IoT and machine learning ensure reliable, available aircraft at a moment’s notice

Lockheed Martin transforms aircraft maintenance and fleet management with SAS analytics and AI

With an innovative history spanning a century, Lockheed Martin is a leader in the aerospace industry, building and maintaining aircraft for researchers, non-government agencies, municipalities and commercial entities worldwide. Each aircraft is meticulously designed for a unique purpose, whether it’s lifting heavy cargo from a sinking ship, supporting cutting-edge scientific endeavors at extreme altitudes, or traveling at supersonic speeds. Despite their remarkable capabilities, they have a common problem … parts fail over time.

From the highest landing strip in the Himalayas to the roughest, most remote dirt airstrip, the C-130J Super Hercules has proven to go where other transport aircraft can’t, don’t and won’t go. As a dependable workhorse, the C-130J is the aircraft of choice for 28 operators across 23 nations around the globe.

When Lockheed Martin began producing the C-130J in the late 1990s, it included a limited number of sensors that collected about 30MB of data per flight hour. Today each aircraft features more than 600 sensors generating upwards of 3GB of data every hour. Using models based on SAS analytics and AI solutions, Lockheed Martin ensures that data is used to revolutionize aircraft maintenance and performance so its airlifters can always be “ahead of ready.” One of the most important outcomes of that use of data is the ability to predict when parts will fail, contributing to improved availability and reliability of the fleet at a moment’s notice.

Improving aircraft readiness for global research and humanitarian missions

Countries around the world rely on the C-130J for:

• Around-the-clock maritime search and rescue.
• Oil spill monitoring and cleanup.
• Coastline and ice sheet mapping.
• Monitoring of natural disasters and atmospheric phenomena.
• Providing humanitarian relief.

“The relief efforts are what I’m most proud of,” says Mike Isbill, a Lockheed Martin Fellow specializing in data analytics. “After tornadoes, hurricanes or earthquakes, C-130s quickly bring emergency supplies and aid. They can operate in the most austere conditions, landing in fields or on damaged runways.”

Isbill’s team uses data and machine learning to make Lockheed Martin’s aircraft as reliable and cost-effective as possible. When a C-130J sits on the ground waiting for an ordered part to arrive, that downtime can slow delivery of vital aid or hinder the effectiveness of humanitarian relief operations. By predicting failure before it occurs, the part can be available when and where it’s needed.

Sensors on the C-130J detect and record about 6,000 data elements up to 10 times every second. Data elements range from vibrations and temperature readings to the location of a part on the aircraft.

“By knowing everything happening on the aircraft when a failure occurs, we can build stronger anomaly detection algorithms with SAS Viya,” Isbill says. “The machine learning and AI algorithms help us see into the future — to see when things start to degrade. We can say with better accuracy, this part is going to fail in the next 50, 30 or 20 hours. Customers can plan for maintenance downtime prior to failure, especially before a critical deployment.”

Enhancing efficiency and cost savings

Lockheed Martin began using SAS in 2014 to better manage and perform C-130J maintenance.

“We’ve gone from writing on a whiteboard, taking notes and trying to figure out in our heads what the information was telling us, to being able to draw so many insights that weren’t there before,” Isbill says. “Now I can take the data from 2.5 million flight hours for the C-130J and I can slice it, dice it and combine it in ways and at speeds never possible before.”

With the cloud-based SAS Viya platform, Isbill and his team can expand and contract memory as needed, even when running huge simulations.

“We’ve seen some amazing benefits in using SAS Viya in all our processes, including time savings and productivity gains,” Isbill says. “The biggest for me was in cleaning and aggregating the data. The algorithms we’ve developed clean the data for us when it comes in, taking only about 30 seconds, and about 85% of that is done automatically. That alone saves us 40 hours a month.”

Analytics with SAS Viya

Using SAS, Isbill’s team developed a menu of analytics services for Lockheed Martin customers. A fleet management service uses in-depth analyses of maintenance activity, sensor data, flight profile and repair data to help operations and maintenance teams make faster, more reliable decisions to maximize aircraft availability.

“The customer can go into the fleet management dashboards prior to flight to diagnose the aircraft and monitor the health of their fleet,” Isbill says. “With insights about which parts in an aircraft might fail, they can make better decisions about which aircraft to use and which spare parts to have with them on the trip”

Harnessing the Internet of Things (IoT), Lockheed’s team is working to use SAS Event Stream Processing to reside on and communicate with the aircraft. It will stream and read data in real time to alert the maintenance team on the ground if any issues exist. Tools and parts can be ready before the airplane lands. Using a tablet, the flight crew can review the health of any system on the aircraft at any time. If they see anomalies in the data for a critical system during flight, they can decide if they need to abort the mission.

The event-stream processing unit also feeds information to Lockheed Martin’s customer service center, which monitors the health of its fleet across the world. Data trends from flights add to the knowledge base so they can help other customers troubleshoot problems and prevent failures.

Isbill is especially pleased with the security of the IoT solution that offers a “disconnected edge” option for ground crews who work on secure laptops not connected to data in the cloud. “The crew can gather information, run SAS Event Stream Processing in real time, do the analytics as they’re flying, and then we can securely send that back through a secure 5G network connection,” he says.

Intelligent diagnostics via machine learning optimize troubleshooting in real time

With aircraft data streaming in, Isbill and his team developed the Intelligent Diagnostics tool that applies machine learning and IoT analytics to C-130J flight data. The system combines that data with system knowledge from Lockheed Martin engineers and parts vendors to form a central repository on more than 300 aircraft parts. It tracks all the decisions and related events associated with repairing a specific part and learns from them.

Intelligent Diagnostics can recommend troubleshooting steps, repairs or replacements in similar situations – creating real-time best practices for aircraft maintenance. For example, the Intelligent Diagnostics service lowers no-fault-found rates, a particular area of aircraft downtime and added expense.

Lockheed Martin further reduces downtime by applying survival analysis to a broad family of line-replaceable units for spare parts optimization.

Taking the correct set of spare parts into a hostile environment, such as a wildfire zone or to evacuate refugees, is crucial. For example, the company worked with one of its largest C-130J operators to track 20 aircraft and 50 parts over six months. The predictive maintenance models forecasted a 2,000-hour reduction in downtime. That’s 2,000 hours delivering fire suppression during peak summer wildfires or desperately needed food and water to starving refugees — not sitting in a hangar waiting for a part.

“That’s a 2.6% increase in mission-capability rate,” Isbill says. “That might not sound like a lot, but moving the needle on mission-capability rate is usually done in increments of 0.1 or 0.2%.”

Optimizing part supplies improves the readiness of spares for existing customers as well as how the company sells spares to new customers and deploys parts around the world for faster distribution.

Redefining aircraft analytics

Isbill’s team plans to increase its use of event-stream processing and intelligent diagnostics to turn that data into more valuable insights.

“I’m excited about the future of our partnership with SAS,” Isbill says. “With analytics behind the data, we’re going to give the maintenance personnel on the ground information and insights they’ve never had before. They can keep pilots in the air longer, and aircraft won’t have to be on the ground any second longer than necessary.”