Understanding the ageing of materials for electronic application is a complex challenge, in particular for composite material as electrically conductive adhesive consisting of a nonconductive polymer binder and conductive filler particles. This research involves a large amount of parameters related to both operating conditions and material structure, which act together. Accelerated testing, with the aim to consume rapidly lifetime without inducing incorrect failure mechanisms, is difficult to optimize and its modelling to describe the ageing process is a challenge. To reach this aim could be interesting for the characterization of the material but above all for the general validity of the proposed methodology.

In the building domain, components or equipment are often subjected to severeenvironmental conditions. In order to predict the reliability and the life-time of suchequipment, accelerated life testing can be carried out. Severe conditions are applied toaccelerate the ageing of the components and the reliability at nominal conditions is thendeduced considering that these nominal conditions are not constant but stochastic. In thispaper, the accelerated life testing of photovoltaic modules is carried out at severemodule temperature levels. The module power losses are monitored and the limit state isdetermined when a threshold power is reached. The stochastic data and the reliability aresimulated during a period of twenty years. Finally, the life time of the component isevaluated.

Nous proposons dans cet article une méthode d'essai accéléré applicable pour des systèmes mécaniques complexes dont la durée de vie ne peut être exprimée par une relation analytique du niveau de sollicitation. Dans de tels cas, une approche numérique peut rester envisageable. Il s'agit alors d'ajuster le modèle numérique à partir de tests sévérisés au cours desquels les défaillances apparaissent plus rapidement qu'en conditions nominales. Il est ainsi possible de prévoir à partir d'essais très courts des défaillances susceptibles d'apparaître après une longue période d'utilisation. Cette méthode est mise en œuvre dans le cas de sollicitations en fatigue, l'endommagement cumulé conduisant à la rupture ne pouvant être modélisé que par un code numérique en éléments-finis.