Development and Implementation of a Framework for Aerospace Vehicle Reasoning (FAVER)

Ian K Jennions,Cordelia Mattuvarkuzhali Ezhilarasu

doi:10.1109/access.2021.3100865

Ian K Jennions, Cordelia Mattuvarkuzhali Ezhilarasu

Open Access

https://doi.org/10.1109/access.2021.3100865

Copy DOI

Journal: IEEE Access	Publication Date: Jan 1, 2021
Citations: 7	License type: CC BY 4.0

Affiliation: Cranfield University

Abstract

This paper discusses the development and implementation of the architecture of a Framework for Aerospace Vehicle Reasoning, `FAVER'. Integrated Vehicle Health Management systems require a holistic view of the aircraft to isolate faults cascading between aircraft systems. FAVER is a system-agnostic framework developed to isolate such propagating faults by incorporating Digital Twins (DTs) and reasoning techniques. The flexibility of FAVER to work with different types and scales of DTs and diagnostics, and its ability to adapt and expand for previously unknown faults and new systems are demonstrated in this paper. The paper also shows the novel combination of relationship matrix and fault attributes database used to structure the knowledge of FAVER's expert system. The paper provides the working mechanism of FAVER's reasoning and its ability to isolate faults at the system level, identify their root causes, and predict the cascading effects at the vehicle level. Four aircraft systems are used for demonstration purposes: i) the Electrical Power System, ii) the Fuel System, iii) the Engine, and iv) the Environmental Control System, and the use case scenarios are adapted from real aircraft incidents. The paper also discusses the pros and cons of FAVER's reasoning via demonstrations and evaluates the performance of FAVER's reasoning through a comparative study with a supervised neural network model.

Highlights

Integrated Vehicle Health Management (IVHM) is an evolving capability that enables the Condition Based Maintenance (CBM) of complex vehicles like aircraft
Evaluation Criteria 1: Accuracy of diagnostics at system level The data-driven reasoning to identify if a symptom vector belongs to an output class using neural networks has an advantage over FAVER’s reasoning, which is, its independence of system diagnostic functions
The trained neural network was able to detect the faults at system level with 100% accuracy without system level diagnostic functions

Summary

INTRODUCTION

Integrated Vehicle Health Management (IVHM) is an evolving capability that enables the Condition Based Maintenance (CBM) of complex vehicles like aircraft. Such unexpected fault propagation paths lead to unplanned downtime costing the airlines time and reputation Troubleshooting and isolating these cascading faults requires a holistic view of the aircraft considering the interactions between its various systems, i.e. vehicle level health monitoring. Airbus Defense and Space demonstrated the effect of a fault in a fuel-cooled oil cooler on other independent systems using a modular framework that employed Bayesian Networks for reasoning [18] These examples of IVHM systems show that the area of vehicle level health monitoring is still underexplored. This paper explains how FAVER’s architecture, influenced by the Open Standard Architecture for Condition Based Maintenance (OSA-CBM), is developed and how it explores vehicle level health monitoring for the aircraft by employing reasoning and digital twins. The ECS model is a MATLAB SIMSCAPE model called SESAC [25], and its system level diagnostic function uses a Linear Discriminant Analysis (LDA) algorithm and can isolate six fault modes

CURRENT CONTRIBUTION

Output for simulation models of DT

Communications layer

ACCIDENT SCENARIO 1

ACCIDENT SCENARIO 2

Summary and Future works