Fatigue Analysis of Welded Joints Using Local Reference Stress
MetadataShow full item record
Of the many parts of a machine, welded joints are the most susceptible to fatigue loading. The unusually complex geometry of a welded joint combined with the heating process during welding produces a high amount of stress that can result in fatigue cracks and failure. In welded joints, fatigue cracks usually emanate from critical locations such as a weld toe, where the stress is highly elevated. It is important to design welded structures based on the fatigue failure criterion in tandem with the usual design requirements. There is also a need to accurately assess the fatigue damage of welded joints for maintenance purposes because existing welds may contain defects or fatigue cracks. Therefore, accurate estimates of the fatigue life of welded joints are thus essential components of structural design and maintenance. The three most common methods of estimating fatigue life, and the ones selected for consideration in this work, are the nominal stress-life (S-N) method, the local strain-life (ε-N) method, and the Linear Elastic Fracture Mechanics (LEFM) method. Challenges arise from the fact that each method requires different and precise information concerning critical stresses that affect the fatigue life of welded joints. For each method, different stresses defined at the critical locations of a welded joint must also be obtained. The S-N method requires a determination of the nominal stress σn, the ε-N method requires an estimate of the peak stress σpeak, and the LEFM method requires the calculation of the stress distribution in the prospective crack plane σ(y). The complex geometry and loading that characterize welded joints make these critical stresses difficult to determine, so that variations in the definition of stress constitute the primary current source of inconsistencies in weldment fatigue analysis. The goal of the study presented in this thesis was to develop one universal weldment stress analysis method that can supply accurate and consistent stress information for all contemporary fatigue analysis methods. A determination of accurate and consistent stress data at weld-critical zones requires that the stresses in that area be reviewed and related to a reference stress that must be located at the most critical weld zone and must also be related to the actual stresses there. The work conducted for this thesis involved the development of this kind of reference stress, which has been termed “local reference stress.” A number of issues complicate the determination of stress in the most critical areas of a welded joint, including the weld toe or weld throat, such as residual stress generated by the heating process during welding and elevated stress due to an abrupt change in geometry. Reliance on the well-known nominal stress that is dependent only on the cross-sectional area and ignores other geometric factors associated with the weld is therefore insufficient. Peak stress is also difficult to determine even with the use of a three-dimensional (3D) finite element (FE) method because the magnitude of the stress on the surface of the weld toe can be captured only by a fine mesh. The through-thickness stress distribution in the critical cross-section of a weldment is even more challenging to establish using conventional FE packages because of the steep stress gradient associated with the prospective crack plane. The methodology presented in this thesis offers a convenient means of evaluating the critical stresses that affect the fatigue life of welded joints. The new method is based on an FE technique, which was used effectively for generating well-defined reference stress values at critical areas in a weld. The local reference stress on the weld toe line and the associated linear through-thickness stress distribution constituted the basis for determining the necessary stress data required as input for each of the three fatigue-analysis methods. The objective was to analyze an entire welded structure using a shell FE model that involves a relatively small number of elements. Subsequent post-processing of the local reference stress produced the required stress data, such as the peak stress or the through-thickness stress distribution. While the stress data are based on a simple shell FE model and the local reference stress concept, they yet were able to supply accurate and detailed stress information for any of the fatigue life estimation methods. The developed method is based on a unique reference stress related to the actual critical non-linear stress across the weld toe. Determining weldment fatigue life using the local reference stress and the associated stresses resulting from the post-processing enable the inclusion of the weld macro; micro-geometrical features; and other factors that affect the fatigue life of welded joints, such as the stress concentration and residual stress. The shell FE modelling method also has the advantage of simulating a full structure in a relatively short time, thus requiring substantially less computational time and resources than 3D FE models. The tests also revealed that, with the new method, the thickness of the shell element that simulates the weld is very important because it affects joint stiffness. For this reason, the thickness of the weld shell element should be equal to the thinner of the welded plates. The stress data obtained based on the shell FE local reference stress were validated against detailed 3D FE models. These data were then applied for the prediction of fatigue life using the strain-life and the LEFM methods. Five case studies were modelled and analyzed in order to compare the fatigue life predictions produced using the developed methodology against experimental fatigue life data. The proposed shell FE local reference stress proved able to simulate the stress fields in a variety of welded joints, such as T-joints (fillet joints), and square or circular tube on a plate. The shell FE local reference stress data were compared to the results obtained using detailed fine mesh 3D FE models. The accuracy of the stress data based on the shell FE models was under 15 % in all the joints except for the gusset, which was around 20 %. The difference between the peak stress resulting from the shell FE local reference stress and the peak stress resulting from the 3D FE modelling is within 10 % for all the case studies, with the exception of the first case, which was within 20 %. With respect to the number of elements, the difference between the shell and the 3D FE models was very large. The data produced by the proposed shell FE model matched that produced by the 3D fine mesh elements with respect to accuracy, but required nine to 204 times fewer elements. As a result, the proposed shell FE modelling method offers the advantage of simulating a full structure in a relatively short time and with fewer computational resources, as was particularly proven in the last case study involving a complex tubular joint. Such accurate stress data then enables fatigue life to be evaluated using the post-processed shell FE local reference stress data.
Cite this version of the work
Badr Bedairi (2017). Fatigue Analysis of Welded Joints Using Local Reference Stress. UWSpace. http://hdl.handle.net/10012/11949