Some observations on Sih's strain energy density approach for fracture prediction