The purpose of the Detailing EDM is to “fine-tune” a previously ameliorated design. First, in order to determine the best fillet radius for formability, forging simulations are conducted using commercial software. Once the fillet dimension is ascertained, the 3-D model is generated and maximum stresses determined for a trial force. Two different commercial programs were used to determine the three dimensional stresses. The only statistical quantities involve the loading (Gaussian distribution), the material “strength” in the Analytical Criteria for Failure (ACF), and possibly, the boundary conditions in the 3-D models. This paper considers the ACF to be the resistance to fatigue-fracture under complete reversal of loads at 5 × 108 cycles. The paper overviews three different methods of combining stochastic behaviour with FE analysis, and presents a methodology for using the interference-method with non-symmetrical distributions. The EDM are then presented for the three Product Criteria of forging — formability, Prime cost and 3-D reliability with respect to the selected ACF.

No abstract

This paper introduces the concept of a three-level hierarchy of economic-decision matrices (EDM) and discusses how finite element (FE) and stochastic modeling can be coupled in decisions on quality control. A disadvantage of the deterministic-engineering approach is that quality control cannot be properly accessed by a corresponding EDM. A five-step design-strategy is outlined and various suggestions associated with these steps are presented in the paper. This design-strategy is then quantified by the consideration of a hypothetical reverse-engineering problem involving a U-shaped bracket with cut-outs and holes. The paper presents the results of numerical experiments (sensitivity study) involving six preliminary variates (Yield Strength, Young’s Modulus, thicknesses, magnitude and phase of the forces, and displacement B.C). The final primary variates are assumed to be governed by a Gaussian distribution, and the methodology of applying the classical Z-statistic method to the FE results are presented.

No abstract