Abstract

Reliable and unceasing exploitation of power electronic converters plays a major part in every application. This PhD thesis comes up with new opened-up reliability assessment frameworks and demonstrates the feasibility of using multistate degraded system analysis and interval analysis as well for attaining much more accuracy in reliability evaluation. Considering self- and coupling degradation effects of the items and assessing reliability in the system level are important issues which are still lacking in the previous studies and the present thesis has made an effort to overcome these problems. The thesis tries to launch two distinct reliability assessment frameworks, namely interval reliability and Multistate degraded system reliability.Interval reliability is capable of introducing an interval useful lifetime for DC-DC power electronic converters in which the self and coupling degradations effects of items have been taken into account. Fundamentally, it is based on the conventional physics of failure reliability assessment. Based on this method, instead of obtaining an inaccurate reliability value, one can attain an interval for the reliability of a global system whose useful lifetime undoubtedly lays between the boundaries. This method is able to consider degradation dependencies and assess multi-component system level reliability but not redundant system. An attempt was made to enhance the conventional reliability framework to reach a systematic approach capable of estimating a reliability assessment considering self and mutual degradation effects and being able to evaluate system-level reliability including redundant system. Multistate degraded system reliability analysis is capable of estimating system-level reliability, while mission profile and physics of failure of the system’s items are taken into account. In addition, the self and mutual degradation effects of items on the operation of the global system have been considered. The multi-state degraded system reliability model exposed to multiple failure processes has been generalized. The operating condition of the global system is defined by a finite number of states. Not only can the proposed framework be employed in determining the reliability of the degraded systems in terms of multi-state functions, but also it can obtain the states of the systems by estimating the system state probabilities.As an application, a DC-DC power electronic system containing three critical items has been studied. In this case study, two power semiconductors, namely IGBT and diode, and a power capacitor have been considered as three degradation processes and their aging effects on the useful lifetime estimation of the power electronic system have been discussed. For having a sense about newly proposed reliability frameworks, a 3000W and 200/400 V conventional DC-DC converter for electric vehicle application exposed to WLTP driving cycle is considered.; La fiabilité et la sureté de fonctionnement de l'électronique de puissance jouent un rôle majeur aujourd'hui dans l'industrie. Cette thèse propose un nouveau cadre d'étude de la fiabilité à partir de deux méthodes afin d'améliorer la précision des études de fiabilité : une analyse utilisant la méthode des système dégradés multi état, et une analyse par intervalles. La considération des effets de dégradation propres et couplés de chaque composant est également une problématique qu'il reste aujourd'hui à étudier et qui est traitée dans ce travail.L'étude de la fiabilité par intervalles permet d'estimer des intervalles de durée de vie des convertisseurs DC/DC en prenant en compte les dégradations propres et mutuelles des composants. Fondamentalement, cette méthode est basée sur la compréhension des phénomènes physiques. A partir de cette méthode, on peut obtenir un intervalle pour la durée de vie d'un système complet. Cette méthode n'est toutefois pas adaptée aux systèmes redondants.De son côté, la fiabilité des systèmes dégradés multi-états est capable de proposer une étude au niveau système en prenant en compte le profil de mission et la physique des dégradations.Cette méthode a été généralisée dans cette thèse au procédés subissant de multiples dégradations. L'exemple d'un convertisseur DC/DC conprenant trois composants critiques a été étudié comme cas d'application. Deux composants semiconducteurs, diode et IGBT, et un condensateur sont considérés comme trois process de dégradation et l'effet de leur vieillissement sur la durée de vie du convertisseur est discutée. Le convertisseur étudié a pour puissance 3 kW (200/400V) et est dédié aux applications de type véhicule électrique. Il subit un profil de mission typique automobile (cycle WLTP).

Abstract

International audience; Reliability evaluation is a critical task in computing systems. From one side, the results must be accurate enough not to under-or over-estimate the overall system reliability (thus either resulting in a non-reliable system, or a system for which too expensive solutions have been adopted). On the other side, the time required for the analysis should be kept at the minimum. This paper presents some new advances in the reliability assessment of computing systems, by showing techniques targeting both hardware and software levels, and the combination of both.

Abstract

n important problem in the assessment of reliability benefits of transport projects is that link level improvements must be translated to network level, so that they can be economically valued based on users’ trips from origins to destinations. For intermodal transport, shipments follow a chain with more than one mode. Generally, this requires aggregation of travel time distributions that are not additive. We propose an approach that estimates the change in transport time reliability of an intermodal transport chain based on the changes for links of that chain. We demonstrate the framework of reliability assessment for a case study of network improvement for rail-truck intermodal transport in China. Also, we demonstrate the application in a cost-benefit analysis context with user valuations of transport reliabilities from the case at hand. The application leads to the result that projects for the renovation and expansion of the transshipment terminal perform better compared with project that improve rail haulage speed. Another finding is that the effect of reliability improvement projects can be super-additive at network level. In comparison with traditional methods, we conclude that the proposed method can better estimate transport time reliability benefits when the distribution of link travel times is highly skewed. Also, it opens new possibilities for further research for measuring correlated reliability measures within networks and for performing network resilience analysis.