AI and Reliability: How Much, How Fast?

By Dr. Klaus M. Blache, Univ. of Tennessee Reliability & Maintainability Center (RMC)

Interest in Artificial Intelligence (AI) has been gaining speed due to big data, the cloud, increased computing power, greater connectivity, and advancements in sensors and signal processing. Machine perception, i.e., using cameras to recognize objects, has been around for years. Today’s machine-learning systems can do much more.

For purposes of this column, the term “machine learning” refers to computer algorithms that improve with experience. Think of enhanced speech and facial-recognition technology, Tesla autonomous vehicles, IBM Watson, and reliability and predictive maintenance (PdM) modeling efforts. Another example comes from Lufthansa Airlines, which maintains more than 1,000 planes and is using machine learning in real-time data collection and decision making. Recommended failure-avoidance actions come as a result of error messages and sensor data, among other things.

Less than two years ago, I visited Lufthansa Technik (airplane maintenance, production, and development) near Hamburg, Germany, where they were still mostly “talking” about machine learning. Now, these operations are leveraging it to reduce downtime, component failures, and cost.

Around the same time, my study on evaluating the state of reliability-related modeling in North America found that fewer than one in 100 companies were sufficiently using the analytical tools to gain competitive advantage. Then again, some companies that are technically capable often don’t have sufficiently granular or quality data. I see this situation changing in the next 5+ years in all parts of industry and government.

Yes, companies are currently setting up “virtual twins” to emulate their critical assets. Visual models and algorithms are being used to simulate manufacturing capabilities and predict failures and when to do maintenance repairs or replacements, as well as for product performance/customer experience. At the same time, the world-renowned scientist Stephen Hawking has expressed his concern to BBC News that “artificial intelligence could spell the end for the human race.” Maybe, but not in the near future. Risk-versus-progress decisions will always be with us.

Progress in AI is inevitable. Thus, it’s better to embrace it, understand it, use it to improve, and be part of controlling it to the extent necessary. Today, AI is still very specific or applied, i.e., controlling drones and cars. AI that can do anything, or most general tasks, is still only in the movies. The time to be concerned is when scientists figure out how to use a neural network, like the human brain.

What we need to improve immediately is the ability to apply big data and analytics to practical applications for daily use. We should also take the opportunity we have to engage Millennials who like to be in this space.

If you have examples of successful machine learning, I would like to hear from you. EP

Based in Knoxville, Klaus M. Blache is director of the Reliability & Maintainability Center at the Univ. of Tennessee, and a research professor in the College of Engineering. Contact him at kblache@utk.edu.