Optimization or Bayesian Strategy? Performance of the Bhattacharyya Distance in Different Algorithms of Stochastic Model Updating

Abstract The Bhattacharyya distance has been developed as a comprehensive uncertainty quantification metric by capturing multiple sources from both numerical predictions and experimental measurements. This work pursues further investigation of the performance in different methodologies for stochastic model updating, thus to prove universality various currently popular updating procedures. first procedure is Bayesian where utilized define an approximate likelihood function transitional Markov chain Monte Carlo algorithm employed obtain posterior distribution parameters. In second procedure, construct objective optimization problem. defined between samples prediction target data. comparison study performed on four degrees-of-freedom mass-spring system. A challenging task raised this example assigning distributions parameters with imprecise coefficients. requires calibrate not themselves, but their properties. employs GARTEUR SM-AG19 benchmark structure demonstrate feasibility presence practical experiment raising measuring techniques, equipment, subjective randomness. results universal updating.