Modern Manufacturing: Machines Learning from Machine Learning


Industry Trends

The manufacturing production value stream hinges on efficiency and a surefire guarantee of as little error as humanly possible from concept to delivery of a service or product. However, as competition and a need for responsiveness and agility increases, what is “humanly” possible is no longer good enough. Production facilities are integrating more artificial intelligence (AI) technologies into daily operations. This integration of artificial intelligence not only helps project managers increase efficiency, but also helps modern manufacturing facilities of all types (assembly, machining and raw material processing) predict and prevent future failures before cost impacts or catastrophic events, much better than a human could.

Smarter, Better, Faster

AI is already solving some of the biggest problems facing industry today in terms of efficiency and production output. For example, many project managers face the challenge of synchronizing product design, process capabilities and production capacity, which directly impact schedule, cost, and scope required to meet customer expectations. This is exacerbated when the factory has non-collocated or remote design engineering support with limited visibility into operational health (inventory, design, supplier, machine health, scheduling).

Forward-looking manufacturers are also using AI to predict work-in-progress (WIP) queuing up in front of constraint machines within the factory floor. By decreasing WIP, the manufacturer can cut down on lead time and service more customer orders, leading to higher revenue per factory. They are also harnessing AI to extend their product lines by anticipating what process steps will result in scrapped parts.

Additionally, predictability in operations has long been unexplored by manufacturers: most of the data that operators have been able to collect and analyze in the past are simply capable of presenting past information for human processing, but never before have machines been able to essentially anticipate what’s coming next. We’ve never had technology that’s been able to critically think for us, or further, been able to demonstrate its reasoning.

The benefits of this machine-driven predictability and learning for manufacturing are proving to be significant. For example, AI-based machine prognostics models have delivered substantial improvements in reducing unplanned downtime of critical equipment and spare parts inventory cost by predicting machine failures with sufficient forewarning. This allows operators and GMs to streamline their operations and supply chains, maximize production output, and reduce costs from idle machine operators.

ROI of Industrial IoT

The ROI of applying machine intelligence to industrial operations is undeniable, though it varies based on the complexity of the operation and the equipment. For example, a product design issue for a turbine at a large industrial operator was missed by the original equipment manufacturer and the operator itself. The product design issue caused damage that could have led to catastrophic failure, destroying the compressor and the rotor. Machine intelligence predicted this failure before it could occur. The operator estimated the cost of this failure would have been over $100 million in parts and replacement. This is easily 50 to 100 times return on investment.

Challenge in applying AI and ML to industrial processes

Machine learning can be used in almost every facet of modern industrial processes, including improvements to product design collaboration and predictive operations. While industrial AI has proven its value, like most things, it needs to overcome a few challenges.

For many manufacturers, the lack of effective IT support is one of the biggest challenges. Even at Fortune 500 firms, production in the factories is not as sophisticated as it can be due to restrictions from IT, which can limit an operator’s ability to digitize. These limitations have included things like a prohibitively high bar for security standards, resisting cloud adoption, or favoring approved vendors over new ones. Secondly, IT organizations are struggling with managing industrial data, with their wide variety of formats and often incompatible systems.

Next wave of industrial automation

Today, AI is solving narrow problems with unparalleled accuracy and precision. The next wave of AI will expand its role in solving even more challenging problems.  Some examples may include making sophisticated decisions due to enhanced situational awareness and supporting strategic decisions by finding nuanced relationships in large data sets.

In the next two to five years, we’ll likely see the adoption of “mixed reality,” which is the combination of augmented reality (AR) and virtual reality (VR). AR and VR are still early on their adoption curves; however, mixed reality is gaining popularity due to its potential for tremendous benefits, especially for aging workforce issues. For example, augmented reality, driven by natural language processing, could walk a technician through steps to make repairs, highlighting parts involved, displaying instructions, and facilitating a call with a remote expert.

To increase efficiency and remain competitive, project managers and manufacturing facilities will require artificial intelligence. As companies overcome obstacles to adoption and recognize the value, AI technologies will become increasingly critical to staying viable. Early adopters of machine learning are likely to progress to incorporating mixed reality in order to maintain competitive advantage.

About the Author:

Usman Shuja is founding executive and general manager of industrial IoT at SparkCognition, a global leader in cognitive computing analytics. He is responsible for the company’s industrial IoT go-to-market efforts, strategy and product offerings.


Engage with ARC Advisory Group