Probabilistics and Reliability

The effects on the normal use of buildings are regulated by the relevant standards. In contrast, there are no clear rules for defining these effects for exceptional effects from terrorist attacks or industrial incidents. The subject of special considerations in the CWE is therefore the methodological development of risk assessments with the identification of risk scenarios. Among other things, the approaches of probabilistic safety analyses for power plants, which have proven themselves in practice, are used. Probabilistic Security Analysis (PSA) is a systematic method for the holistic assessment of system security, which is based on error tree and event flow analysis. The research priorities at the CWE in this area can be summarised as follows:Probabilistik_1

  • Risk considerations
  • Probabilistic security analyses
  • Fragility Curves and Consequential Analysis
  • Response Surface Methods
  • Polymorphic blur models
  • Numeric Design Methods
  • Reliability analysis
  • Protective measures

Error tree and event flow analyses require fragility curves that describe the relationship between impact variables and probability of failure as cumulative distribution functions.

The aim of the CWE is to develop reliable methods with which fragility curves can be systematically determined on the basis of experimental studies or simulations. As a classic solution approach, Monte Carlo simulations are used in which stress intensities are considered stochastic on the impact side. The dispersions on the resistance side are taken into account by appropriate distribution functions, which can be described continuously or with random fields.

Probabilistic

The concept of the “Design of experiments” (DOE) is used at the CWE to reduce the computational effort. For each intensity of action taken into account, a “Response surface model” (RSM – “Metamodel”) is established depending on the significant influence parameters, which represents a functional relationship between the input parameters and the response variables. The methods described can also be used to derive multidimensional fragility curves that can be used to describe the probability of failure of multiple components at the same time. The methods are implemented in the context of probabilistic safety considerations at the CWE by coupling common program systems (ANSYS, Abaqus, LS-Dyna…) with specially developed evaluation programs and software packages.