Fueled by a combination of industrial data and artificial intelligence, a new era of smart and flexible manufacturing is underway.
The technology promises to facilitate an environment where production can be seamlessly calibrated to pump out customized goods while factory floor equipment is maintained proactively to avoid costly downtime.
That was the focus of a recent panel exploring the evolution of smart design and manufacturing at EmTech Next, a virtual event hosted by MIT Technology Review. Presenters said artificial intelligence has the potential to touch and transform the entire design and manufacturing continuum, including:
- Early-stage ideation.
- Custom and flexible production.
- Predictive maintenance.
- Finished goods.
- Machinery out in the field.
The connective tissue underpinning this seamless workflow is computational models and data, with AI deployed to sort through a sea of options and identity optimal outcomes.
“Connecting models at different scales to the data you get from the real world is especially critical,” said panelist Saigopal Nelaturi, research area manager, principal investigator at PARC. The way forward is to incorporate both models and data together in an artificial intelligence framework that’s capable of parsing through large data sets to zero in on optimal designs or outcomes, he explained.
Here are four key tenets to keep in mind as you embark on your smart manufacturing journey:
Let AI help with design, process planning, and production
Whether the goal is getting to the optimal product design or uncovering the most efficient and cost-effective manufacturing method, the possibilities are endless when you set specific parameters and goals and work backwards from there.
Take conceptual design, for example. Typically an engineer comes up with potential designs for a widget or component based on a set of core requirements, but there are limits to the number of ideas they might explore. Not so in the world of AI-driven design, where an engineer can specify parameters like cost, weight, and strength targets and let software do the heavy lifting to computationally churn through all the possibilities to come up with a spread of viable candidates.
AI can work the same magic for the process planning and production methods that go into making that widget. By creating full-scale 3D and behaviorial models of plant floor equipment, teams can leverage AI-based tools to run through different virtual scenarios and simulations to determine what materials and systems will produce the goods in the most efficient and cost-effective manner.
“Your goal here is the part. You’re starting from raw stock and trying to figure out the best way of making it,” Nelaturi said. “Once the AI runs its analysis, you come up with an automated process plan that tells you how to orient the part, what materials needs to be cut out, what materials remain, and in which way the part can be built in the most efficient manner.”
Collect data where it lives — on the factory floor
Early on, the push was to connect industrial machines together via the Industrial Internet of Things (IIoT) and let machine-to-machine communications drive insights and automation. However, the complexity of plant floor data and a lack of clarity around goals hobbled a lot of early efforts, according to Matt Wells, vice president of digital product management for GE Digital and a presenter on the EmTech panel.
GE Digital advises companies to put more systems in at the plant floor level (often called edge computing) to aggregate industrial data and get it into shape so it can be mixed with enterprise data in the cloud for further analysis. Standardizing and creating context for plant floor data, which is typically disparate and stored in silos, is a crucial step, he said.
Wells cited Intel as an example of a manufacturer able to successfully leverage IIoT and data analytics to optimize the performance of its clean rooms. Using a sensor network, Intel collects data on its fan filter units, brings that data directly to the cloud where it runs analytics that predict when units might go down.
“Instead of having an unscheduled break, Intel can now schedule maintenance during planned downtime,” Wells said. As a result, Intel has been able to improve uptime by 97% while reducing unscheduled downtime by two thirds.
The next wave of efficiency will come from process analytics and “digital twins” — virtual models of an entire manufacturing plant, including the physical assets and their behavior. Used in concert with machine learning models, digital twins help manufacturers automatically detect issues that a human can’t and make recommendations that a human couldn’t otherwise figure out, Wells added.
Mind the gap between the C-suite and the factory floor
One of the biggest challenges to successful data-driven manufacturing is the disconnect between managers on the shop floor and leadership in the C-suite. Continuous improvement teams and teams local to a plant tend to work on specific challenges like increasing uptime or transitioning over a line; executives seek large-scale digital transformation to create new business models.
“There is a gap between what these two groups are doing, and often, they’re not talking to each other,” Wells explained. “This is a missed opportunity because the continuous improvement teams are close to the key problems . . . while the digital transformation teams are close to the technology and have the skills. Marrying the two creates a better ROI.”
Embrace a future of humans and machines working in concert
Because parts are geometrically complex and fixtures and tooling are so variable, it’s next to impossible for a human to run through all possible scenarios without help from AI. And as the demand for manufacturing flexibility increases, many believe the future is not a lights-out factory with machines humming day and night and not a human in sight, but more of a hybrid operation.
“It’s a flexible, dynamic collaboration between humans and machines,” said panelist Clara Vu, co-founder and CTO at Veo Robotics. “Where machines bring their strength, repeatability, and precision, humans can bring their flexibility, judgment, and dexterity. We can use both sets of qualities to build not just the same things better, but new things all the time.”