Work Package 2

Predictive Maintenance

Our Challenge

Comprehensive maintenance is a crucial part in aviation and smooth airline operations. Most components of an airplane must function at all times to ensure safety during flight, but are also exposed to highest stress on a daily basis. Therefore, deterioration is an inevitable consequence in the aviation business and is answered mainly with preventive or reactive strategies. What these strategies have in common, is their reduced cost efficiency. For preventive maintenance actions the costs can explode very quickly, because the danger of replacing components too fast and therefore unnecessarily early is very high. Reactive maintenance on the other hand can mean significant disruptions for airline operations and relates by itself to huge risks and high costs for an airline business. Moreover, certain incidents, in the cabin for example, are currently only reported during turnaround operations which leads to increased delays or tail swaps and consequently increased costs for the airline. For analytics purposes, means for secure cabin data lake access and ground transfer are of high interest. This data should serve analytics purposes ranging from data driven product enhancements to operational optimizations providing operational (near) real time data of the digital twin. Data security and sovereignty is one of the most crucial aspects when handling this valuable asset.

Our Vision and Mission

We want to enable airlines to concentrate on planning their maintenance schedule more accurately instead of dealing with disruptions caused by failing components. Therefore, we focus on optimizing the costs between reactive and preventive maintenance by utilizing predictive models to find the sweet spot for replacing fatigued components. To achieve this goal, we implement and integrate predictive algorithms into the maintenance, repair, and operations (MRO) process for airlines and aim for an exact prediction of remaining useful lifetimes and failure probability of cabin and turbofan components. These predictive models enable us to include MRO processes into the holistic optimization in respect to flight plan and trajectory as well as turnaround and crew constraints. Therefore, maintenance and turnaround activities can be prepared and planned well in advance. This includes the effective tracking of spare parts and efficient dispatch of aircrafts to appropriate locations in time. Another objective aims at detecting in-flight cabin incidents and providing appropriate immediate solutions well ahead of turnaround operations. Further direct on-demand transactions are combined with the near real time cabin data to visualize and analyse these in cloud infrastructures. Beside the real time analytics part, a product history based approach in identifying a predictive health algorithm is ongoing with data scientists.

our use cases and Cooperation

MTU / TWT GmbHPredictive Maintenance Module
Responsible for processing engine data and generating a failure probability for selected auxiliaries.

Diehl Aerospace Cabin Management System
Controls, monitors and collects data from cabin equipment including lighting, lavatory, passenger service units etc. in order to predict their remaining useful life.

Jeppesen in cooperation with TU BraunschweigAirplane Health Management System
The Aircraft Health Management-Module offers an interface to the health data of aircraft systems. It collects error codes and monitors the system parameters in order to initiate maintenance requests before a malfunction occurs. During maintenance it provides all required data for maintenance crews to isolate and repair dysfunctional systems.

JeppesenTail Assignment
This system assigns the available assets (aircraft) to the flight schedule. In other words: the tailnumber of an aircraft is assigned to the flight number of the flight schedule. Tail assigment can support recovery by applying a tail swap (= A/C type unchanged, but different tail number) or aircraft swap (= other A/C type)

Rolls-RoyceLRU Tracking System
Provides instant and dynamic information about the availability of required LRUs along with relevant data.

SAPData Intelligence
Cloud based data orchestration system which allows one to build data-driven processes and pipelines across complex enterprise landscapes with a unified governance.

  • Oil and fuel filters: Oil and Fuel Filters Differential Pressures are used as indicators for when a replacement of these filters is necessary. These values increase as more flights are completed. Exceeding a certain threshold, the filters must be replaced. Both linear and exponential functions were tested to predict the number of flights left until the threshold is exceeded.
  • Data based anomaly detection using autoencoders: Measurement parameters of engines change in a certain way over time, whereby these parameters heavily interdepend. Deviations from this normal change behaviour indicate malfunction or other exceptions worth a second look in the add-on devices. Different types of autoencoders are used to represent normal behaviour in order to detect abnormal data points and thereby detect anomalies.
  • Lavatory water and waste system:
    A line-replaceable unit located in the water and waste system of the lavatory is analyzed. Here the database of product lifecycle information is curated and used to determine how its efficiency changes over time. Consequently, the remaining useful life of the equipment can be predicted and this information delivered in-flight so that necessary maintenance preparations are made before the aircraft lands.
  • Live incident communication: Incidents in the cabin are reported live to the ground operation centre and corresponding solutions are sent back to try and resolve the situation immediately so as to avoid unnecessary turnaround activities and delays.