DFM Specifications

DFM 2.2

Oct 2010

The DFM 2.2 specification will be released on 1 Oct 2010. It introduces:

  • Time-dependent variables
    • States can be assigned time-dependent probabilities if:
      • The node is binary
      • The node is multi-state and the states are disjoint
    • The time duration can be:
      • Constant (specified by the user)
      • Measured from the start of the analysis
      • Measured from the occurrence of an intermediate DFM Event
    • Time-dependent probabilities can be calculated from:
      • Rates
      • Probability distributions

DFM 2.1

Mar 2010

The DFM 2.1 specification introduced:

  • Variable-delay Transition Boxes
    • Previously, the delay was a characteristic of the Transition Box and every Edge into a particular Transition Box necessarily had the same delay.
      • Modeling transfer functions that used state information from multiple time-steps required the use of history Nodes that stored state information from previous time-steps.
      • A Transition Box was defined as a Transfer Box with a non-zero delay.
    • From DFM 2.1, the delay is a characteristic of the Edge that enters the Transition Box.
      • Each Edge into a particular Transfer Box can have a distinct delay.
      • Feedback loops across multiple time-steps can now be modeled directly and history Nodes are now obsolete.
      • A Transfer Box can have multiple input Edges from the same Node, as long as each Edge has a different delay
      • A Transition Box is now defined as a Transfer Box with at least one input Edge with a non-zero delay.

DFM 2.0

Jan 2009

DFM 2.0 introduced two new features: probabilities and quantification.

  • Probabilities
    • The states of a source node can be assigned point probabilities.
    • If a model contains states with probabilities, it's DFM Events, Prime Implicants, and DFM Results have probabilities determined by the probabilities of the states they contain.
  • Quantification
    • Upper Bound
      • The quantification of a Complete Base produces an upper bound for the probability of the Top Event occurring.
        • If the Prime Implicants are mutually exclusive, the upper bound is also the real probability.
        • If Prime Implicants share events, the shared events are over-counted and the result is an upper bound of the true result.
      • The upper bound is the equivalent of the Min Cut Upper Bound used by Fault Tree/Event Tree methods.
    • Exact Quantification
      • The Complete Base can be converted into a set of Mutually Exclusive Implicants.
      • Because they are mutually exclusive, quantifying the set of Mutually Exclusive Implicants always produces the exact result.

DFM 1.0

Jun 1993

DFM 1.0 was the first release of DFM.

It included all of the current DFM structures:

  • Directed graph representation
  • Nodes
    • Discrete nodes
    • Continuous nodes
    • All nodes are discretized into multiple states
  • Transfer boxes
    • Transfer boxes with no time delay
    • Transition boxes with time delays
    • Boxes contain a decision table that maps discrete input combinations to discrete output combinations
  • Directed edges
    • Causal Edges
    • Conditioning Edges

It also contained the basic DFM analysis capabilities:

  • Inductive Analysis
    • Directed-flow (simulation mode)
      • Explores a single path through the DFM model
    • Exhaustive (analysis mode)
      • Generates all possible events that can result from the Initiating Event
  • Deductive Analysis
    • Generates all possible initial states that could have produced the Top Event