Analytics Engines for Mitek’s Performance Digital Twin (PDT) software have been deployed in several airline and NASA projects to process historical fleet data sets collected from aircraft Flight Data Recorders (FDRs). One of the projects used RDT for analysis of aircraft safety in collaboration with NASA Ames Research Center and an FAA-funded MITRE center. Other projects looked at consistency of airline Flight Operations Quality Assurance (FOQA) data collection and aircraft fuel performance analysis for airline fleet.
The analytics build upon baseline models based on engineering knowledge. The models include airframe, flight actuators, sensors, propulsion, and analytical redundancy relationships (both kinematic and dynamic). One view of the Predictive Analytics is that is builds engineering models of aircraft dynamics by calibrating parameters like aerodynamic coefficients on the data. Similarly, for engine gas path performance model, performance map coefficients are calibrated. This allows analysis results to be easily understood by aerospace engineers. The PDT analytics scale up the baseline model to multi-level models that are fitted to fleet historical data (Big Data) and characterize the parameter variation across the fleet and parameter trends over time.
One example PDT analytics application project illustrated in the figure involved data from aircraft, including jet engines, collected from flight data recorders. The data was collected for hundreds of sensors every second of each flight in a million flights over 3 years of commercial airline operation. Predictive Analytics estimated multi-level model with 10 million regressors from 1 trillion data points (flight data for 300-aircraft fleet over 3 years); these computations take 12 hours on a 20-core cluster. Predictive Analytics characterizes trends and anomalies of key performance parameters such as flight surface offsets, aircraft drag and lift coefficients, and engine gas path performance efficiency parameters. These parameters are hidden, are not directly measured in the data.
A Proof-of-Concept project included Report Generation to show the predictive and prescriptive analytics results as 1GB of portable hyperlinked HTML reports. The executive fleet-level reports allow drill-down access to unit (aircraft-level) reports, record (single-flight-level) reports of anomalies and specific data channel problems. The reports provided visibility into fuel efficiency, safety, and data quality issues at fleet scale.
Jet Engine Fuel Efficiency
In another example of PDT deployment, the Prescriptive Analytics deliver value by identifying engines that consume excess fuel or have sub-standard thermal efficiency. Targeted maintenance of the bad actor engines commonly brings cost savings of 2% or more. The analytics also enable condition based maintenance by detecting the anomalies and isolating the faults in the data. More information about the jet engine fuel efficiency solution is available on request.