistor).

The perception of system quality by the user is close due to the noise factor of sensitivity of this system. So it is necessary to reduce the noise factor impact on the system performances. There are two ways to do that:

• eliminate noise sources.
• eliminate system sensitivity to this noise.

It can be very expensive and very time consuming to eliminate the noise factors themselves because of the uncontrolled nature of these factors and some of them are too difficult and too expensive to control:

So we can give this definition of robust engineering:

A product of a process is robust if it has no sensitivity to the effect of a variability source even if these variability sources have not been eliminated [2].

The objective of the R&D team is to develop a product or a process that works as the specification and in different condition during all its lifetime. Robust engineering design is a method allowing that enables a design where the performance is not affected or affected as little as possible by the noise factors.

For Taguchi [2], robustness is the essential component of quality. So it is natural that the tools employed in robust design are "quality" tools like the design of an experiment.

The purpose of the design of an experiment is to organise experiments intelligently in order to obtain the maximum of information using a minimum of arrays [3].

The design of the experiment method is employed to answer two kinds of problems:

• screening problems: the purpose is to identify influent parameters, comparing factors impact on system performance; and
• the response surface method (RSM): the purpose of this method is to model the system performance in order to optimize it.

High power IGBT modules used in traction applications are complex multilayered structures consisting of different materials, which have to provide a good mechanical stability, good electrical insulation properties, and good thermal conduction capabilities [4]. During the life cycle, the modules suffer thermal cycles. The main problem in this multilayer structure is the difference between the coefficients of thermal expansion (CTE) between the associated materials. These differences cause important shearing constrains at interfaces which provoke first the lift off of the interface and then, the destruction of the module.

1

URL

2

W.Y. Fowlkes, C.M. Creveling, "L'ingénierie robuste - utiliser la démarche Taguchi pour concevoir des produits et des systèmes robustes", Édition DUNOD, 1998

3

M. Pillet, "Les plans d'expériences par la méthode Taguchi", Les éditions d'organisation, 1997.

4

M. Ciappa, "Selected failure mechanisms of modern power modules", Microelectronics Reliability 42, p. 653-667, 2002. doi:10.1016/S0026-2714(02)00042-2

purchase the full-text of this paper (price £20)

go to the previous paper
go to the next paper
return to the table of contents
return to the book description
purchase this book (price £140 +P&P)