Testing, inspection, and certification (TIC) of equipment and machinery is an essential, if underappreciated, part of the interconnected industries that make modern living standards possible. The TIC industry acts as an important intermediary between private and public entities, and is a rapidly growing sector with a market size forecasted to balloon from $199B in 2020 to $272B in 2022. Our safety hinges on shared industry standards and the dedication and constant vigilance of machinery and equipment experts that ensure stringent TIC standards are met.
Machinery and equipment inspection is not a check-the-box activity. Inspections help with detecting issues early on when they are easier to fix, and avoiding the costs and dangers associated with breakdowns that require major repairs or result in equipment failure. Despite current best efforts, machinery and equipment does routinely fail. A recent study looking at just the world’s 72 largest manufacturers found that almost $1T is lost each year to machine failures.
Inspection procedures often involve routine, repetitive tasks that are time consuming and distracting for operators in the field. Although artificial intelligence (AI) can’t (yet) solve the issue of equipment failure, it can automate processes that previously required human workers—saving time, freeing up resources, and providing more accurate results. This article explains how AI can be used to automate an important yet seemingly simple piece of the equipment inspection process: nameplate extraction.
Every equipment or machine test, inspection, or certification, requires capturing specific information about the machine or piece of equipment in question. This identifying information, such as the model number or technical specifications, is found on the nameplate. Nameplate extraction is the process of copying pertinent details from physical machine and equipment nameplates to a centralized repository.
Extracting the right information, such as serial and model number, is typically done from a photograph of the nameplate, taken at the time of inspection. The process of reading each photograph to identify the correct numbers, then correctly entering them into the system, is time-consuming, error-prone, manual, and mundane work.
And it’s work being done by highly-skilled inspectors and engineers, taking them away from the real work at hand: the inspections, tests, and certification assessments and reports.
Something as mundane and repetitive as nameplate extraction is an ideal candidate for intelligent automation—yet it remains a manual process for most providers of TIC services. In the words of the Industrial Analytics Platform (IAP), “the Fourth Industrial Revolution is more than a technological leap forward”, rather it is a new era where communication between all materials and people is redefining the possibilities of industry and manufacturing.
The potential of this great technological leap forward stands in stark contrast to our case of nameplate data extraction, which represents part of a critical process to ensure equipment and machine safety, and yet remains out of reach of basic business process automation. Why?
Nameplate data primarily exists as unstructured image data (photographs taken at the time of inspection). Unstructured data, or data that doesn’t follow a pre-defined model or schema, is far more difficult to process and analyze than structured data—which off-the-shelf automation tools like robotic process automation (RPA) rely on.
Extracting data from images is a perfect use case for artificial intelligence (AI). But that doesn’t mean it’s an accessible solution. In the past, the manufacturing industry enthusiastically initiated AI projects to realize the potential of digital transformation only to find themselves reassessing the feasibility of applying AI, primarily due to underestimating the time and resources required to create the data infrastructure that underpins AI models.
For nameplate extraction, we can see the challenges to building an automated process, starting with building a dataset that requires manual transcription and labeling nameplate image data by highly-skilled staff. And if and when you reach the point where your data is ready, sourcing talent or outsourcing the AI build project means a massive investment in both time and costs.
There is good news, an “off-the-shelf” AI solution for nameplate data extraction does exist. Super.Image, super.AI’s automated image processing solution, represents the future of unstructured data processing by taking an assembly line approach to artificial intelligence.
To build a training dataset, super.AI allows any combination of AI, bot, and human workers to collaborate. Simply upload images (you can even drag and drop image files), define the data fields for extraction, and let super.AI’s Unstructured Data Processing (UDP) Platform do the rest. Experts can be involved during training to supervise dataset labeling and ensure output quality. That’s it. Super.Image turns automated nameplate data extraction into a streamlined process that can be up and running in a matter of hours.
Automating a seemingly simplistic task like nameplate extraction can have a profound impact on a business. Super.Image makes automating nameplate data extraction easy, reducing process complexity and time by orders of magnitude, while increasing data accuracy. Some specific benefits of applying AI to nameplate extraction include:
Nameplate extraction may sound like a trivial task, but as we've seen in this article it can have a massive impact on the bottom line for TIC providers and the productivity of testing, inspection, and certification experts. However, artificial intelligence has broad applications and is capable of disrupting business processes in virtually every industry. This means the potential for AI to disrupt the TIC services sector goes far beyond automating nameplate data extraction. For more information on AI for testing, inspection, and certification, check out the following resources: