A procedure was developed to analyze layered elastic flexible pavement systems in terms of probability and reliability. A computer program RELIBSIA was prepared to carry out the computations. Rosenblueth method, instead of the conventional Taylor series expansion, us used to estimate the expected value and variance of the strains (dependent parameters) based on the input mean values of independent parameters, i.e., aircraft load, layer thicknesses, and material moduli. The relationships between the reliability level and the allowable strain repetition of the designed system which is established with results computed using RELIBISA provide a decision-making tool for engineers to design pavements at desired reliability level. The design can be optimized by selecting thicknesses of the bituminous concrete and the base layers so that the pavement is failed in fatigue cracking and subgrade failure at nearly the same traffic level and the same reliability level. The reliability-strain curves have steeper slopes with the bituminous concrete strain failure criterion than with the subgrade strain failure criterion, indicating that for flexible pavements designed using the Corps of Engineer's failure criteria, the design has a greater degree of uncertainty in preventing subgrade failure than fatigue cracking of the bituminous concrete surface corps. However, this may not be true in real cases because the bituminous concrete failure criteria are determined based on controlled laboratory test data which do not consider the uncertainties existing in the laboratory-to-field correlations.

The California Bearing Ratio (CBR) design method for flexible airfield pavements was analyzed using a probabilistic approach. The design parameters considered were the load P (or the equivalent single-wheel load), the sub-grade CBR, the tire contact area A, and the pavement total thickness t. The expected value and variance of the dependent variable performance factor, alpha (which is logarithmically related to the number of passes to failure) were estimated by using the Taylor series expansion and the Rosenblueth method. Differences in computed results between the two methods were found to be small, although the derivation of the expressions for Taylor series expansion was very complicated. A procedure was developed to estimate the reliability of the designed pavement system based on known variabilities of design parameters. Results of the reliability analysis indicate that prediction of pavement performance is most influenced by variations of pavement thickness t and is least influenced by variations of tire contact are A. The effects of variations of wheel load P and subgrade CBR are identical. The weighting factors for parameters t, CBR, P, and A, in general, are approximately 1, 0.34, 0.34, and 0.01, respectively (Table 3).