Visualization Techniques for Simulation-Based Dependent Failure Analysis

Abstract

<div class="section abstract"><div class="htmlview paragraph">Failure analysis of engineering systems typically emphasizes identification and mitigation under an independent failure assumption with dependent failures treated as the exception rather than the rule. Some frameworks for addressing dependent failures through analysis appear in standards including NUREG 0492, ISO 26262, MIL-1629-A, and ARP4761 amongst others. The purpose of identifying these dependencies is to allow system analysts to determine and quantify the factors that influence dependent fault probabilities. Once identified, failure relationships can be incorporated into a Discrete Event Simulation (DES) of the system, providing a mathematically rigorous estimate of system utility (e.g., availability, reliability). The output of a simulation can provide an expected value of performance but additionally, can also allow the analyst to identify the downstream impact of probabilistic dependencies between system elements. Detailed simulations of complex systems have the potential to generate rich data sets, which once analyzed may aid in the formulation of corrective actions to mitigate any detrimental effects identified from simulated failure dependencies. However, human factors research has demonstrated that effective decision making requires that only the key information is provided to the user, which in turn should be supported through effective data visualization techniques. In this paper, case studies are conducted using automated failure dependency mappings generated using a model-based RAMS software and then simulated using a DES framework. Key mappings and transformations of temporal data from the simulation are then presented along with various supporting visualization techniques that together aim to elevate the situational awareness of analysts when performing complex system evaluation.</div></div>