A statistical approach involving the uniform design of experiments and regression analysis is used to investigate the effects of thermal bonding process parameters, dwell time, thermal bonding temperature, and hot air velocity, on the pore size of three-dimensional (3D) nonwoven filters. These filters are produced from polypropylene (PP)/polyester (PET) (sheath/core) bicomponent staple fibers. The pore structures of the filter samples were examined using the capillary flow porometer. Results reveal that the statistical approach is effective in identifying the effects of the investigated process parameters on both the bubble point pore diameter and the mean flow pore diameter for the thermally bonded nonwoven filter samples. Under the optimized processing parameters for achieving the minimized pore size, the predicted minimum bubble point pore diameter is 111.12 mu m and the predicted minimum mean flow pore diameter is 63.4 mu m for the filter sample with the area density of 60 g/m(2). They are in good agreement with the experimental values of 111.71 and 60.91 mu m, respectively. Microstructure features observed using scanning electron microscope indicate that the effects of the investigated process parameters on the pore size are closely related to the thermal energy delivered to the fibers and the pressure drop acting on the fabrics. (c) 2006 Wiley Periodicals, Inc.