Dimensional Analysis and Buckingham's Pi Theorem

Description

Write a candidate law so it is unchanged when the units of its base quantities change, which means expressing it as a relation among dimensionless products of its variables.

Every physical quantity has a value and dimensions such as mass, length and time; changing units rescales values but should not change the truth of a law. Buckingham's Pi theorem states that any law invariant under such unit changes can be written as a relation among dimensionless products of its variables. Listing the variables and their dimensions, then finding the independent dimensionless groups, reduces the number of quantities a model must relate and rules out forms that would change with the choice of units. Constants that carry dimensions, such as a spring constant, are treated as further quantities so the law stays invariant. In UX the technique is mainly a check on proposed performance models: a formula mixing seconds, pixels and millimetres should reduce to dimensionless terms, as Fitts' index of difficulty (distance over width) does.

Attributes

Loss type
Unmigrated
Effort
Hours

Steps

  1. List every variable in the candidate law with its dimensions in a chosen base (for example time, length, count).
  2. Write the matrix of dimension exponents and find the independent dimensionless products of the variables.
  3. Express the law as a relation among those dimensionless groups, introducing any dimensional constant as an explicit quantity.
  4. Check that the proposed functional form is unchanged when each base unit is rescaled; reject or revise forms that are not.
  5. Use the reduced set of groups to design experiments that vary one group at a time.

Sources

  • Foundations of Measurement, Vol. 3: Representation, Axiomatization, and Invariance (Luce et al., 1990)
  • Foundations of Measurement, Vol. 1: Additive and Polynomial Representations (Krantz et al., 1971)

Also known as

  • Buckingham Pi theorem
  • Dimensionless groups
  • Dimensional invariance
  • Dimensional analysis

Categories

  • Information Theory, Physics & Mathematics