Aircraft Manoeuvre Classification Using Neural Network Time Series
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for classifying aircraft manoeuvres are hindered by the complexity of different manoeuvre durations and require advanced data processing techniques, along with a high computational burden and the need for careful feature extraction and time window definition.
Innovation Solution
A method utilizing a neural network with an encoder-decoder structure that processes time series data from aircraft monitoring systems to classify manoeuvres into predefined macrocategories without requiring specific feature selection or time window definition, enabling accurate classification of manoeuvres performed by an aircraft.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If advanced data processing techniques are used to classify manoeuvres, then measurement precision is improved, but device complexity increases
Solution Approach 1:
The patent transforms the raw time series data into a different parameter space using wavelet transforms and empirical mode decomposition. This changes the representation of the data from time-domain signals to frequency-time domain coefficients, making the manoeuvre characteristics more distinguishable and easier to classify automatically without complex processing algorithms
Solution Approach 2:
The patent introduces intermediate processing stages including wavelet decomposition, empirical mode decomposition, and automatic feature extraction algorithms that act as mediators between the raw sensor data and the final classification. These intermediaries automatically transform the data into a form that can be classified with simpler algorithms, reducing the overall system complexity
2Manufacturing precision
If manual feature selection and time window definition are performed, then manufacturing precision is improved, but loss of time increases
Solution Approach 1:
The patent implements automatic feature extraction and automatic time window selection algorithms that make the system self-configuring. The algorithm automatically identifies relevant features from the decomposed signals and determines optimal time windows based on the data characteristics, eliminating the need for manual parameter tuning and reducing processing time while maintaining high classification precision
Solution Approach 2:
The patent performs preliminary decomposition of the time series data using wavelet transforms and empirical mode decomposition before classification. This preliminary action transforms the data into a form where features are already separated and organized, making subsequent classification faster and more accurate without requiring extensive manual feature selection
3Adaptability or versatility
If different manoeuvre durations are considered, then adaptability is improved, but device complexity increases
Solution Approach 1:
The patent uses dynamic time warping and adaptive time window selection that allow the analysis to adapt to different manoeuvre durations. The algorithm dynamically adjusts the analysis window size and timing based on the detected manoeuvre characteristics, enabling it to handle variable duration manoeuvres effectively without requiring complex fixed-structure analysis frameworks
Data Source
Figure 1A~2
Figure 1B
Figure 3~18
AI summary
A computer-implemented method (12) for classifying manoeuvres performed by an aircraft (1), including: acquiring (120) a data structure (910) including at least one unknown data matrix (TMX) including a plurality of time series of samples of quantities related to the flight of the aircraft (1), the samples being relative to a succession of instants of time; applying to the unknown data matrix (TMX) a neural network (29) generating a corresponding probability matrix (MXprob(NUM_C,W)) including, for each instant of time of the succession of instants of time, a corresponding probability vector (CX) including, for each class of a plurality of classes of manoeuvres, a corresponding estimate of the probability that, in the instant of time, the aircraft (1) has performed a manoeuvre belonging to the class; and selecting (123), for each instant of time of the succession of instants of time, a corresponding class of manoeuvres, based on the probability estimates of the corresponding probability vector (CX).