Roundtable 1: AI for the Factory Floor

Roundtable 1 Banner

The benefits of employing real time sensing with modeling, in particular predictive modeling, to control production quality in-process have been recognized almost since the advent of digital computation. This concept lies at the core of Industry 4.0, Smart Manufacturing, Digital Manufacturing, etc. and is fulfilled in the notion of a digital twin (recognizing that definitions of digital twin vary widely). Over the past half century of computational modeling, manufacturing science has progressed to the point where almost any manufacturing process can be brought under computer control but the solutions are expensive, time consuming, and often require the capabilities of highly trained professionals. Worse, they lack generality and often can be applied only to a limited range of processes or machines and are not easily maintained. This has limited the penetration of solutions to expensive, difficult to produce products, such as jet engines, or very high volume production as in semiconductors, automotive components, and materials in the process industries.

Computational modeling challenges for harnessing mountains of data from factory operations led to a focused discussion on machine learning methods and their potential to provide the generality and the associated dramatic cost savings that computational modeling methods have so far failed to achieve.