Large-eddy simulation (LES) has been used to predict the statistically three-dimensional turbulent boundary layer (3DTBL) over a rotating disk. LES predictions for six parameter cases were compared to the experimental measurements of Littell & Eaton (1994), obtained at a momentum thickness Reynolds number of 2660. A signal-decomposition scheme was developed by modifying the method of Spalart (1988) to prescribe time-dependent boundary conditions along the radial direction, entrainment towards the disk surface was prescribed by satisfying global mass conservation. Predictions of the mean velocities and r.m.s. fluctuations are in good agreement with data, with the largest discrepancy occurring in the prediction of the wall-normal intensities. The primary and two secondary shear stresses are also in good agreement with the measurements and one-dimensional energy spectra of the velocity fluctuations agree well with established laws, i.e. a −1 slope in the buffer region and −5/3 slope near the edge of the boundary layer.

Conditionally averaged velocities provide new evidence in support of the structural model of Littell & Eaton (1994) concerning the interaction of mean-flow three-dimensionality and shear-stress producing structures. Inside the buffer region under strong ejections, the conditionally averaged crossflow (radial) velocity is larger than the unconditioned mean, and the profile conditioned on strong sweeps is smaller than the mean. This is consistent with the notion that streamwise vortices having the same sign as the mean streamwise vorticity, and beneath the peak crossflow location, are mostly responsible for strong sweep events; streamwise vortices with opposite sign as the mean streamwise vorticity promote strong ejections. Comparison of two-point spatial correlations with previous measurements in two-dimensional turbulent boundary layers (2DTBLs) indicates interesting structural similarities, e.g. the correlation of wall pressure and surface-normal velocity fluctuations is an odd function of streamwise separation, being positive downstream and negative upstream. These similarities offer quantitative indirect support to the hypothesis advanced by Littell & Eaton (1994) and Johnston & Flack (1996) that structural models describing 2DTBLs may be employed as a baseline in (equilibrium) 3DTBL structural studies.