The objective of the thesis is to develop and apply testing methodology for fracture initiation of actual components of a ship's hull and to provide engineering design data to be used by the industry in predicting the effects of extreme loads such as accidents at sea or earthquakes. Since cleavage is more dangerous than ductile fracture, attention was focused on conditions controlling cleavage fracture. A careful inspection of damaged components of ships and structures damaged in the 1995 Kobe earthquake showed that fracture often initiates in the zones of large bending strains. It was found from the literature that tensile strains resulting from bending alone are not sufficient to cause fracture. A certain amount of reversed strain is needed. Also, temperature is an important parameter of the process. A comprehensive experimental program was undertaken to study the effects of pre-strain, reversed strain, and temperature. A total of 39 bending specimens cut from actual ship plating were tested and on that basis a 3-D failure surface was constructed for the AH36 steel, suitable for practical applications. The failure surface was formed in terms of compressive pre-strain, temperature, and reverse tensile strain. As-rolled plates pre-strained by -0.43 at room temperature with a Charpy 20 J work transition of 6⁰ C, cleaved with very small reversed strain of 0.011 at low service temperatures of-40⁰ C. Machining the surface, as done with conventional test specimens, raised the reversed strain to cleavage to 0.33.