Standard Article

Multiple-Model Structural Identification

Civil Engineering Applications

  1. Ian F. C. Smith

Published Online: 15 SEP 2009

DOI: 10.1002/9780470061626.shm169

Encyclopedia of Structural Health Monitoring

Encyclopedia of Structural Health Monitoring

How to Cite

Smith, I. F. C. 2009. Multiple-Model Structural Identification. Encyclopedia of Structural Health Monitoring. .

Author Information

  1. Ecole Polytechnique Fédérale de Lausanne (EPFL), Institute of Structural Engineering, Lausanne, Switzerland

Publication History

  1. Published Online: 15 SEP 2009

Abstract

Complex structures create important challenges in structural health monitoring. Direct measurement of all phenomena is not possible. Modeling of behavior is rarely accurate. Measurement systems also have errors that may compensate modeling errors so that wrong models are identified. Generation and subsequent filtering of candidate models using rationally defined error thresholds provides systematic support to structural identification. Five orders of identification are described. At the lowest level (order 0), the model used for design is calibrated with measurements. At the highest level (order 4), candidate models are generated using the criterion that the differences between measurements and predictions fall below a threshold value that is fixed probabilistically. At identification order 4, candidate models are then filtered using data-mining techniques and the addition of specially targeted measurement data. The best order to choose depends on available information and resources, as well as the level of decision support required for specific infrastructure-management tasks.

Keywords:

  • models;
  • model-based diagnosis;
  • measurements;
  • errors;
  • data mining