Aircraft Manoeuvre Classification Using Multi-Window Flight Data

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current methods for accurately identifying aircraft manoeuvres during flight are hindered by the complexity of different manoeuvre durations, requiring advanced data processing and struggling to correctly classify manoeuvres due to varying load patterns.

Innovation Solution

A method and system that utilize sensors to acquire and process time-series data from aircraft, employing a training data structure and multi-class classification techniques to differentiate between manoeuvres, regardless of duration, by extracting feature vectors and applying supervised learning classifiers to accurately categorize manoeuvres.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If advanced data processing techniques are used to identify manoeuvres from time trends, then manoeuvre identification accuracy improves, but system complexity increases

Engineering Contradiction:
Improvemanoeuvre identification accuracyVSAvoiddata processing complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent segments the continuous time trend data into discrete manoeuvre events by identifying start and end points based on threshold criteria. This segmentation transforms the complex continuous analysis problem into simpler discrete event detection, reducing processing complexity while maintaining identification accuracy

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent extracts key特征的 parameters from the full time trend data, such as maximum values, minimum values, and threshold crossings. By extracting only the essential features needed for manoeuvre identification rather than processing the entire data set, the system achieves accurate manoeuvre detection with reduced computational complexity

Inventive Principle:
Principle #2Taking out (Extraction)

2Quantity of substance

If monitoring systems acquire large numbers of measurements to study manoeuvre history, then data completeness improves, but data processing difficulty increases due to varying manoeuvre durations

Engineering Contradiction:
Improvemeasurement data volumeVSAvoidmanoeuvre classification difficulty
Core Design Contradiction:
Quantity of substanceVSDifficulty of detecting and measuring

Solution Approach 1:

The patent employs dynamic time windowing that adapts to varying manoeuvre durations. The analysis window automatically adjusts its length based on the detected manoeuvre characteristics, allowing the system to process complete manoeuvre data regardless of duration variations. This dynamic adaptation resolves the difficulty of classifying manoeuvres with different time scales

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The patent changes the parameter of time window length to accommodate different manoeuvre durations. By varying the analysis window size based on the specific manoeuvre being detected, the system can accurately process both short and long-duration manoeuvres from the same data set, eliminating the need for separate processing procedures

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentEP4016221B1Method and system for detecting and classifying manoeuvres executed by an aircraft on the basis of measures acquired during a flight of the aircraft
Publication Date: 2023.06.14 LEONARDO SPA
  • EP4016221B1 patent drawingFigure 1A~2
  • EP4016221B1 patent drawingFigure 1B
  • EP4016221B1 patent drawingFigure 3

AI summary

Method implemented through a computer (12) for detecting the execution, by an aircraft (3), of a manoeuvre belonging to a macrocategory among a plurality of macrocategories (MC), including: receiving a data structure (205) with a plurality of time series of values of quantities relating to a flight of the aircraft (3); for each time duration (TW'p) among a plurality of predetermined time durations (TW'), selecting (310) a corresponding subset of the data structure (205) and extracting (320) a corresponding feature vector (FVXkp); on the basis of the feature vectors (FVXkp), generating (330,335;500,510;720;820) a corresponding input macrovector (MPVX'k; MPVX"k, MPVX'''k; MFVXk) and applying (338;520;730;830) to the input macrovector (MPVX'k; MPVX"k, MPVX'''k; MFVXk) an output classifier (151;251;561;721) to generate estimates indicative of the probability that the aircraft (3) was performing manoeuvres belonging to the macrocategories (MC).