It is observed clinically that some years after MOD restoration of molar and premolar teeth, cuspal failures commonly occur. The period between restoration and failure is usually from eight to fifteen years. Evidence has been found, both from observations of fractured cusps and from a mathematical model of an idealized molar tooth, that allows a clear description of the mechanism of these slow failures. Suggested changes in clinical practice are recommended if such failures are to be prevented in the future.

A nonlinear finite element method (FEM) model has been developed to model thin-surfaced, unbound granular pavements. The FEM model incorporates a nonlinear anisotropic material model for the granular material and a nonlinear material model for the subgrade. The coefficients for the material models were determined from laboratory repeated load triaxial tests. A test pavement at the Canterbury Accelerated Pavement Testing Indoor Facility (CAPTIF) was instrumented with an inductive coil soil strain system to measure vertical compressive strains and pressure cells to measure the vertical compressive stresses. The test pavement was subjected to loading by a falling weight deflectometer (FWD) device at four different load levels. The initial FEM model configuration was undertaken at one load level using the surface deflection measurements from the FWD; however, there were significant differences between the computed and measured stress and strain values. The scalar coefficients for the granular and subgrade materials were adjusted so that the measured and computed stress and strain values matched within a reasonable tolerance (10%). After the calibration process was finished, it was found that the computed and measured surface deflections also matched. The model was rerun with the three remaining load cases; the computed and measured stress, strain, and surface deflection values were in close agreement. The range of computed stiffness values varied with both the pavement and load level. The results from this analysis showed the need to use a full nonlinear model to obtain realistic results when FWD data are modeled.