How the US Navy will save billions of dollars using Google Cloud AI to fight corrosion

About US Department of the Navy

The US Department of the Navy has a vision to relentlessly innovate, adapt, and cultivate a culture of continuous learning and professional development. Its mission is to deliver combat-ready naval forces to win conflicts and wars while maintaining security and deterrence.

Industries: Government & Public Sector
Location: United States

Tell us your challenge. We're here to help.

Contact us

About Simple Technology Solutions

STS specializes in multicloud engineering and applied AI/ML and delivers automated, scalable, and secure mission solutions to federal agencies.

Google Cloud and Simple Technology Solutions (STS) rapidly built an AI-based corrosion-detection and analysis system for the US Navy, which will eventually be used to automate inspections of vessels, aircraft, and vehicles—saving billions of dollars.

Google Cloud results

  • Detected and analyzed corrosion on vessels with 90% accuracy. Currently expanding to other platforms, including aircraft and vehicles
  • Placing cutting-edge new tools into the hands of Navy and Marine Corps workers to stem the tide of corrosion
  • Projecting billions of dollars will be saved across the US federal government
  • If successful, AI-based corrosion detection and analysis system has potential to support other groups beyond the federal government

In six months, AI system moves from pilot to production

The US Department of Defense (DoD) spends more than $19 billion annually to manually inspect and repair its fleets of aircraft, submarines, tanks, and other equipment for corrosion—a critical aspect of maintaining resiliency and naval wartime readiness. The US Navy alone spends approximately $7 billion per year.

The Navy’s Naval Enterprise Sustainment Technology Team (NESTT), a grassroots organization made up of “sustainment” experts, teamed with the Navy’s Small Business Innovative Research (SBIR) program and the Office of Naval Research (ONR) to develop an agile and scalable approach to automating fleet management inspection for ships, aircraft, and vehicles.

“Our ultimate goal is to productionize this system and bring it to market. For many reasons—speed-to value, cost savings, the richness of its native AI capabilities—Google Cloud was the obvious choice.”

Betsey Hutton, principal at STS

ONR awarded a phase I contract to STS to explore the viability of an artificial intelligence (AI)-based corrosion-inspection system to drastically reduce the labor burden and safety risks associated with these maintenance operations.

After choosing Google Cloud as the platform—including Cloud Storage, Google Cloud AI, and AutoML—the team successfully completed the phase I proof of technical viability in record time and is now ready to take the next big step: operationalizing the system.

Phase II will get the system out into the field, not just in the Navy, but with other partners in the DOD and the US Department of Transportation (DOT). Eventually, commercialization of the system will provide benefit well beyond the US federal government—helping industries around the globe address corrosion.

“Although in phase I we were only trying to prove viability of the technology, we’re always thinking long term,” says Hutton. “Our ultimate goal is to productionize this system and bring it to market. For many reasons—speed-to-market, cost savings, the richness of its native AI capabilities—Google Cloud was the obvious choice.”

“We can do research forever, but it doesn’t make any difference unless we can actually bring it to production, and fast,” agrees McKee. “I can’t accentuate enough our absolute drive for speed, which the STS and Google Cloud partnership is enabling.”

“In the end, it was about speed to production to support our mission, and Google Cloud and STS are showing how fast this can be done.”

Steve McKee, lead, NESTT, US Department of the Navy

Addressing a mission-critical challenge

Corrosion is a universal problem. Any organization, public or private, possessing metal equipment, machinery, or infrastructure, needs to continuously inspect, maintain, repair, and replace it to ensure maximum operational efficiency as well as safety.

“The Navy exerts a tremendous amount of effort inspecting for corrosion to make sure that we have structurally sound platforms—ships, airframes, vehicles, and so forth—so they can do what they need to do,” says McKee.

Thousands of people are employed on a daily basis to do those inspections, look at the results, and judge the efficacy of those platforms. “By shortening the timeframe for inspections, we recognize the potential to save billions of dollars,” says McKee.

A key aspect of the proposed system was that it absolutely had to be platform agnostic. Another critical prerequisite was speed. NESTT was adamant that the system be developed at an accelerated pace to address the urgency of the problem.

The work was originally awarded to STS as a phase I SBIR project, and Google Cloud was chosen as the technology platform.

“The project’s genesis was that it was not only for ships, not only for aircraft, not only for vehicles, but a system that truly transcended all of those platforms,” says McKee. The Google platform fit the platform-agnostic requirement. “And in the end, it was about speed to production, and Google and STS are showing how fast this can be done,” says McKee.

