Battle situations can create loss of essential ship functions and capacities such as communications, power, cooling, weapons, etc. System state restoration today is limited by state constraints imposed by the extremis conditions, resource availability, access to critical system information, and by operator proficiency and experience. The explosion of data from various interdependent sensors, each competing for the operators attention through numerous information displays and alerts/alarms can result in cognitive overload. The Navy recognizes that a different approach is needed to make sense of the situation, and present timely, contextually-coherent, actionable, and goal-oriented options to the operator. We propose research into the selection and fine-tuning of machine learning algorithms to develop a system state awareness for mission-critical Navy ship systems. We propose to add the machine learning engine and meta information generation algorithm into the Navys Condition Monitoring System (MCS) baseline software. The ML engine is trained on both machinery and other sensor data as well as historical event data. We also propose use of synthetic event data where possible to expand/enhance training data sets.
Benefit: Provide consistent near-term outcome predictions/awareness for complex system conditions using advanced machine learning algorithms. Improve feature prediction in complex multi-dimensional data sets. Achieve greater predictive accuracy at lower cost using expand synthetic data sets. Train not only algorithms but also new operators using Simulation environment. Ensure that algorithms continuously learn by collecting and integrating live operational data. Improve operator cognition of system behavior and risk.
Keywords: Operator cognitive overload, Operator cognitive overload, Predictive Analytics, Digital twin, Machinery Condition Monitoring, Machine Learning Algorithm, Synthetic Data, Operating under uncertainty, Situational Awareness
