Traditional structured and unstructured grid generation methods need a ‘water-tight’ boundary surface grid to start. Therefore, these methods are named boundary to interior (B2I) approaches. Although these methods have achieved great success in fluid flow simulations, the grid generation process can still be very time consuming if ‘non-water-tight’ geometries are given. Significant user time can be taken to repair or clean a ‘dirty’ geometry with cracks, overlaps or invalid manifolds before grid generation can take place. In this paper, we advocate a different approach in grid generation, namely the interior to boundary (I2B) approach. With an I2B approach, the computational grid is first generated inside the computational domain. Then this grid is intelligently ‘connected’ to the boundary, and the boundary grid is a result of this ‘connection’. A significant advantage of the I2B approach is that ‘dirty’ geometries can be handled without cleaning or repairing, dramatically reducing grid generation time. An I2B adaptive Cartesian grid generation method is developed in this paper to handle ‘dirty’ geometries without geometry repair. Comparing with a B2I approach, the grid generation time with the I2B approach for a complex automotive engine can be reduced by three orders of magnitude.