(Reliability Analysis of Components and Systems)
Description. Computer program RELACS evaluates probabilities of certain events defined in terms of random variables and one or more smooth limit-state functions. These events correspond to the failures of components or of systems (made up of multiple components). RELACS uses advanced first- and second-order reliability (FORM/SORM) techniques, which makes it ideal for the evaluation of small probabilities arising in structural-reliability applications. The program has the following features and capabilities.
- RELACS is easy to use. It has an interactive, menu-driven user interface and extensive on-line help. The analyst may specify or modify parameters interactively or read them from an existing input file. Program output is displayed on the screen and may also be sent to an output file.
- RELACS can model virtually any probability distribution. RELACS has more than 30 built-in probability distribution types. Additional distribution types are modeled by specifying the first four moments or by reading tabulated values from a file. The analyst can specify correlation between any two random variables, regardless of their distribution type. All this is done without making any problem-specific modifications to the source code.
- The analyst specifies component limit-state functions by means of a FORTRAN subroutine. A system is constructed in terms of several components using minimal cut sets (i.e., unions of intersections).
- RELACS has four options for the evaluation of component failure probabilities with increasing levels of accuracy. The FORM option uses a linear approximation; the SORM option uses a more accurate quadratic approximation. A third option refines the SORM result by means of importance sampling. The fourth option uses Monte Carlo simulation.
- RELACS has extensive capabilities for sensitivity analysis and for the modeling of uncertainty. RELACS evaluates the sensitivity of a failure probability to changes in a distribution parameter or a parameter of the limit-state function. RELACS can also treat any parameter as being itself an uncertain random variable, by using its unique built-in "nested FORM/SORM" capability. This capability is also useful in problems involving both short-term and long-term random quantities.
- RELACS incorporates the SORM and SYSREL subroutine packages developed by Dr. Rudiger Rackwitz and his co-workers (under license from RCP Gmbh, Munich, Germany). These subroutines have been used extensively in structural reliability analysis.
Advanced Capabilities. RELACS contains a number of innovations to allow the solution of a wide range of reliability problems. Some of the most important innovations are described below.
- Modeling of time-variant quantities. The limit-state functions for components may contain one or more stationary time-variant quantities (e.g., wave elevation), not necessarily gaussian.
- Specification of joint distributions using the Rosenblatt transformation. The analyst can specify a joint distribution as a product of conditional distributions; the conditional parameters of a random variable are specified as functions of the values of other random variables.
- Equality constraints. This capability allows the calculation of system failure probabilities, given the state of one system component. This capability is useful in the probabilistic analysis of structural fatigue and fracture.
- Adaptive Response Surface. RELACS has a powerful search algorithm based on an adaptive local quadratic approximation to the limit-state function. This algorithm has proven to be efficient for problems involving many variables or for noisy limit-state functions.
- Environment and Response Contours. These contours allow the visualization of the load and capacity random quantities.
RELACS is being enhanced on an regular basis, under the sponsorship of a Users' Group and as part of in-house consulting projects.
Sample Applications. RELACS has been used by Risk Engineering personnel, by oil companies, and by other consultants in a number of projects. The following is a partial list of these applications.
- Calculation of Extreme Loads and Responses of Several Tension-Leg-Platforms (TLPs) under hurricane loads. Results from these analyses were used to formulate an LRFD design procedure for TLPS.
- Calculation of the Reliability of the risers of a recently built TLP. This study considered extreme winds and waves under hurricanes and fatigue from wave-action and from vortex-induced vibrations during loop-current events.
- Reliability of a partially damaged fixed offshore platform under wave loads. Results from this analysis were used to make decisions about upgrading damaged members.
- Probabilistic analysis of the performance of offshore structures and their foundations in the Gulf of Mexico that were exposed to Hurricane Andrew. Results from this analysis were used to re-evaluate the implied capacity in conventional design practice.
- Calculation of extreme waves with annual exceedence probabilities of 10-4 from hurricanes, without requiring Monte Carlo simulation or hindcasting.
- Reliability of downhole tubing and casing in oil wells. Results from these studies have been used to optimize the placement of different quality tubing and to explore more rational design procedures.
- Tendon-redundancy effects in the Reliability of TLPs. This study investigated the effect of the number of tendons, and assumptions about the correlation in their capacities, on the reliability of a TLP.
- Reliability of structures under ice loads in northern latitudes.
- Fatigue reliability of structural members and the effect of inspection.
- Fatigue reliability of offshore structures during trans-oceanic tows.
- Probabilistic economic analysis of oil development projects.