Description: Many complex engineering problems which model when a structure will fail have significant amounts of uncertainty which are present in the results. This session illustrates the importance of understanding the physics of the failure mode in order to produce physically meaningful statistical results which quantify this uncertainty. Several papers in this session will illustrate this concept through a specific engineering case study.
Organizer: Eric P. Fox, Pratt & Whitney, P. O. Box 109600, M/S 731-82 W. Palm Beach, FL 33410
Chair: Don H. Ebbeler, Jet Propulsion Laboratory, California Institute of Technology 4800 Oak Grove Dr., M/S 125-224 Pasadena, CA 91109
Paper #1: "Uses of Statistical Models for Fatigue Crack Growth" Speaker: George P. McCabe, Purdue University Dept. of Statistics W. Lafayette, IN 47906-1399 Co-Authors: Ben M. Hillberry, Purdue University, School of Mech. Eng. Alten F. Grandt, Purdue University, School of Aeronautics and Astronautics Bruce A. Craig, Purdue University, Dept. of Statistics
Abstract: Variation in fatigue crack growth is an important phenomenon that plays a central role in the understanding of problems related to the failure of aircraft structures. Probabilistic models have been developed for describing this variation. Although these models are interesting in themselves, simulations based on these models have provided valuable engineering information. These models will be described, and examples of their use, including multiple site damage problems, will be discussed. It will be argued that stochastic models should be evaluated in terms of their potential uses.
Paper #2: "Evaluating Crack Detection Capability for Aircraft Risk" Speaker: Peter W. Hovey, Air Force Wright Aeronautical Laboratories 484 Twinning Dr. Dayton, OH 45431 Co-Author Alan P. Berens, University of Dayton Research Institute
Abstract: Fatigue crack growth is the dominant source of damage to aircraft structures and can lead to catastrophic failures. The strategies used to ensure safety include redundant load carrying structures and periodic maintenance. Nondescructive inspections (NDI) are used as part of the maintenance program in both military and civilian aviation to help manage the extent of fatigue crack growth in aircraft. Inspection intervals are based on predictions of fatigue crack growth under typical operational conditions. The capability of the NDI system provides the initial flaw sizes used in the fatigue crack growth predictions. Statistical methods for evaluating the capability of NDI to find cracks will be discussed with an emphasis on the influence of NDI capability on the risk of failure of the inspected structure. The history of NDI evaluation methods will be discussed, and the current US Air Force standard method for NDI of engines will be described.
Paper #3: "Stochastic Simulation of Liquid Propellant Aging" Speaker: Nicholas R. Moore, Jet Propulsion Laboratory, California Institute of Technology 4800 Oak Grove Drive, M/S 125-224 Pasadena, CA 91109 Co-Authors: Don H. Ebbeler, Jet Propulsion Laboratory, California Institute of Technology N. W. Ferraro, Jet Propulsion Laboratory, California Institute of Technology Laura E. Newlin, Jet Propulsion Laboratory, California Institute of Technology J. J. Blandino, Jet Propulsion Laboratory, California Institute of Technology
Abstract: A stochastic model of liquid propellant aging based on mechanics of the aging process has been developed. This model captures life limiting effects of elevated temperatures and iron leaching from metal surfaces. Stochastic simulation was used to reflect variability of data from tests performed to characterize aging mechanics. Probabilistic prediction of spontaneous rapid partial decomposition resulting from propellant aging is compared with observation.
Paper #4: "A Probabilistic Assessment of Fiber Failure During Creep in Continuous Fiber Reinforced Composites" Speaker: Eric P. Fox, Pratt & Whitney P. O. Box 109600, M/S 731-82 W. Palm Beach, FL 33410 Co-Author: David P. Walls, Pratt & Whitney
Abstract: This paper discusses a probabilistic approach to calculate the statistical variation in creep life of continuous fiber reinforced composites based on a structural model of creep behavior. Creep is the elongation of a material system under load at elevated temperature. Creep life is the time required to deform a structure so that it is no longer functional. The variation in creep life results from variation in fiber strengths. Several statistical methodologies are considered to determine the best strategy for estimating the distribution of creep life.
Paper #5: "Optimal Design of Neutron Lifetime Experiment" Speaker: Kevin Coakley, National Institute of Standards and Technology Statistical Engineering Division 325 Broadway Boulder, CO 80303 Co-Author: None
Abstract: An isolated neutron decays into a proton, electron, and neutrino with a mean lifetime of about 900 seconds. Physicists from Harvard and NIST are planning a high precision measurement of the mean lifetime of the neutron. There will be many cycles of a two stage experiment. In the fill stage of each cycle, a neutron beam is guided into a magnetic trap. After filling the trap to a chosen level, the beam is blocked. In the decay stage of each cycle, neutron decay and background events are recorded as a function of time. The event data is summarized as a histogram. The break points of the bins are selected so that the expected number of decay events per bin is approximately constant. Based on data pooled from all the cycles, the lifetime is estimated by a nonlinear fitting procedure. The optimal allocation of time, between the fill and decay stages is found by minimizing an asymptotically valid approximation for the lifetime estimate. Stationary and nonstationary background models are considered.