Keep an Eye on AI
Big data is just the beginning. In the future, optimizing the system will be a function of the system itself.
Will artificial intelligence prove to be a technology that transforms your shop? Probably not next year, but possibly sooner than you think. Doug Woods, president of AMT – The Association For Manufacturing Technology, gave a compelling talk at this year’s Manufacturing For Growth meeting arguing for artificial intelligence as a development that manufacturers ought to watch.
A definition first. AI does not refer to sentient or human-like machines as presented in science fiction—let go of that image. The AI we are discussing involves algorithms that iteratively construct algorithms. The programming of a complex process need not be written line-by-line by a human, but instead a higher-level program might construct that program itself through trial and error. Mr. Woods’ definition of AI is “machine learning applied to big data to make predictions about future states.” Such a capability is being used today, for example, to enable learning-equipped robotic systems to develop human-like capabilities such as grasping a knob to open a door. With smoothly opening the door defined as the aim, the system gradually develops the programming for realizing this goal based on the details of its failures.
No dramatic invention was needed to make this kind of capability possible. Instead, several pieces have come together. They include cheap and reliable sensors, open data standards and wireless networks, as well as cloud computing for both storing and working with large quantities of data.
The move in the direction of big data and its application is already under way, whether manufacturers realize it or not. Think of the process data already captured in management systems such as ERP. Think of the machine-tool status information now available via MTConnect. Is there additional relevant data a manufacturer could capture? Is there more that could be done with the data than to construct status displays for humans to observe? Yes and yes.
A robot, instead of opening a door, might learn to recognize and handle a family of parts. Or—and here is the real promise—the production system overall could be structured to aim toward a carefully defined combination of accuracy, repeatability, productivity, safety and cost. What if the AI with these goals in mind was free to take in data throughout the shop, gradually identifying and advancing all the interacting variables affecting this set of metrics? The result would be a process that gets better. It would be something like continuous improvement realized not through human discipline, but as an output of technology.
Crucially, though, we are talking about artificial intelligence, not the real kind. Humans will remain vital. Successful businesses will be those with people best able to master this technology. As a result, in manufacturing, roles will change. Think of how we take it for granted now that “programmer” is a manufacturing role, where once that role would have been unknown. In the same way, says Mr. Woods, in a future not far off, manufacturers might come to employ team members with titles such as information architect or data engineer.
For machine shops, the transformation that data-driven manufacturing promises to bring begins with machine monitoring, and there is a human component to this.
Expanding capacity into the unattended hours calls for counterintuitive new ways of thinking about the work.
This shop discovered the most profitable thing it could do with certain RFQs is to prevent them from ever reaching the quoting team.