Verification and Validation

TITLE: Verification and Validation AUTHOR: gepr FORMAT: raw -----

There has been alot of use and abuse of the terms "validation" and "verification" in the world of simulation. There are lots of ways these words can be used. But, I'd like to clarify some of the uses with a little text. The terms may seem incompatible with Logical Validity and Logical Soundness; but, they really aren't. Logical validity is the internal consistency of a set of statements, wherein the conclusion follows from the assumptions. Logical soundness is when a conclusion (or any one of the statements in a proof) is externally confirmed as true. Logical validity is comensurate with validation in simulation because the realization is clear that when using simulation, there exists no simple way to establish soundness. Instead, validation is extended to check the consistency of the data used to drive the simulation. All of this presupposes the acceptance of a measurement process.

The Air Force:

AIR FORCE INSTRUCTION 16-1001

1 JUNE 1996
Operations Support

VERIFICATION, VALIDATION AND ACCREDITATION (VV&A)


Taken from (now defunkt):http://204.34.204.77/kb/

  • Verification: Verification is the process of determining that M&S accurately represent the developer's conceptual description and specifications. This is accomplished by identifying and eliminating mistakes in logic, mathematics, or programming. This process establishes that the M&S code and logic correctly perform the intended functions, and to what extent M&S development activities conform to state-of-the-practice software engineering techniques.
  • Validation: The process of determining the degree to which a model is an accurate representation of the real-world from the perspective of the intended uses of the model. Validation process can be used to identify model improvements, where necessary. It has two main components: structural validation, which includes an internal examination of M&S assumptions, architecture, and algorithms in the context of the intended use; and output validation, which determines how well the M&S results compare with the perceived "real world."

My Opinion:

  • Verification is "model matching" or ... establishing that two models are isomorphic or that the differences between models are stated completely (e.g. variant assumptions or impact of implementation on requirements flowdown).
  • Validation: two types
    • Structural -- Structural validation is to ensure that the model assumptions lead to the conclusions of the model.
    • Output -- Output validation is to ensure that the measurements on the model is equivalent to the same measurements on the system being modelled. The output of the model is simply the result of measurements taken on the model. Data is the result of measurements taken on the system. "Equivalence", here, means either "consistent with" or "consistent up to the point defined by well-defined and expected differences" (e.g. commensuration with simplifying assumptions).
glen e. p. ropella