Patterns for Handling Numerical Uncertainty in Software Applications

In this paper, the UNCERTAIN VALUE and two other supportive design patterns (UNCERTAINTY PROPAGATION and VALUE INFER-ENCE) are presented. UNCERTAIN VALUE manages uncertainty and tolerance limits in numerical data values to represent proba-bilities and likelihoods, as well as error-margins. This uncertainty is propagated throughout the system, updated in calculations according to the UNCERTAINTY PROPAGATION rules, to be finally considered in comparisons and decisions to fulfill a needed deci-sion confidence. For this some VALUE INFERENCE technique must be used. Altogether, this can be used to evaluate the precision and accuracy of values and to make better informed and more precise decisions. This can also be used for sensor fusion, probabil-istic models, predictive maintenance, and in safety to model failure scenarios for mitigating hazards and risks.