*See also RISKMAN – 20 Years After*

## SOFTWARE FOR RISK ANALYSIS

RISKMAN is a completely integrated menu and window-driven software system for performing quantitative risk analysis. This software has been used to perform large and comprehensive probabilistic safety assessments for aerospace, nuclear power, and chemical process industries. RISKMAN is designed to run on personal computers with certain versions of the Microsoft operating system; with Windows-XP, Windows-Vista and Windows-7 (32 bit). It insulates the user as much as is possible from having to know about the computer and, therefore, helps to automate activities such as data backup and printer control.

RISKMAN is composed of four main modules: Data Analysis, Systems Analysis, Event Tree Analysis, and Fragility Analysis. It also contains three smaller modules for follow-on applications: Mini-Monitor (for assessing time-dependent plant configuration risk profiles), fire scenario screening (to quickly identify the most important from many postulated fires scenarios), and Big Loop Monte Carlo (for uncertainty analysis).

There are also options for software system maintenance, backup and restore, software updates, printer font, and page control. Also integrated into the software are interfaces to fault tree programs such as: CAFTA, FTREX, and SAPHIRE.

RISKMAN embodies the ABS Consulting’s scenario-based engineering approach. It uses event sequence diagrams, fault trees, and event trees for display and calculations, allows for system-specific data updates, and accounts for dependencies between systems.

The development of RISKMAN is a guided by the RISKMAN Technology Group (RTG), a software users� group now in its 21st year. The RISKMAN Technology Group (RTG) is an association, composed of licensees of the RISKMAN integrated software package, formed to advise and support ABS Consulting in the continued development of the RISKMAN code.

The RTG was formed in 1989 and is composed of member organizations representing probabilistic safety assessment analysts and practitioners in the United States, Europe, and Asia.

The RTG meets yearly to establish priorities for planned enhancements to RISKMAN, determine an appropriate membership fee, and elect a steering committee. A new release of RISKMAN is developed a year, incorporating the enhancements approved by the RTG, and is supplied to each RTG member.

RISKMAN, Release 13.0 is currently in use. The annual RTG membership fee per site is $18,000.

Brief descriptions of the four main modules are provided below.

**Data Analysis Module of RISKMAN**

The data analysis module of RISKMAN (software for risk management) is an integrated package for Bayesian data analysis and database management. The program consists of several distinct functions which are described below:

- Development of distributions for event frequencies and mean durations and update of distributions using Bayesian data analysis techniques.
- Perform simple mathematical operations on distributions using Monte Carlo Sampling.
- Database management functions and reports.

**Development of Distributions**

Here the user is allowed the option of creating a distribution, or performing either a one-stage or a two-stage Bayesian update. The following options are available for creating distributions:

- Lognormal distributions
- Normal distributions
- Beta distributions
- Gamma distributions
- Uniform distributions
- Weibull distributions
- User input histograms and Discrete Probability Distributions (DPD)
- Distribution based on empirical data and expert opinion
- Addition and multiplication of distributions
- Addition and multiplication of distributions with constants
- Merging distributions
- Maximum and minimum of distributions
- Editing titles and notes of distributions
- Deleting distributions and input data for two-stage distributions
- Cloning input data and output distributions
- Distribution reports
- Prior-posterior distribution reports
- Two-stage distribution input data reports
- Probability density and cumulative probability graphs
- Export/Import of distributions in ASCII format
- Distribution graphs and experts in .bmp and .pdf formats
- Summary report of distributions presenting the main characteristics of 5th, 50th, 95th, and the mean values, and if applicable, the mean of the prior distributions and the data used in updating.

In a simple one-stage update, the prior distribution is specified by the user using the above options for distributions or an existing distribution in the database may be retrieved. This prior distribution is then updated using site-specific data.

The first stage of the two-stage update allows the use of data from many sources (empirical and expert opinion) to be first combined into a “generic distribution”. Such a distribution maintains the source to source variability and is used as a prior for the second stage. In the second stage, this prior distribution is updated with site-specific data to generate a posterior distribution.

**Perform simple mathematical operations**

Simple operations may be performed with the distributions that exist in the database such as addition, subtraction, multiplication, scaler multiplication.

All of the above except the operation with constants are performed using Monte Carlo Sampling. Operations may be combined into a one-step calculation by writing equations.

**Database management functions and reports**

The software allows the basic management options for the database. These are:

The software is user-friendly and interactive for all operations. Batch functions are available for single-stage Bayesian updates and for printing reports. In addition, if a prior distribution were to be revised, a one-step operation is available to identify all the posterior distributions in the database based on this prior. A batch run for the revision of all these posteriors can then very quickly be assembled.

