No abstract

The Mobility Aircraft Availability Forecasting (MAAF) model prototype development and study effort was initiated to help the United States Air Force Air Mobility Command (AMC) answer the question, "How can we accurately predict mission capable (MC) rates?" While perfect prediction of aircraft MC rates is not possible, we investigate a simulation-based risk analysis approach. Current prediction methods utilize "after the fact" analyses and user opinion, making it difficult to perform quick, accurate, and effective analyses of potential limiting factors and policy changes, particularly in timesensitive situations. This paper describes the MAAF proof-of-concept model and decision support system built to provide AMC managers the dynamic, predictive tools needed to better forecast aircraft availability. The simulation component featured new capabilities for mobility modeling to include dynamic definition of the configuration of a mobility system, dynamic definition of the capabilities of the individual airbases within a mobility system, improved representation of the aircraft objects within the model, and a new approach to modeling aircraft maintenance including the realistic consideration of partially mission capable aircraft. The development efforts and sample experimental results are recounted in this paper.

Artificial neural networks (ANN) are powerful tools for pattern matching and have proven useful in fault detection and diagnosis. Rule based systems (RBS) have the ability to process complex reasoning paths and explain the results of an analysis. The Hybrid Intelligent Logistics Diagnostic Assistant (HILDA) combines these technologies to augment the abilities of a logistics analyst. HILDA aids a logistics analyst working with the Mobility Aircraft Availability Forecasting (MAAF) simulation in trying to balance the requirements of a mission schedule versus the personnel and aircraft resources available at a set of airbases. This paper describes the design and implementation of the hybrid architecture, with details on the domain specific knowledge embedded in the ANN and RBS modules. Examples show how the HILDA prototype is successful in aiding a decision maker in converging to a successful mission/resource scenario. RÉSUMÉ. Les réseaux de neurones artificiels (RNA) sont de puissants outils pour l'appariement de formes qui ont fait leurs preuves dans le domaine de la détection et du diagnostic des anomalies. Les systèmes experts (SE) sont en mesure d'élaborer des parcours de raisonnement complexes et d'expliquer les résultats d'une analyse. L'assistant intelligent de logistique/diagnostique hybride (HILDA -Hybrid Intelligent Logistics Diagnostic Assistantregroupe ces technologies en vue d'augmenter les capacités de l'analyste logistique. HILDA supporte l'analyste logistique qui fait usage du modèle de simulation des prévisions de déploiement aéromobile (MAAF) -Mobility Aircraft Availability Forecastinget s'efforce de concilier les exigences du calendrier des missions en fonction de la disponibilité des ressources en hommes et en avions dans un groupement de bases aériennes. Cet article décrit la conception et la mise en place de l'architecture hybride et fournit des détails sur les connaissances propres au domaine qui ont été intégrées aux modules RNA et SE. Des exemples montrent de quelle façon le prototype HILDA aide le décideur à élaborer un scénario mission/ressources réussi.

No abstract