Model based predictive control and recommendations
Examples:
Optimize efficiency of sugar extraction (joint with I. Oseledets)
Input X: Beet chips shape, quality, temperature, sugar content and flux; wash water temperature, pH and flux; the temperature inside the diffuser, etc.
Output Y: Costs, losses, efficiency of sugar extraction
Minimize fuel consumption of a cargo vessel, detect frauds with fuel, optimize expedition route
Input X: Dimensions (length, height, width), load capacity, type (ferry, barge, freighter, etc.), number of engines, etc.; Route data; information about weather and sea currents (historical, predictive and real-time); controls (vessel speed, etc.)
Output Y: Fuel consumption
Some challenges are the presence of heterogeneous data and noise, large volumes of high-dimensional stream data, missing values, outliers/incorrect values, etc.
Large-Scale Shape Retrieval and Classification via 3D Deep Neural Networks
Examples:
3D data is widespread, e.g.
- 3D CAD models
- Remote sensing data from satellites
- 3D medical images, etc.
For applications it is necessary to
- Recognize/categorize 3D shapes (e.g. CAD models)
- Retrieve similar shapes
- Predict characteristics of 3D objects
Used methods
- Voxelization
- Sparse 3D convolutional deep neural networks
- Local features based on differential geometry
Some challenges are the presence of heterogeneous data and noise, large volumes of high-dimensional stream data, missing values, outliers/incorrect values, etc.
Applications of massive data processing and predictive maintenance technology
- Prediction of failures in auxiliary power units
- Aerodynamic design of efficient layouts for passenger aircraft
- Design of the side panel of F1 car
Applications of deep learning technology
- 3D data processing
- 3D shape recognition
Development & implementation of software libaries
pSeven Core (MACROS Library)
The library was developed for optimization and modeling of surrogate functions in collaboration with DATADVANCE
Quality assurance:
- Technology Readiness Level 6 (NASA classification)
- According to Airbus experts , application of MACROS “provides the reduction of up to 10% of lead time and cost in several areas of the aircraft design process”
- Several joint projects with industry partners have been successfully completed
Use cases:
- Structural analysis of composite stiffened panels on aircrafts
- Aerodynamic design of layouts for passenger aircrafts
Library for predictive maintenance (PHM core)
Quality assurance:
- Technology Readiness Level 5 (NASA classification)
Use cases:
- Airplanes leakage detection
- APU failures prediction
- Oil filter clogging detection
- Software-intensive systems: detection of outages of internet user services