Machines making machines. The idea has been a staple of science fiction over the past 200 years. With the age of the Industrial Internet at hand, we're still not quite in the era of self-replicating devices, but we are witnessing the rise of self-aware machines.

Industrial assets, components of large-scale infrastructure, are gaining sensibility and smartness. Many of them are not only capable of reporting their own stats, but can communicate with other similar devices and cooperatively adjust themselves to achieve a collective overall goal. But with the emergence of these smart assets, the very idea of what an asset is comes into question. What distinguishes an asset from a component? Are there advantages to treating functionally connected assets like wind turbines as one entity? How far can you go in creating smart, self-aware objects?

Currently, there seems to be a top-down approach to deploying and managing smart assets. Let's say a utility wanted to know something specific, like the utilization of particular transformers throughout the grid, it would invest in transformers that are able to report their load to an overall information system. That information system likely operates a specific set of analytics that are called for in most situations. This approach is cost-effective and works well for answering specific questions, but one of the great capabilities of the Industrial Internet is to answer questions that are non-obvious. To that end, there might be an advantage in building in smartness on a more granular scale. Assets are made of several subcomponents.

What if each of these subcomponents themselves were self-aware and capable of reporting their health and their interactions with other subcomponents? The vision could be for a product like a washing machine that can diagnose its own failure by identifying the responsible component through a clever method of gathering data from the components themselves and the way they interact with one another. After using multiple queues to identify a faulty component, the machine itself could request a replacement part and schedule maintenance. Extend this idea across a fleet of deployed assets to identify a recalled component or execute preventive maintenance and data-driven decision making of this kind could be the ultimate realization of the Industrial Internet, where exigencies take drastically less time to resolve and potential failures are remedied far in advance of actual breakdowns.

But in order to make components to become assets themselves, there needs to be a common information platform that can scale from the smallest subcomponents to the largest intertwined systems. This requires both hardware and software platform unity on every level. Before, a lack of architectures that worked well both in situations that called low-power as well as high performance limited development. Today, there are several architectures like ARM's Cortex, Intel's Quark, and MIPS that can provide the hardware backbone for smart assets. Scalable, adaptable platforms like GE's Predix can be deployed within machines, on location, and in the cloud for a seamless flow of information and analytics from component to decision maker.

But what if we can take the Industrial Internet one step further? What if we can use the Industrial Internet to help build the next version itself? Feeding back data from the very smart assets it uses, the Industrial Internet could impact manufacturing in a very big way. The machinery used to build smart machines could be smart themselves, providing tighter control over manufacturing, quicker identification of defects, higher quality control, and even more efficient production and design of equipment. It might not be exactly the kind of machine self-creation John von Neumann talked about over 60 years ago, but with the Industrial Internet, the idea of machines designing themselves suddenly doesn't seem so far off.  

About the author

Suhas Sreedhar

Strategic Writer at GT Nexus