Frank T-norm (Frank Product-Sum Family)

The Frank t-norm is a family of rules that combines two membership degrees, or the two components of an intuitionistic fuzzy number, and changes shape according to an s parameter set by the user. It is not a decision method in its own right: it runs the combination step inside methods such as the intuitionistic and q-rung orthopair fuzzy weighted average, where several experts' or several criteria's assessments are brought together. Its output is not a ranking but a single number, or a new membership–non-membership pair. Frank (1979) defined the family while investigating which binary operations satisfy both the t-norm axioms and a specific algebraic property (mutual associativity) together with their own co-operation, the t-conorm; the family was later used as an aggregation operator in q-rung orthopair fuzzy decision-making research (Seikh and Mandal, 2022).

A t-norm answers one question: if two conditions are each satisfied to some degree, to what degree are both satisfied together? The Frank family arose from mathematicians seeking the most general answer to the question "what algebraic property can a t-norm and its co-operation, the t-conorm, jointly preserve"; for this reason the Frank family is known as the only family that satisfies both the t-norm and t-conorm axioms and this particular algebraic property at the same time. The parameter s sets how "hard" or "soft" the combination will be. As s falls (towards 1) the rule approaches ordinary multiplication; as s grows the combination rule follows a different curve and drifts towards two extreme behaviours, the minimum rule and the Łukasiewicz rule. In a decision context, this is the mathematics of being able to stretch the shape of a combination rule to fit the data.

The method proceeds through three steps.

First, choosing the parameter s. s can be any number greater than zero other than 1; as s approaches 1 the rule reverts to ordinary multiplication. For example, taking s=2 for 0.6 and 0.7, the Frank product comes out at approximately 0.403, which is smaller than the ordinary product of the same two values, 0.42.

Second, the Frank product and sum. The two values are combined using the chosen s. The product rule corresponds to "AND" logic and the sum rule to "OR" logic, and the two mirror each other. At s=2, the sum of 0.6 and 0.7 comes out at approximately 0.897.

Third, combining two intuitionistic fuzzy numbers. When two assessments are combined, the membership degrees are brought together with the Frank sum and the non-membership degrees with the Frank product. The combination becomes softer or harder depending on the chosen s, and the result remains a valid intuitionistic fuzzy number.

The formulas behind each step, the intermediate tables and the citation formats are given on the DecisionMind method page; this card carries no formulas.

The output of the Frank t-norm is not, on its own, a decision result; it is an intermediate step inside a host method. The score derived from the output, membership minus non-membership, shows the overall direction of the combined evidence. This score can be compared only against other scores computed with the same parameter s. A score computed at s=2 and one computed at s=20 may look as though they sit on the same scale, but they come from different curves. As s grows, the gap between alternatives typically shrinks; this does not mean "the result became more reliable as s grew," but can instead signal that the analysis has entered a region sensitive enough that a small change in the data could change the ranking.

Thus instead of writing:

"The Frank aggregation clearly favoured this alternative"

the report should read:

"These assessments were reduced to the following scores under the Frank rule with s=X; the report should show how the ranking changes if s is varied"

The Frank t-norm works with intuitionistic fuzzy data: each assessment consists of a membership and a non-membership degree, and the two must not sum to more than 1. The same rule is used with q-rung orthopair fuzzy and Pythagorean fuzzy data types. DecisionMind has no separate extension for this building block. You need at least two intuitionistic fuzzy values and a choice of s (s>0, s≠1). The Frank t-norm neither requires nor produces weights; weighting is handled separately within whichever host method it runs inside.

The Frank t-norm is a sound choice if your data is intuitionistic or q-rung fuzzy and you want to be able to stretch the shape of the combination rule to fit your data set. It should not be used in the following cases: if your data is crisp, a t-norm is unnecessary. If you intend to use a value of s very close to 1, the rule already reverts to ordinary multiplication, so a parametric family offers no additional benefit. If you have no justification for s and will simply use "the software default," the choice stops being transparent.

Intuitionistic fuzzy data, the shape of the combination curve needs to flex with the data → the Frank t-norm (s)

Same need, but a fixed, uncontested rule will do → the Einstein t-norm

Adjustability is wanted, but a power-based family is preferred over product-sum → the Schweizer-Sklar t-norm (p)

Data is crisp, no combination rule is needed → a direct weighted average

The Frank t-norm's chief strength is its mathematical distinctiveness. It is the only family that satisfies both the t-norm and t-conorm axioms and the property of mutual associativity, which is why it is favoured in theoretical work. Parameter s covers a wide range of behaviour, from ordinary multiplication to distinct extreme behaviours, within a single formula. It is computed with a closed-form formula and requires no iteration. Applications exist in the q-rung orthopair fuzzy decision-making literature (Seikh and Mandal, 2022).

Its limitations are the price of its flexibility. The choice of s mostly comes from the researcher's preference rather than from the data; the same data set can produce different rankings under different values of s (Xia, Xu and Zhu, 2012). Second, the formula becomes mathematically undefined at the boundary s=1 and a special limiting rule (ordinary multiplication) must be applied, a technical detail that can be overlooked in practice. Third, as s grows, the score gaps between alternatives tend to shrink, which can make the ranking more vulnerable to small changes in the data. Fourth, reports frequently give no justification for s beyond stating that a particular value was used, which makes the choice look arbitrary.

The most common mistake is using s=1 directly in the formula; this leads to division by zero, and the ordinary-multiplication limit should be used instead. A second mistake is using s=0 directly; here the rule should switch to the minimum rule. A third mistake is confusing the Frank t-norm with the Hamacher t-norm; the two are distinct parametric families, and the same s and the same γ produce different results. A fourth mistake is sharing the result without reporting the parameter s; without knowing which s was used, the result cannot be reproduced.

The governing principle is this:

The Frank t-norm does not fix an aggregation result but the parameter s under which that result was produced; the choice of s and its justification must be recorded in the report.