Show simple item record

A Study of Software Input Failure Propagation Mechanisms

dc.contributor.advisorSmidts, Carolen_US
dc.contributor.authorwei, yuanen_US
dc.description.abstractProbabilistic Risk Assessment (PRA) is a well-established technique to assess the probability of failure or success of a system. Classical PRA does not consider the contributions of software to risk. Dr. B. Li and C. Smidts have established a framework to integrate software into PRA which recognizes the existence of four classes of risk contributors: functional, input, output and support failures. Input/Output failures have been shown to make up 57.4 % of the failures experienced during software development of major aerospace systems and have been at the origin of a number of major accidents such as the Mars Polar Lander. This research quantifies the contribution of the input failures. More specifically, this dissertation 1) defines the concept of input failure, 2) studies the related propagation mechanisms, 2) estimates the propagation probability for different types of input failures, and 3) applies the fault propagation analysis to the framework of integrating software into PRA. The dissertation defines the concept of artifact as a reference point to identify expected inputs and consequently input failures (inputs which differ from the expected ones). Input failures are divided into value-related failures (including value, range, type and amount failures) and time-related failures (including time, rate and duration failures). Value failures are examined first. The concept of masking areas and flat parts is defined, and the dissertation proposes an Image Reconstruction Method (IRM) to estimate the propagation probability of input value failures. This method is proven to require less number of test cases than one that could be based on random testing to reach the same relative error. For the other input failure modes, the dissertation reveals how they transform to the data state error and formalizes their propagation criteria so that the IRM can be applied to estimate the propagation probability. The contributions are thus: 1. Clear definition of the concept of input failure; 2. Definition of a systematic process of identification and quantification of the contributions of input failures to risk; 3. Systematic analysis of the propagation mechanisms of each type of input failures.en_US
dc.format.extent1385727 bytes
dc.titleA Study of Software Input Failure Propagation Mechanismsen_US
dc.contributor.publisherDigital Repository at the University of Marylanden_US
dc.contributor.publisherUniversity of Maryland (College Park, Md.)en_US
dc.contributor.departmentReliability Engineeringen_US
dc.subject.pqcontrolledEngineering, Mechanicalen_US

Files in this item


This item appears in the following Collection(s)

Show simple item record