OPTO-KNOWLEDGE SYSTEMS INC — National Aeronautics and Space Administration SBIR Phase I: S5

Description

Testing and verifying rocket engines requires the careful placement, calibration, and analysis of dozens of sensors to measure important system characteristics such as temperature, density, flow, vibration, etc.Due to the volume and complexity of the raw outputs of these sensors, most are not monitored live and require extensive amounts of time from engineers to analyze and verify tests in an offline manner.In recent decades, Machine learning (ML) technologies have been proven to speed up the process and reliability of identifying anomalies.However, as the volume and complexity of sensor data has grown, traditional ML methods have failed to scale and require carefully crafted sensor models to get meaningful results.More recently, Deep Learning (DL) methods have been proven to achieve superhuman performance and reliability in extremely complex domains such as object recognition, natural language processing, and games.They achieve this by identifying patterns and relationships in data that humans are unable to quantify and encoding them hierarchically within “deep” neural networks.With DL technology having been proven in many domains, we will build a DL-based, real-time engine sensor diagnostic and health management system to enable superhuman-level analysis and decision-making during firing tests and launches.