Most industry professionals agree: The future of manufacturing is smart. It makes use of modern technologies to revitalize old plants and production methods. That’s why “smart manufacturing” is, according to a new survey by Information Services Group (ISG), a chief priority for many enterprises right now. Top consulting firms have dedicated lots of time, money, and energy to implementing smart manufacturing across a range of enterprises, and this effort has resulted in the continued development of important concepts like “factories of the future,” “lighthouse plants,” and “digital manufacturing.”
Central to the smart manufacturing effort is a hot-ticket item: Industry 4.0 — or I4.0, which represents that future of automation and machine-to-machine communication. Think of I4.0 as a strategy whose approach is to modernize manufacturing operations, and whose goals are to improve efficiency and productivity across the industry, thus increasing profitability. This is not only about technology, but also those who work with that technology: The Industry 4.0 roadmap leads to a near future in which the workers of tomorrow will find their jobs more interesting and engaging in part because of the 21st-century technologies introduced into their workplaces.
A great strategy, though, requires great tactics. For I4.0 to succeed, the manufacturing industry will need to plan carefully for the future it wants to create.
Your Best Tactics Are the People You Work With
What, concretely, are the tactical steps to manifesting I4.0, and what kinds of technology and processes will it take to transform the industry? To answer these questions, we need to look at the evolution of Industry 4.0. Since the term was coined in 2011, AI and ML were seen as natural prerequisites for I4.0. Given the way these technologies have become an everyday part of our lives, their potential to transform the way we work and produce materials was obvious — from changing the way we process information to improving mechanical operations.
How to leverage these technologies effectively is another matter. At one point, many industry leaders predicted a near-future in which neural networks would run plants with little to no human intervention. A decade later, this prediction has failed to materialize for most manufacturers—especially those who think that new technologies will change everything all on their own.
AI and ML, as analytical tools, are only one part of the solution. People and processes are required to put technology to work and effect meaningful change. This becomes clear by looking at the technological landscape of industry, where we find AI embedded:
- in the CAD tools that design engineers use
- in the sourcing systems leveraged by procurement teams
- in vision systems that count products and detect defects
- and in the sentiment analysis that drives marketing decisions
These AI-powered solutions add value not on their own, but because they put actionable information in the hands of people making decisions and developing products.
The takeaway? If you want a factory of the future, your tactics need to be centered on the people who make use of the technologies you adopt. You need to equip workers with a modern, data-driven approach to decision making. And you need to develop a collaborative process for sharing, analyzing, and executing decisions based on vast amounts of data from a myriad of systems and vendors. It's not enough to have intelligent programs push unexplained data to the cloud in the hope that a data scientist, somewhere, will understand it.
Organizing Around Data, Data for Organization
Let’s pause on the data problem for a second. Data is the material out of which any smart manufacturing future will be made. This is not a new revelation, and yet it often goes overlooked: Data is what feeds AI and ML models, and it is what those models convert into insights for workers to consider and implement.
It’s not uncommon for manufacturing leaders to bemoan the failure of predictive models to scale beyond one type of machine at a given site, or to wonder why AI itself can’t independently uncover known failure modes that weren’t included in the training data. But it’s the process that failed in these cases, not the data or the people.
Often the problem reverts to the fact that many companies disperse their data analytics operations across multiple, siloed domains, which don’t necessarily work together and sometimes use different, incompatible tools and software. The inverse case is its own problem: An enterprise that centralizes all of its advanced analytics within a closed team of data scientists will immure that team against what some of the data actually means. A data scientist might misunderstand an input that otherwise makes sense to a controls engineer. The result of this misunderstanding is most often an inaccurate and unusable model.
Between these two extremes lies a happy medium: building an AI Center of Excellence (CoE). At any enterprise, a functioning CoE will comprise an integrated network of data collaborators, across a variety of teams from engineers to data scientists to business analysts, that makes use of a single system to share insights and solve problems. This is especially important for manufacturing. Manufacturing data contains noise and lots of highly correlated (sometimes redundant) variables often collected at random intervals with obscure naming conventions. In the quest to predict and forecast, you need cross-team agreement about what you’re trying to predict and what kind of information you need to predict it.
If we look at the centrifugal motor, for example, what data is needed to predict failure? Is electricity consumption enough? What about vibration? How typical is a given motor relative to others in the plant, and can they all be sensorized in the same way? A powerful, flexible system of coordination is needed to tackle such multivariable problems.
Building Bridges From IT to OT and Beyond
What we’re talking about here is the challenge of bridging the gap between information technology and operations technology (the IT/OT gap). The data-informed solutions to manufacturing problems of all kinds don’t reside in AI and ML models, but in the collective brainpower of any company’s engineering, operations, and maintenance teams. The key to making smart manufacturing a reality is finding a way to tap into that collective brainpower.
Imagine a single enterprise problem-solving framework that provides engineers, analysts, and domain experts with a means to collaboratively analyze large data volumes across a wide array of data sources. And imagine if that same framework also provided the teams using it with advanced analytical methods that suggested meaningful correlations and provided quick forecasts. That’s what it will take to bridge the IT/OT gap, and that’s what Dataiku offers through its infinitely customizable platform combined with out-of-the-box manufacturing solutions.
Analytics platforms like Dataiku’s let OT teams mine the data they know best and come up with new solutions to existing problems — solutions that IT teams then enhance, operationalize, and support with the technologies that they know best. Most organizations have the people and the technology in place to realize the I4.0 vision. They just need the right tool to bring it all together.
Beyond Hype: Real AI Is Everyday AI
Hype cycles come and go, but complex problems demand discipline and innovation if they are to be solved. There will be no shortage of books and think pieces claiming that the manufacturing industry of the future will be run entirely by intelligent robots, or that AI has the power to transform the industry all by itself. But this is science fiction compared to how AI is actually used everyday by a variety of industries across the globe. Like all technologies, AI exists to be wielded, not to wield us.
The greatest technology, for all its potential and power, still benefits from human input and help — this is as true now as it will be in the I4.0 future. Many tasks can be automated, but there is still no substitute for a domain expert defining the operational theory of equipment to determine where, amongst a sea of noise, a data scientist should look for signals — for correlations and causation. Any smart manufacturing vision is only attainable when a systematic process for bringing smart people and smart technology together is in place.