NeT bead-on-plate round robin: Comparison of residual stress predictions and measurements

https://doi.org/10.1016/j.ijpvp.2008.11.017Get rights and content

Abstract

The members of the European network NeT have undertaken parallel round robin activities measuring and simulating the residual stresses generated by laying a single Tungsten Inert Gas (TIG) weld bead on an AISI Type 316L austenitic stainless steel flat plate. This is a strongly three-dimensional configuration with many of the characteristics of a repair weld. The round robin finite element predictions of weld residual stresses are compared with each other in order to identify the effects on the predicted residual stresses of material hardening model, global heat input, mechanical and thermal boundary conditions, and the handling of high temperature inelastic strains. Comparison with the residual stress measurements then leads to the optimum choices for these variables.

Introduction

NeT Task Group 1 (TG1) was formed in May 2002 to examine the benchmark problem of a single weld bead laid down on the top surface of an austenitic stainless steel plate. This weld geometry produces a strongly three-dimensional residual stress distribution, with similar characteristics to a weld repair, in a compact, portable specimen which is amenable to residual stress measurement using diverse methods. The single weld pass is relatively straightforward to model using finite element methods, and allows full moving heat source residual stress simulations to be performed in practical time scales. However, because the weld bead is laid onto a relatively thin plate, the predicted stresses are very sensitive to key modelling assumptions such as total heat input and the thermal transient time history, making the bead on plate a challenging benchmark problem. TG1 have organised and performed two parallel round robin activities centred around the single bead-on-plate specimen, covering residual stress measurement and prediction.

Four nominally identical bead-on-plate specimens (see Fig. 1) were manufactured under closely controlled and documented conditions, and appropriate materials characterisation testing was performed on both plate and weld material. Two protocol documents were written to define and control the experimental residual stress measurement and finite element residual stress prediction round robins, respectively. These are summarised elsewhere [1]. The specimens were then circulated to TG1 measurement round robin participants to perform residual stress measurements. In parallel, the finite element analysis protocol was circulated to TG1 analysis round robin participants to allow them independently to predict the transient welding temperature history and the weld residual stresses in advance of measurements being available. The results from this initial round robin are referred to as Phase 1 results. Predicted and measured temperatures and stresses were compared in detail in December 2004.

After completion of the phase 1 activities, two of the bead-on-plate specimens were destructively examined to confirm the weld fusion boundary geometry, and a number of parametric studies performed to confirm the importance of selected finite element solution variables. The understanding gained from the phase 1 round robin and subsequent parametric studies was then used to update the finite element analysis protocol and initiate a phase 2 analysis round robin, which is still underway.

This paper considers the phase 1 residual stress predictions made by Task Group 1, and has the following objectives:

  • 1.

    To assemble together the results of the Phase 1 mechanical finite element analyses and compare predicted residual stresses with measurements.

  • 2.

    To understand the influence of key thermal analysis variables on predicted residual stresses.

  • 3.

    To understand the cyclic thermo-mechanical loading history in the bead-on-plate specimens.

  • 4.

    To understand the influence of the assumed material constitutive behaviour on predicted residual stresses.

It makes use both of the phase 1 analyses submitted by December 2004, and of the parametric studies performed up to May 2005. Any analyses performed after that date are deemed to be phase 2 analyses, and are not considered here. A companion paper in this special issue describes and reviews the phase 1 transient temperature predictions [2], and should be read first.

Section snippets

Mechanical modelling assumptions

The participants in the phase 1 residual stress prediction round robin are listed in Table 1. The most important analysis variables are summarised in Table 2 for the main study and Table 3 for the parametric studies.

All the participants took advantage of symmetry to analyse half models, with a mesh size varying from 4850 to 25,522 elements, using both eight and twenty node solid elements. Three analysis codes were used, ABAQUS [3], ANSYS [4], and SYSWELD 2004 [5]. All but one analysis made use

Overview of predicted residual stress field

Fig. 2 presents typical contour plots of longitudinal and transverse stresses viewed from the top of the plate, with a cross-section along the centre of the weld bead.

  • High tensile transverse stresses are predicted beneath the weld bead, peaking beneath the weld stop end, and also on the top surface of the plate alongside the weld bead, again peaking at the weld stop end. The tensile stresses are balanced by transverse compressive stresses beyond the ends of the bead.

  • High tensile longitudinal

Discussion

The response on the mid-length through-wall line B–D provides the best insight into the impact of mechanical solution variables on the predicted stresses. Here, weld metal undergoes a simple tensile cool-down cycle, and the final stress is determined by the level of equivalent plastic strain and the assumed stress–strain behaviour. The choice of isotropic or kinematic hardening is not relevant. Transverse stresses in weld metal are generally not sensitive to the weld metal material properties,

Detailed conclusions

  • 1.

    The use of isotropic hardening leads to over-conservative predictions of stress in the single bead-on-plate specimens, particularly in the longitudinal direction.

  • 2.

    The use of non-linear kinematic hardening leads to the most accurate predictions of stress.

  • 3.

    Correlation between the “best” non-linear kinematic finite element prediction and the mean of the measurements is reasonable, but there is much room for improvement. No participants used mixed isotropic-kinematic hardening schemes, so their

Acknowledgements

We acknowledge the contributions of all members of NeT TG1, in particular the simulation round robin participants listed in Table 1. The specimens were manufactured and welded by Doosan Babcock. The support of British Energy Generation Ltd for the comparative analysis is also gratefully acknowledged. This paper is published by permission of British Energy Generation Ltd.

References (26)

  • Truman CE, Smith MC. Editorial,...
  • Smith MC, Smith AC. NeT bead on plate round robin: comparison of transient thermal predictions and measurements,...
  • ABAQUS/standard user's manual version 6.4

    (2004)
  • Ansys. ANSYS Inc, Canonsburg,...
  • SYSWELD v2004, ESI...
  • J. Lemaitre et al.

    Mechanics of solid materials

    (1990)
  • Ficquet X, Smith DJ, Truman CE, Kingston EJ, Dennis RJ. Measurement and prediction of residual stress in a...
  • El-Ahmar W, Jullien J, Gilles P. Robustness analyses of numerical simulation of fusion welding: NET-TG1 application:...
  • Shan X, Davies CM, Wangsdan T, O' Dowd NP, Nikbin K. Thermo-mechanical modelling of a single-bead-on-plate weld using...
  • Smith MC, Smith AC. NeT Task Group 1 single weld bead on plate: review of phase 1 weld simulation round robin. British...
  • Ohms C, Wimpory RC, Katsareas DE, Youtsos AG. NET TG1: residual stress assessment by neutron diffraction and finite...
  • Bate SK, Charles R, Warren A. Finite Element Analysis of the single bead on plate specimen using sysweld,...
  • Hofmann M, Wimpory RC. NET TG1: residual stress analysis on a single bead weld on a steel plate using neutron...
  • Cited by (102)

    • A probabilistic study of welding residual stresses distribution and their contribution to the fatigue life

      2020, Engineering Failure Analysis
      Citation Excerpt :

      In their study, Naser et al. [3] are limited to study the welded parameters without investigating how these parameters will affect the magnitude and the depth of the residual stresses. To answer the challenge which is “determination of welding RS distribution”, many models [8,9,10,11] have been developed based on finite element analysis using different welding parameters. However, numerical simulation of welding processes is a complex and time-consuming task that requires extensive and accurate input to provide reliable results.

    View all citing articles on Scopus
    View full text