Automated Generation of Temperature Fields for Numerical Welding Simulation

• PITTNER Andreas

  Corresponding author; [e-mail]

• SCHWENK Christopher

  Federal Institute for Materials Research and Testing (BAM), Division "Safety of Joined components", Berlin, Germany

• RETHMEIER Michael

  Federal Institute for Materials Research and Testing (BAM), Division "Safety of Joined components", Berlin, Germany

• WEIßr; Dietmar

  Sauer Danfoss Aps, Sensor Core Competence Centrc, Nordborg, Denmark

Abstract

The investigation of welding induced distortion requires knowledge regarding the heat effects of welding in terms of a precise description of the temperature field. Welding simulation enables to understand basic phenomenological mechanisms and to evaluate different welding configurations which is, in contrast to experimental investigations, more time- and cost-efficient. The complex physical phenomena of welding processes demand the calibration of phenomenological models against experimental reference data. The calculation of a temperature field that approximates the experimental one very accurately is still a time consuming and difficult task. This is caused by the fact that the model input which yields the optimal temperature field cannot be derived from parameters of the real process directly. This paper focuses on a recently introduced approach that combines empirical and phenomenological modeling techniques. The calibration of the weld thermal models is performed automatically by means of a global optimization scheme and does not require the definition of initial model input parameters. This key feature confirms the benefit potential when it comes to practical utilization of welding simulation because the calibrated temperature field is provided within few minutes without the need for expert knowledge. This is the basis for an efficient determination of the welding induced distortions for industrial applications. Furthermore, the representation of the relationship between model input parameters and process parameters by a neural network enables to predict the temperature field for unknown process parameters that were not considered for calibration. Consequently, the combination of phenomenological and empirical models permits to reduce the experimental effort and bridges the gap between numerical and experimental investigations of the heat effects of welding. The application of this modeling approach is independent of the welding technique under investigation. Exemplarily it is tested for laser beam and GMA-laser hybrid welding. All numerical calculations are validated experimentally.

Journal

• QUARTERLY JOURNAL OF THE JAPAN WELDING SOCIETY

  QUARTERLY JOURNAL OF THE JAPAN WELDING SOCIETY 27 (2), 219s-224s, 2009

  JAPAN WELDING SOCIETY

Keywords

• Welding Simulation
• Temperature Field Generation
• Optimization
• Neural Networks

Details

• CRID: 1390001204724756096
• NII Article ID: 10030149228
• NII Book ID: AN1005067X
• DOI: 10.2207/qjjws.27.219s
• ISSN: 24348252
• ISSN: 02884771
• Text Lang: en
• Data Source:
  • JaLC
  • Crossref
  • CiNii Articles
• Abstract License Flag: Disallowed