Industries
Travel and Logistics
Data and AI in Travel Industry for a Memorable Passenger Experience
Highlights
- Maintenance is a major cost factor for airlines in the MRO function, and scheduled maintenance practices followed traditionally add to the wastage of resources and components.
- Predictive maintenance (PdM) strategies that collect and analyze various aircraft-related data using AI-ML not only save cost and reduce wastage by enabling preemptive actions, but also ensure airworthiness and quality control.
- Predictive maintenance will give rise to a maintenance culture that guides and recommends MRO actions to minimize unscheduled events, unplanned grounding of aircraft, and costly flight delays and disruptions.
On this page
MRO practices today
For airlines, maintenance is a major cost factor in the maintenance, repair, and overhaul (MRO) function.
This is a big challenge over and above headwinds such as fluctuating fuel costs, geopolitical instability, global supply chain crisis, and shortage of skilled workforce. This can be overcome by adopting predictive maintenance (PdM) strategies that help MROs improve performance and productivity.
A lack of efficient planning that results in over or understocking of parts is one of the main causes for the increase in operating costs for MRO. While overstocking leads to non-moving inventory, understocking results in AOG (aircraft on ground), a strain on supply chain, unscheduled grounding, and flight cancellations. Costly components provisioned based on recommended spare parts list (RSPL) during initial provisioning for new aircraft usually end up as non-moving inventory.
Moreover, since maintenance is carried out as per scheduled tasks at time intervals, hours, or cycles, aircraft components may get replaced before they fail, or their remaining useful life (RUL) gets over. This causes wastage as resources are spent on components that don’t require replacement or maintenance at that point of time. But any failure of components between scheduled maintenance could result in AOG situations.
By adopting a predictive maintenance psychology, airlines can minimize the possibility of a potential component or system failure and retain operational efficiency. Predictive maintenance involves continuous tracking of components’ health and assessing failure trends using AI-ML algorithms and interpreting the data for faster decision-making.
Predict, preempt component failure
Predictive maintenance (PdM) strategies provide airline operators with a well-calculated estimate of component health.
They help assess the possible RUL in a component life cycle and the tentative time frames when maintenance will be required. This helps predict and preempt component failures.
In aircraft servicing, scheduled maintenance can either be driven by maintenance tasks from approved maintenance program (AMP), based on hours, cycles, days, and elapsed time, or a health assessment of the components and systems. In predictive strategies, the health of systems and components drive maintenance.
The goal of PdM is to predict a potential failure of on-board components and systems. This can be done by collecting and analyzing various onboard sensor data and historical data related to maintenance, repair, and replacement, to retain systems and components in operational condition on the best possible scale of accuracy. PdM continuously tracks the conditions and assesses the failure trends through IoT, AI-ML, and data analysis, while complying with airworthiness regulations and quality control requirements.
Predictive maintenance strategies
MRO can adopt data-led PdM strategies to gauge the variation in performance of components and systems from normal, well before any breakdown occurs.
AL-ML algorithms can analyze the various supervised and unsupervised data generated from on-board sensors, along with past maintenance information and experiences, and current maintenance plans.
PdM models can deploy techniques such as trend analysis, vibration monitoring, anomaly detection, and validation with various failure models. Powerful data analytics and pattern-detection capabilities of AI-ML algorithms can analyze real-time and historical data to perform in-depth analysis of components’ inflight performances and derive useful maintenance tips.
Research indicates the possibility of adopting reinforcement learning in PdM models for optimizing scheduling of maintenance tasks and producing effective maintenance plans. Static algorithms can be deployed for long-term scheduling, while adaptive algorithms can be used for rescheduling based on new maintenance information such as new faults identified and RUL at task level.
Gaining confidence with stakeholders’ support
PdM can be imagined as a maintenance culture, which is evolving over time.
It ensures maintenance is performed only when required, helping facilities to cut cost, save time, and maximize resources.
While there are obvious advantages of PdM over preventive maintenance, stakeholders will have to address some inherent challenges. There is a lack of confidence in the maintenance tips that PdM provides through health assessment, vis-à-vis the confidence built over years in a well-proven scheduled maintenance process based on hours and cycles.
Healthy PdM practices and experiences and active support from relevant stakeholders over time will improve confidence. One area where PdM is already gaining confidence is in engine health monitoring (EHM). PdM is quite effective in monitoring and assessing aircraft engines for performance and predicting and managing their maintenance requirements. As this aspect of engine maintenance evolves, it will provide confidence over time to build suitable predictive maintenance models for other aircraft systems and components. These models will have to be complemented with trained personnel with the right skill sets.
PdM strategies require basically two sets of skills—domain expertise to analyze and interpret huge sets of data to make meaningful suggestions, and the ability to create effective algorithms to provide meaningful predictive inputs for maintenance.
The skill sets required to train PdM models can evolve with active adoption and analysis of data using previous maintenance experiences.
Data sharing through consortium
Every moment of an aircraft’s operational life generates a huge chunk of data.
The data pertains to flight routes, flight schedules, passenger information, and operational behaviour of components on the aircraft.
As data is often considered a valuable commodity, MRO may be hesitant to share theirs with aircraft manufacturers and component OEMs or vice versa. But sharing this data will provide mutual benefits to the parties concerned. The Independent Data Consortium for Aviation (IDCA) acts as a platform for OEMs, operators, and MROs to share data related to maintenance and prognostics such as technical problems common to all operators and information on parts from back to birth. The aim of the consortium is to make data immediately available for the diagnosis of fleet-wide problems, develop mutually beneficial analytics, and even help create opportunities to develop new data analytical tools. Sharing data on common operational problems related to equipment among relevant stakeholders will help PdM strategies.
The IDCA platform uses blockchain for sharing data securely. But unlike other blockchain consortia, it is decentralized and governed by paid members.
As stakeholder cooperation evolves through platforms such as IDCA, PdM strategies for MRO will start gaining more traction. Onboarding regulatory authorities and aircraft, engine, and component manufacturers is essential for a foolproof maintenance psychology based on PdM.
