Authors:
Ekaterina Auer
;
Roger Cuypers
and
Wolfram Luther
Affiliation:
University of Duisburg-Essen, Germany
Keyword(s):
Numerical Verification Assessment, Validation, Uncertainty, Result Verification.
Related
Ontology
Subjects/Areas/Topics:
Industrial Engineering
;
Informatics in Control, Automation and Robotics
;
Performance Evaluation and Optimization
;
Quality Control and Management
;
Systems Modeling and Simulation
Abstract:
Common verification and validation methodologies suggest using verification benchmarks in the cases where no theoretical proof of the model and algorithm accuracy is possible and interpreting the outcome of the process of interest on these. In this paper, we propose to shift the focus on the process itself. Helped by appropriate questionnaires, users are urged to subdivide their processes into smaller tasks and to classify the input/output data and algorithms, which allows us to assign a certain initial verification degree to the (sub-) process from a recently proposed four-tier hierarchy. After that, the initial degree can be improved as well as uncertainty quantified by with the help of a selection of specialized interoperable data types and tools. Besides, we address issues of software quality and user support by describing a comparison system for verified initial value problem solvers.