Towards Accreditation of Diagnostic Models for Improved Performance

Abstract : The research community mainly concentrates on developing new and updated diagnostic algorithms to achieve high diagnostic performance which is necessary but not sufficient for the diagnostic models that are embedded in software. The focus of this paper is to understand the requirements for accrediting diagnostic system models to meet high performance and safety criticality in case of both models and embedded system (model + software). For embedded systems, models need to be accredited first to allow a more accurate distinction of whether the model or the code within which the model is embedded is the cause of degraded performance. This is because, neither standards for models and simulations (NASA-STD-7009) nor software engineering requirements (NPR 7150.2A) are sufficient to accredit the models in embedded systems. NASA-STD-7009 assesses the correctness of the physics in models and simulations and NPR 7150.2A lists software engineering requirements for NASA systems. Thus, it is important to understand the accreditation standards in terms of performance requirements of models in embedded systems that can smoothly transit from NASA-STD-7009 to NPR7150.2A.