Ru

News and Events

Artificial Intelligence has started changing non-destructive testing (NDT)

December 4, 2018,10:00 am CEST

In recent years, the use of Machine Learning and Artificial Intelligence algorithms has exploded across different fields — from beating humans at board games, to trying to predict consumer behavior across channels, to identifying abnormal operating conditions of industrial equipment, just to name a few prominent examples.

We sat down with Head of Product Management, Isaak Tsalicoglou, and Co-Head of Technology, Cengiz Küpçü, to discuss how Artificial Intelligence plays a role in shaping the future of NDT, especially through products recently launched by Proceq.

There’s a lot of talk about Artificial Intelligence these days — why is that so?

Isaak Tsalicoglou (IT): The answer is “data”. As data sets grow in size and complexity, so does the effort in managing that data. And so does also the cognitive load of interpreting data in a timely and useful manner —ideally even on-site, which has so far been a challenge. Machine Learning and Artificial Intelligence help to generate and reuse insights that augment our human intelligence. This way, we can get things done easier, faster, and more reliably.

And how does this relate to the non-destructive testing industry?

Cengiz Küpçü (CK): As sensors become more advanced and high-performing, the data that they generate also get richer. Furthermore, when more than one sensor technology is involved in assessing the quality of an asset, even more data are added to the mix. As a result, data sets become more “dense” and more multi-dimensional. We must facilitate that this data flood can be absorbed and understood. One side effect of the difficulty of interpreting large data sets is that it also takes a lot of time to do so.

IT: I totally agree, Cengiz, and in practice it’s also about reaching a robust conclusion in a timely manner, ideally without touching your laptop or going back to the office. After all, a measurement plan by itself is worth little, and data by itself do not “speak”. From start to finish, an NDT application is a process, and it needs to be kept short, productive, and collaborative. It delivers value to customers and stakeholders primarily when a verdict is reached — meaning: is the asset OK? If not, why not? If so, what should be done? In a way, it’s all about helping NDT inspectors help their stakeholders make better decisions, faster…

CK: …and this is where Artificial Intelligence can assist the world, by generating better, faster, stronger insights!

How are those ideas reflected in the Proceq product portfolio?

CK: Proceq has been a pioneer of Artificial Intelligence in NDT! Artificial Intelligence already found its way into Proceq products in 2016, with Profometer 6 AI. And, more recently, especially Pundit Live Array Pro brought a lot more Artificial Intelligence into daily use.

How does it work, what does AI actually do for the users of Proceq equipment?

IT: In Pundit Live Array Pro, AI is currently applied in two functions. The first one is that the Pundit Array transducer head can be moved anywhere along a line. Artificial Intelligence assists in effectively tracking the transducer head and stitching the data records together seamlessly. This means that less care is needed during the recording — measurement speed thus increases. The overall productivity increases, and the cost per inspection is reduced. As an additional result, the data images are generated at a higher quality than possible in the manual mode. In turn, this enhanced quality makes the image easier to interpret. Hence, in this implementation Artificial Intelligence helps both directly, during the measurement, and indirectly, during its analysis.

CK: The other function where Artificial Intelligence helps in Pundit Live Array Pro is to automate manual efforts in the analysis phase, i.e. during data interpretation. In the past, when you needed to determine the thickness of a concrete slab, you had to manually tag the back wall in the image every so many centimeters, or inches. For the human eye and mind, this required a lot of attention and diligence — in other words, it was operator-dependent. We now introduced an AI-assisted tagging feature for back walls. You just tap the “magic” AI button, and our iPad app recommends tags based on an algorithm that is typically more accurate than the human eye, and certainly faster than manual tagging! Nevertheless, the user still has to check, confirm and make sense of the tags and images based on her or his understanding of the real structure that is being inspected.

Does this mean that you envision Artificial Intelligence ever taking over the task of interpreting NDT data?

IT: Our Artificial Intelligence features never take over control from you, the user. They enable you to get things done better and faster, they assist you — but they do not hijack your user experience. This is a subtle and important distinction, which shows the user-centric focus of Proceq Live products, even with something as revolutionary as Artificial Intelligence.

CK: Machine Learning and Artificial Intelligence have both existed for decades. However, computational capacity and access to algorithms has grown massively only in recent years. This means that, gradually, Artificial Intelligence algorithms for computer vision are also being used to generate insights from NDT images — such as what Proceq tomography equipment nowadays generates; for example, also in Augmented Reality.

IT: It seems to me that the future of NDT holds as many surprises as it holds certainties, partially thanks to Artificial Intelligence. It is part of our company vision to help our users put material testing equipment to use with confidence, with increasing value — and surely also with some fun, why not!

What does the future hold for Artificial Intelligence in Proceq products?

CK: On our end at Proceq, we are committed to defining the future of NDT with cutting-edge technology implementations — both in sensors and software, including Artificial Intelligence. Our products deliver sensor data streams in a way that people can understand them, collaborate on them with others, and reach conclusions.

IT: And let’s not forget that no single technology is perfect for all applications. This is why at Proceq we are also uniquely committed to a multi-technology approach, i.e. making available “the right combination of instruments for the job”. We therefore also assist in combining those sensor data streams into a meaningful holistic approach for the quality assurance of infrastructure and assets. Artificial Intelligence will play an increasing role in doing exactly that with Proceq Live products, especially as our technology and product portfolio expands further.

This all makes a lot of sense indeed. So, to get there faster, what is the biggest potential obstacle that you foresee?

CK: The availability of talent! Proceq is recruiting new talents across all functions, and especially in R&D. Whilst there are recruiting platforms to support this process, nowadays even with Artificial Intelligence, it is the proactive current and future engineers who will create the new solutions. At Proceq, young professionals and talents like Isaak and many of my team members can have a unique challenge in Proceq if they contribute commitment, creativity, and, well — human intelligence!