**Limits of the Data Analysis Module**

The Data Analysis module does not have a limit on the number of distributions.

## Systems Analysis Module of RISKMAN

The Systems Analysis module is used to create failure frequency and unavailability models for individual systems. Fault tree logic as well as algebraic models can be developed.

The standard approach is to input fault tree graphics using RISKMAN’s own fault tree software package or standard textual fault tree input. The module also interfaces with other fault tree programs such as CAFTA and SAPHIRE and can read ASCII input files. The fault tree logic models are solved using a built in cutsets algorithm, the same algorithm as programmed in to SAPHIRE.

The unique features of this module are:

- Automatic addition of common cause failure modes to specified component groups.
- Ability to quantify a system for several different initial system configurations or alignments (such as entire system available, one train in maintenance, etc.).
- Ability to quantify the top event models for multiple boundary conditions (such as all support systems available, one power train unavailable, etc.).

The quantification of the cutsets for each top event can be performed using point estimates, or Monte Carlo sampling for uncertainty propagation. The user may select the rare event approximation, or the min-cut upper bound approach to cutset totaling. The fault tree defined for a system may be used to evaluate the initiating event frequency. The module reads data variables from the database created in the Data Analysis module as well as local variable defined in the Systems module. The quantified Top Event and Conditional Split Fraction values are available to the Event Tree module.

The Binary Decision Diagram (BDD) quantification algorithm is also available. The BDD approach translates the fault tree into an equivalent BDD, and then walks the BDD to evaluate leaf probabilities. Advantages of the BDD approach include; speed of quantification, no limit on minimal cutsets quantified, larger fault trees allowed, exact treatment of success events and not gates, exact totaling of cutset contributors, and no frequency truncation. Point estimate and Monte Carlo BDD quantification are possible.

The reports from the module provide detailed listing of contributors by alignment, cutset, and basic event importance. Systems Analysis reports may be generated for System Notebooks.

**Limits of the Systems Analysis Module**

*Fault Tree Limits:*

80 Character limit for symbol/description

16 Character limit for symbol name

*Quantification Limits:*

Maximum Number of Cutsets Saved per Top Event (no limit for the BDD quantification approach): 102,400

Maximum Number of Split Fractions per Quantification Run: 3,000

Maximum Number of Initial System Alignments per Split Fraction: 64

## Event Tree Analysis Module of RISKMAN

The Event Tree Analysis module of RISKMAN will allow the user to create an event tree and perform event tree quantification. The module uses a database for the tree nodes (Top Events). This database may be developed manually or may be developed automatically from the fault tree analysis output.

The module allows linking event tree modules and develops accident sequences. The sequences are placed in a database that can be used for risk management analysis. Any number of branches may be placed at each node, i.e., both binary and multi-state top events are supported.

Cutoff frequencies may be provided so that during quantification, sequences whose probability falls below the cutoff are not quantified in the analysis. This feature speeds the quantification and makes it possible to build very large event trees, quantify the highest frequencies and bypass the insignificant frequencies. The sequence frequencies truncated by this process are totaled exactly to assure convergence.

Using the sequence database, the module performs importance rankings at all levels; i.e., systems, top events, conditional split fractions, boundary conditions, initial conditions, components and basic events. Computed importance measures include Fussel-Vesely, Risk Achievement, Risk Reduction, Birnbaum, and Fractional Importance.

**Limits of the Event Tree Module**

Up to 30 event trees with a total of 500 top events (nodes)

Up to 300 top events for a single event tree

No limit on number of sequences

## Fragility Analysis Module of RISKMAN

This module accepts seismic hazard characteristics (or for other similar hazards such as for high winds), equipment and structure fragility information and develops the likelihood of failure for equipment or buildings subject to the hazards. Discrete intervals of the hazard curves are used to integrate the equipment/structure fragilities to calculate the equipment/structure failure probability at various seismic hazard levels.

This information may be transferred automatically to the event tree module database for the quantification of event trees developed specially for seismic events. Such seismic event trees can be linked with event trees representing non-seismic failures to account for all cases of failures.

After the event trees identify the failure scenarios from the seismic hazards, the uncertainty analysis performs the same integration over the entire range of the hazard curves accounting for uncertainties in both the hazard and fragility curves.

**Limits of the External Events Module**

100 Discrete Initiators

1000 Components

Unlimited Number of Hazard Curves

Unlimited Number of Top Events

50 Components per Top Event