Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Machine Learning based Prediction of Ditching Loads

Henning Schwarz, Micha Überrück, Jens-Peter M. Zemke, Thomas Rung

TL;DR

Results indicate a satisfactory level of predictive agreement for all four investigated surrogate models examined, with the combination of an LSTM and a deep decoder CAE showing the best performance.

Abstract

We present approaches to predict dynamic ditching loads on aircraft fuselages using machine learning. The employed learning procedure is structured into two parts, the reconstruction of the spatial loads using a convolutional autoencoder (CAE) and the transient evolution of these loads in a subsequent part. Different CAE strategies are assessed and combined with either long short-term memory (LSTM) networks or Koopman-operator based methods to predict the transient behaviour. The training data is compiled by an extension of the momentum method of von-Karman and Wagner and the rationale of the training approach is briefly summarised. The application included refers to a full-scale fuselage of a DLR-D150 aircraft for a range of horizontal and vertical approach velocities at 6° incidence. Results indicate a satisfactory level of predictive agreement for all four investigated surrogate models examined, with the combination of an LSTM and a deep decoder CAE showing the best performance.

Machine Learning based Prediction of Ditching Loads

TL;DR

Results indicate a satisfactory level of predictive agreement for all four investigated surrogate models examined, with the combination of an LSTM and a deep decoder CAE showing the best performance.

Abstract

We present approaches to predict dynamic ditching loads on aircraft fuselages using machine learning. The employed learning procedure is structured into two parts, the reconstruction of the spatial loads using a convolutional autoencoder (CAE) and the transient evolution of these loads in a subsequent part. Different CAE strategies are assessed and combined with either long short-term memory (LSTM) networks or Koopman-operator based methods to predict the transient behaviour. The training data is compiled by an extension of the momentum method of von-Karman and Wagner and the rationale of the training approach is briefly summarised. The application included refers to a full-scale fuselage of a DLR-D150 aircraft for a range of horizontal and vertical approach velocities at 6° incidence. Results indicate a satisfactory level of predictive agreement for all four investigated surrogate models examined, with the combination of an LSTM and a deep decoder CAE showing the best performance.
Paper Structure (28 sections, 15 equations, 24 figures, 7 tables)

This paper contains 28 sections, 15 equations, 24 figures, 7 tables.

Table of Contents

  1. Introduction
  2. Ditching Load Analysis Methodology
  3. Ditching Simulation Method
  4. Validation
  5. Guided Impact of a Curved Plate
  6. Free Motion of a Model Fuselage
  7. Machine Learning Methods
  8. Convolutional Autoencoder (CAE)
  9. Non-local blocks
  10. Long Short-Term Memory (LSTM)
  11. Autoencoder-LSTM Models
  12. Koopman Autoencoder (KAE)
  13. Test Case and Data Processing
  14. Fuselage Geometry
  15. Spatial Discretization of the Simulation and Training Data
  16. ...and 13 more sections

Figures (24)

  • Figure 1: Frequently employed ditching investigation phases.
  • Figure 2: Aircraft cross section and the corresponding quantities for ditch simulations.
  • Figure 3: Illustration of the investigated guided impact of a curved plate case (left: outline of pressure sensor locations; right: sketch of the test rig).
  • Figure 4: Comparison of predicted (ditch) and measured pressures relative to the ambient pressure (sensor locations correspond to blue symbols in Fig. \ref{['fig:smaes_1222']}).
  • Figure 5: Comparison of predicted (ditch) and measured motion of the NACA TN2929 fuselage model D. Side and top view of the fuselage model are depicted in the lower graph.
  • ...and 19 more figures