Computational Models to Predict the Structural Reliability of Aerospace Systems

Abstract

Three case studies were presented in which computational-based methodologies were used to assess structural reliability in the aerospace industry. The studies involved hot-section turbine disks of a helicopter engine, fan blades of a commercial airline engine, and bearings in an auxiliary power unit. In all cases, the results of the computational models were used to support the certification process for design and application changes. The statistical variation in design and usage parameters, including geometry, materials, speed, temperature, and other environmental factors, were considered. The response surface approach was used to construct a durability performance function. This performance function was used with the first-order reliability method to determine the probability of failure and the sensitivity of the failure to the design and usage parameters. A hybrid combination of perturbation analysis and Monte Carlo simulation was used to incorporate time-dependent random variables. System reliability methods were used to determine the system probability of failure and the sensitivity of the system durability to the design and usage parameters.The turbine-wheel case study presented the computational model used to simulate the wheel low-cycle fatigue lifetime statistical distribution. The wheel was experiencing field failure caused by fatigue crack initiation in the wheel rim and progressing to the bore. Computational models for gas dynamics, heat transfer, structural stress and fatigue, and fracture were used with multidisciplinary system reliability methods to create a probabilistic simulation of the wheel and predict the probability of failure as a function of flight cycles for a fleet of wheels. The model was calibrated with field data, and the rare event that caused failure was determined.The fan-blade case study involved the certification of a repair process. The titanium-blade leading edges erode over time, causing a reduction in engine efficiency. The repair process involved removing the eroded leading edge and welding a new leading edge. The repair certification required that several dozen repaired blades be tested in vibration for millions of cycles at high vibratory loads with no failures. The testing was expensive because only a few test facilities had this capability, and the tests took several months, so a computation high-cycle fatigue model was developed to predict the blade safety and determine the probability that all of the blades would pass the test. Computational probabilistic material fatigue models were created for the weld gradient of several material microstructural zones, including the original blade parent material, the heat-effected zones, the weld, and the parent material of the new leading edge. The gradient microstructure was overlaid on the finite element model and combined with the material model to create a computation vibratory fatigue model of the repaired blade. A similar model was created with the original parent material throughout to simulate a new (unrepaired) blade. A virtual design of experiment was performed in which a thousand blades were tested at many different vibratory stress levels and many different vibration modes for both the repaired and new blade. It was shown that the repair did not adversely affect the blade fatigue lifetime, with significantly fewer tests than the tradition certification method.A major commercial airliner began experiencing premature failures of turbine-driven auxiliary power units (APU) on one of their aircraft fleets. Three to four units were failing per year. It was determined that the failures were caused by the rupture of the main bearings. Proposed changes, such as operating procedures or bearing lubrication, would normally require ground testing of APUs. These are prohibitively expensive tests. Therefore, probabilistic computational modeling and simulation was used to evaluate proposed options for reducing the component failure. Computational models consisting of structural finite element analysis, lubrication analysis, and microstructural material fatigue analysis were created and used to assess multiple operating conditions, design and material processing variations, and lubricant types. The simulations determined that an operational change in combination with a lubricant change would eliminate the premature bearing failures. Probability charts, such as Fig. 1, were computationally simulated. The model results along with other supporting data were presented to the Federal Aviation Administration, which approved a change in the operating procedures and a different lubricant. Since the change, the airline has had no failures and has documented savings of $3–4 million per year.