Multistate analysis and design : case studies in aerospace design and long endurance systems

Abstract

This research contributes to the field of aerospace engineering by proposing and demonstrating an integrated process for the early-stage, multistate design of aerospace systems. The process takes into early consideration the many partially degraded states that real-world systems experience throughout their operation. Despite advancing efforts aimed at maintaining operation in a state of optimum performance, most systems spend very substantial amounts of time operating in degraded or off-nominal states (e.g. Hubble space telescope, Mars Spirit rover, or aircraft flying under minimum-equipment-list restrictions). There exist relatively few methods and tools to address this at the beginning of the design process. At one end of the spectrum is design optimization, but this typically concentrates on the system in its nominal state of operation, only infrequently considering failure states through piecemeal application of constraints. There is reliability analysis, which focuses on component failure rates and the benefits of redundancy but does not consider how well or poorly the system performs with partial failures. Finally, there is controls theory, where control laws are optimized but the plant is typically assumed to be given a priori. The methodology described within this thesis coordinates elements from each of these three areas into an effective integrated framework. It allows the designer deeper insight into the complex problem of designing cost effective systems that must operate for long durations with little or expensive opportunity for repair or intervention. Specific contributions include: 1) the above methodology, which evaluates responses in system expected performance and availability to changes in static design variables (geometry) and component failure rates, accounting for control design variables (gains) where appropriate, 2) the demonstration of the cost and benefits associated with a multistate design approach as compared to reliability analysis and the nominal design approach, and 3) a multilayer extension of Markov analysis, for translating single sortie vehicle level metrics into measures of multistate campaign performance. The process is demonstrated through three application case studies. The first of these establishes the feasibility of the approach through the multistate analysis of performance for an existing twin-engine aircraft. This analysis was enabled through the development of a multidisciplinary simulation based design model for evaluation of multistate aircraft performance. A medium-altitude long endurance unmanned aerial vehicle is designed in the second case study, first from a single-sortie, ultra long endurance perspective and then from a multiple sortie, mission campaign perspective. Finally, the third case study demonstrates applicability of the approach to a lower level subsystem, that of the lubrication system for a geared turbofan engine. Several major findings result from these case studies, including that: 1) multistate performance output spaces have distinctly unique shapes and boundaries, depending on whether formed through variation of component failure rates, static design variables (geometry), or a multistate combination of both, 2) a region of multistate performance results from the combined variation of failure rates and static design variables that is unachievable through the independent variation of either one, 3) small changes in static design variables may be used to significantly improve system availability, and 4) the general multistate design problem is one of competing objectives between system availability, expected performance, nominal performance, and cost.