Airplane Icon

AI-BASED PREDICTIVE MAINTENANCE SOLUTION AND SOFTWARE INTERFACE FOR IMPROVED NAVAL OPERATIONS AND AVAILABILITY

Industry Icon

Industry

Aviation

Airplane Icon

Asset

Aircraft health monitoring systems

Goal Icon

Goal

Develop a predictive maintenance solution for the early detection and diagnosis of engines and other key subsystems in naval aircrafts

Impact Icon

Impact

Created a successful solution and software interface to predict aircraft maintenance issues before they occur and provide quicker and more accurate diagnosis, lowering maintenance costs, improving reliability, and increasing fleet availability

Overview

Predictronics received a Small Business Innovation Research award under the U.S. Navy Technology Acceleration program to develop machine learning (ML) and artificial intelligence (AI) capabilities to impact mission success (Navy SBIR 2019.3 - Topic N193-A01). This program supports the development of innovative artificial intelligence (AI) and machine learning (ML) capabilities to address a variety of use cases, including Predictronics' focus area of predictive maintenance.


Solution

Our approach for this project was to create an AI-based predictive solution which mimics the continuous learning capabilities of a human. These capabilities were validated with three public data sets, two on aircraft systems and one on gearbox condition monitoring. Our team also developed an intuitive and user-friendly software interface, allowing human operators to label new patterns as data accumulates. This labelling helps to fine-tune our models, leading to greater accuracy and less human effort in the future.

Value

Many other predictive solutions on the market require models to be manually updated on a regular basis, experience a decrease in accuracy over time, or have issues with learning new normal modes of operation and diagnosing unforeseen problems. The predictive maintenance solution developed by Predictronics eliminates these limitations by providing fault diagnosis through the use of our multiple anomaly detection models coupled with our human-in-the-loop labeling and continuous learning from new anomalous patterns. This solution, enabled by our PDX platform, successfully delivered near perfect diagnosis results in phase one, confirming the feasibility of our algorithms and approach.

In June 2020, Predictronics was the recipient of a phase two SBIR award from the U.S. Navy. This will allow us to continue the development of our solution and create a working prototype. The deployment of our solution following phase two has the potential to ensure fast and accurate fault predictions for naval aircraft systems. These data insights would lower maintenance costs, optimize fleet availability, and increase operational efficiency. Predictronics also hopes to expand the solution capabilities, providing labels for anomalous patterns from maintenance and inspection records, automating fault detection to reduce the need for human effort, and optimizing the settings and parameters used in our analysis methodology. Our solution can be applied to numerous applications outside of aircraft systems, such as manufacturing, ground transportation, and energy.