The Journal of Defense Modeling and Simulation Applications Methodology Technology · 2007 · 15 citations · 6 references
EngineeringSystems EngineeringModel-based Systems EngineeringSensorsSystems ModelingImage SensorsInstrument ValidationModel CalibrationModel ValidationThermal SensorsVerificationMeasurement Model ValidationModel TestingSensitivity Analysis Of Model OutputsSensor TestingSensor ValidationValidation MetricsValidation Metrics SelectionVerification And Validation
In generic electro‑optic sensor system models, questions of model credibility pose greater challenges than in application‑specific models because a generic formulation must accommodate many applications. This paper investigates how to test, verify, and validate such a generic electro‑optic sensor system model. The authors propose a structural approach that builds confidence through bottom‑up testing, structured verification procedures, and validation metrics based on a geometrical view of model outputs compared with measurements via qualitative methods or quantitative image‑processing, neural‑network, or fuzzy‑pattern‑recognition techniques. The proposed tools outperform traditional validation methods for complex models, offering insight into validity and sensitivities, and have been applied alongside conventional metrics to test, verify, and validate the generic model configured as a thermal imager system.
In general, questions of model credibility introduce more problems in a generic model than they do in models developed for one specific application since a generic formulation must allow for many applications of the model. This paper addresses the issues of model testing, verification, and validation for a generic electro-optic sensor system model. A structural approach to testing, verification, and validation is proposed that builds increasing confidence through bottom-up testing, structured verification procedures, and carefully selected validation metrics. These metrics are based on a geometrical view of model outputs that may be compared with measurements using qualitative methods or quantitative approaches involving image processing, artificial neural networks, or fuzzy pattern recognition. The advantage over traditional validation methods is most marked in the case of complex models with many key quantities where it not only provides insight about the validity but also about sensitivities. These validation tools have been applied, in conjunction with more traditional metrics, to the testing, verification, and validation of the generic model configured as a thermal imager system.
6