Accelerating maintenance inspections with AI

For phase I, STS leveraged public-domain and drone images of commercial vessels to build a high-quality AI model using AutoML, a simple, secure, and flexible machine learning (ML) service that allows organizations to build custom vision models.

STS collaborated with the US Navy corrosion subject matter experts to label and train the data using the AI Platform Data Labeling Service—feeding in more than 6,000 images. It then iteratively trained and validated the model using custom inspection drone flight data, which was uploaded using Cloud Storage for processing. The model was designed to continually improve and update itself based on the most recently ingested data.

“This is about automation, saving time and money, and keeping inspectors out of harm’s way,” said Aaron Kilinski, chief technology officer, STS. “The initial goal for phase I is to build a model that detects and analyzes corrosion in drone images with a very high degree of accuracy. The ultimate goal, however, is to move from detection to prediction by expanding the subjects and sensors and eventually integrating with Navy systems. We selected Google Cloud AutoML because it allows our engineers to train and test high-quality models quickly. Google Cloud provides an unrivaled degree of specification to meet tough business objectives in compliance with FedRAMP High.”

According to Hutton, three basic reasons drove the decision to choose Google Cloud: speed to market, cost savings, and capability. With any other vendor, it would have been necessary to employ data scientists and mathematicians to build an object-detection algorithm from scratch. “But with Google AutoML, we got to skip that entire step,” says Hutton. “We started out with an 80% solution that just needed to be tailored to our use case.”

Cost effectiveness was the second reason for choosing Google Cloud. “Our ability to leverage the public cloud, commercial hardware, and commercial drones contributed immensely to that overall value,” says Hutton.

And the third reason was the superiority of the technology. Out of the gate, the team asked, what technology is going to provide the most portability? What's going to provide the best storage capacity? “We thought through all these elements. And that’s why we went with the Google Cloud technology stack,” says Hutton.

Phase II and vision for the future for the US Navy, DOD, and beyond

During phase II, the team will expand the model’s corrosion detection and analysis capabilities to include other platforms and sensors, which will require tens of thousands of additional training images. They will develop an intuitive user interface so inspectors can easily use the solution in the field. And they will introduce more advanced modeling capabilities to help drive adoption and business value for maintenance workers.

In addition to more commercial and Navy data, Phase II will also leverage data and images from the US Air Force and DoT to ensure maximum applicability. One of several new features during phase II includes adding a classification scoring mechanism to the algorithm, to grade the severity of corrosion on a scale of one to five. “That will help maintenance workers prioritize their workloads,” says Hutton. Further, phase II brings a user-friendly frontend interface that includes 3D modeling capabilities. This will tell maintenance workers the precise location of corrosion on a vessel or piece of equipment. “The system could find an instance of level 5 corrosion—which is a big deal—but if you can’t pinpoint where it is on a massive ship, that information is not going to help much,” says Hutton.

The team is additionally exploring using other sensor types as data input devices. “We want this to be applicable to the many, many platforms the Navy possesses,” says Hutton. Each platform has different maintenance inspection criteria. So to expand, for example, to inspecting aircraft, the system would need to be able to detect non-visual maintenance indicators using infrared sensor technology.

“We’ll get ultrasonic test data as well, so it won’t just be hyperspectral. You’ll get lots of other types of input that we can train the algorithm on,” says McKee.

The DoD’s cross-department Joint Robotics Organization for Building Organic Technologies (JROBOT) is one organization keeping tabs on progress. NESTT is also partnering with the Advanced Robotics for Manufacturing (ARM) Institute, in which the DoD has already invested more than $80 million. Across the DoD, the Army and the Air Force are also participating. “We think that the STS-Google solution will truly benefit not just the Navy, but ultimately the larger swath of the defense industrial base and allied commercial partners,” says McKee, who says the value of this solution is astronomical when you consider the possible applications.

Lastly, the benefits of this solution could positively impact the private sector and specific commercial industries. “The ability to use this model to determine corrosion or no corrosion is a game changer,” says McKee. “Not just for the DoD, but for maritime shipping, the aviation industry, mining, and more.”

Tell us your challenge. We're here to help.

Contact us

About US Department of the Navy

The US Department of the Navy has a vision to relentlessly innovate, adapt, and cultivate a culture of continuous learning and professional development. Its mission is to deliver combat-ready naval forces to win conflicts and wars while maintaining security and deterrence.

Industries: Government & Public Sector
Location: United States

About Simple Technology Solutions

STS specializes in multicloud engineering and applied AI/ML and delivers automated, scalable, and secure mission solutions to federal agencies.