Effect of Processing on Inelastic Fracture Behavior of 6061 Aluminum

Statistical concepts from Design of Experiments were used to study the effect of thermal processing on the mechanical properties of 6061 aluminum. The soak temperature, cooling rate and soak temperature/cooling rate interaction term were found to be the most influential processing variables. Soak time and heating rate were statistically insignificant in the current experimental design, but modifications to the experimental method are discussed to study the influence of these factors. Micrographs of specimens subjected to two different processing schemes are presented and discussed. The integration of the results of the work into a general study of the ductile fracture behavior of aluminum alloys is discussed. INTRODUCTION Fracture in engineering materials, particularly metals, has been widely studied. The concept of Linear Elastic Fracture Mechanics (LEFM) is a powerful technique for understanding fracture behavior when the amount of plastic deformation prior to fracture is relatively small [1]. Tools such as the crack-tip opening displacement (CTOD) [1] and the J-integral [2] expand our analysis capabilities when significantly more plastic deformation is present prior to fracture. In the presence of massive plastic deformation and ductile fracture, however, even these approaches cannot measure geometry-independent properties [1]. This limitation can lead to severe design restraints or expensive experimental validation procedures in applications such as aerospace and automotive structures where high strength and stiffness with low weight are desired. As a result, a need exists to be able to understand and predict the ductile fracture behavior of metals. In aerospace and automotive applications, the metal of interest is typically aluminum. Ductile fracture in aluminum is associated with the nucleation, growth and coalescence of voids within the microstructure [1]. One approach that has been used to model this behavior follows an analysis after that of Gurson [3], who considered the growth of spherical voids in a homogenous matrix. Various corrections to the original Gurson model have been proposed to allow for instability and final fracture of the material [4,5]. One of the most promising models is the so-called ‘Complete Gurson Model’ of Zhang et al. which combines a yield function based on the void volume fraction with expressions for calculating void nucleation and growth as well as plastic instability in material between adjacent voids [6]. This model was implemented into the finite element code ABAQUS and showed good agreement with strength measurements in two weld geometries [7]. Zhang et al. have also developed a method to link some of the model parameters from the Complete Gurson Model to the processing history of the material using the program WELDSIM and a specialized mapping algorithm [7]. Results of the study of Zhang et al. showed that strength predictions from the Complete Gurson Model were in good agreement with experimental results if slight adjustments were made to the void nucleation parameters for each test geometry [7]. A direct, geometry-independent link between the void nucleation parameters and the processing history is not yet available [7]. To explore the link between strength and microstructure, Myhr et al. have constructed a thermodynamic framework to predict strengthening and hardness based on particle evolution [8]. While the results are encouraging, a fully coupled, predictive processing/fracture model does not yet exist. As a result, the full potential of numerical modeling techniques to reduce the required number of experiments for validation of design performance cannot be completely realized. The current work is part of a larger project to provide a statistical link between material processing and ductile fracture behavior of aluminum alloys. Fracture experiments, microstructural analyses and numerical calculations will be combined to allow the fracture behavior of aluminum alloys to be predicted based solely on knowledge of their thermal and mechanical processing history. The final goal of the overall project is to link every parameter in the Complete Gurson Model to measurable processing variables. This paper presents experimental results and statistical analyses from tests on 6061 aluminum samples that were subjected to a variety of heating and cooling sequences. EXPERIMENTAL METHOD Standard dogbone tensile specimens were machined from 6061-T6 sheet stock (0.1875 in. thickness) in accordance with ASTM Standard E8 [9]. To remove any effects of work hardening on the results, these specimens were annealed at 775 for o F two hours. After two hours, the specimens were cooled at a rate of 50 o F /hour until 500 o F and then the oven was turned off and the specimens were allowed to cool room temperature. It is expected that this treatment will have removed the effects of the original T6 condition [10]. Each sample was then subjected to a heat treatment following the schematic of Figure 1. Figure 1: Heat treatment history of the experimental samples. Statistics based on Design of Experiments (DOE) concepts [11] were used to assign random values of the processing parameters shown in Figure 1 to each of the experimental specimens. The heating rate, , was varied by placing the specimens in a preheated furnace (+) or a room temperature furnace (-), the soak temperature (T 1 T& H) was either 788 (-) or 1022 (+), the holding time (t) was either 60 minutes (-) or 2.5 hours (+) and the cooling rate, , was either furnace cooled (-) or water quenched (+) for each specimen. For a four variable system, a full factorial setup consists of 2 o F o F 2 T& 4 = 16 test specimens. An appropriate randomized test matrix can be found in the reference [11]. Immediately following heat treatment, each specimen was placed in a refrigerator at 38 . After all specimens were heat-treated, all specimens were removed from the refrigerator and allowed to age at room temperature for four days and four hours. The samples were then returned to the refrigerator for approximately two weeks until testing. o F Each sample was tested under displacement control in a hydraulic testing machine (Tinius-Olsen, Super “L” Universal Testing Machine). Load and strain information were provided by the machine load cell and an attached extensometer, respectively. Additional strain data was captured using a 2D digital image correlation system (Correlated Solutions). Good agreement was found between both sources of strain data. Following testing the specimens were placed back into the refrigerator for storage prior to microstructural examination. From the testing data, a true stress-strain curve for each specimen was constructed. Yield stress, fracture stress, fracture strain, and the area under the true stress-strain curve were all calculated. Finally, the plastic portion of the stress-strain curve was fitted using the relationship [12]