A computer that thinks for itself presents an extremely polarizing reality. Some see an all-encompassing solution, while others fear risks to security, whether it be for their job or their information. This is increasingly true in the world of nondestructive evaluation (NDE).
NDE is a multidisciplinary approach to inspecting and evaluating materials, components or assemblies without destroying their serviceability. NDE professionals around the world play critical roles in quality assurance and material science when it comes to essential infrastructure such as pipelines, nuclear reactors, submarines, bridges, manufacturing assembly lines and much more. So, the idea that such work could be reimagined with artificial intelligence and machine learning can be exciting, daunting, concerning or almost frightening, depending on who you ask. Yet, no matter the feelings behind it, the shift is occurring, and it has a name - NDE 4.0.
Evolving into NDE 4.0
To fully understand the emergence of what many consider entirely new technologies, it’s important to know there is history behind them. In 1957, Dr. Frank Rosenblatt, a research psychologist and project engineer at the Cornell Aeronautical Laboratory, programmed a computer to distinguish cards marked on the left from those marked on the right. Though slow-moving and seemingly limited in function, this pioneering work on neural networks led to the development of advanced machines that can learn from experience and specifically from sets of data. These more complex processes are now known as artificial intelligence (AI) and machine learning (ML), respectively.
With a computer that can think through certain operations comes an opportunity to redistribute such operations. Over the past ten years, this concept of a “natural transition of tasks” has taken hold in NDE. In the same way that Rosenblatt programmed a computer to operate like a human brain in its ability to sort cards, the nondestructive evaluation industry began allocating tedious tasks – and the time and learning required to complete them – to computers. This evolution, known as NDE 4.0, revolutionizes the way we take in, evaluate, and infer from large data sets. With AI and ML at our disposal, NDE 4.0 represents a more dynamic future for the industry.
Unfortunately, a haze of misunderstanding around the incorporation of NDE 4.0 leads to hesitation, concerns, and overall stagnation in adoption. To help make the most of this natural transition in our industry, let’s address some common misconceptions.
The future of the NDE workforce.
NDT
ASNT Columns
It’s a concern heard across the world: what happens to my job if a computer can do it? While this worry is a reality in many professions, the role of ML and AI in NDE 4.0 is not to replace practitioners, but to help them do their jobs better. At the end of the day, AI can help us analyze and think through data, but people make the decisions about how to act on it.
Rather than removing positions from the industry, the goal of NDE 4.0 is to amplify the intelligence of the individuals doing the job, allowing them to expand their capabilities, especially when it comes to speed. For example, speed becomes especially helpful with acoustic emission (AE) testing. This NDE process measures the transient elastic waves that result from microdisplacements, or flaws, in a material. To do this, a load must be applied for these waves to be generated. High-speed computers and improved software make it possible to closely examine acoustic emission waveforms and perform evaluations with greater accuracy. For instance, if you’re looking at 1,000 gigs of data from an acoustic emission hardware running at 10 million times a second, AI can evaluate and summarize a data set that would be impossible to review by sight at all, let alone in a comparable amount of time. As a result, issues in the materials can be found, and addressed, much faster.
NDE 4.0 also delivers a next-level capability when it comes to pattern recognition, far advancing our inferencing ability. That said, the practitioner and the organization having a learned understanding of the technology, and how it can best be employed, is a critical component to optimal usage. After returning from field inspections, a practitioner’s turnaround time can be cut from 72 hours to as low as 24—all depending upon how well-trained the staff is and how prepared operations are to comply with AI and ML technologies.
An integrated solution.
On the flip side of the concerns over job security is an erroneous expectation that needs to be greatly tempered: that these systems will complete and fix everything with the click of a button.
While AI and ML technologies have the capacity to propel NDE operations to a whole new level, it is not a magic one-stop shop. A business can only be revolutionized by this type of evolution if the foundation – the team, the procedures, and most critically, the quality of the data sets – is solid to begin with. If you have bad data, it doesn’t matter if you evaluate it or a computer does, the result will be the same. AI merely allows us to get to that result faster.
A balance between security and advancement.
When technology and sensitive data are involved, concerns about security are bound to emerge. Even with bad actors to be wary of, this intense caution often greatly overshadows the potential advancements that could come with AI and ML, to the detriment of the industry.
For example, a common framework for data sharing may set off alarm bells for many. As it is, we are stuck in a seemingly endless loop. Despite the large potential benefits that access to open-source data sets may bring to everyone, there is widespread hesitation to share the data from an experiment because of the perceived financial implications that come with giving up private ownership. Yet, with data sharing backed by proper security measures, the capabilities of AI and ML in NDE would help advance the operations of all involved.
