By John Slitz, Technology Industry Veteran
After attending the inaugural meeting of the Industrial Internet Consortium, it is indisputable that some of the biggest most successful companies in High Tech believe passionately in the transformative power of real-time industrial data when combined with historical information to produce new insights into nearly every aspect of a business or even a person's life. These sensors and their resulting data outputs are not new. Sensor networks have become an increasingly important component in monitoring, controlling and managing industrial processes and machines in the industrial landscape for over fifty years. What is new and incredibly powerful is the ability to merge and evaluate potentially thousands of different flows of data to gain immediate insights and actions never before possible.
As with every major change in the technology landscape, this one will bring a myriad of anticipated benefits and an even larger group of unintended consequences. The first hurdle that companies leading this charge must overcome is the sheer mass of redundant data that is generated from a logarithmically expanding sensor universe. Systems designed to handle terabytes of data are not expandable to petabytes or exabytes. New methods of pre-processing this high flow of information must be enabled to analyze, minimize, and extract value as well as initiate actions from the data before it is stored in a database. Today's thinking is a direct line of development from the methods and practices built into the Data warehousing initiatives of the mid '90s. At that time, each data element meant something and needed to be recognized, processed, retained and analyzed much like pallets of goods in a physical warehouse. The Industrial Internet will challenge this premise with the sheer speed, mass, and complexity of the data generated. For example, according to GE, a single sensor on a turbine blade within a jet engine will generate 550 megabytes of data on a single flight from Los Angeles to New York. What information in that vast data flow is the most important to collect, store, analyze and combine with other data to develop actionable information to be used during the flight and after landing?
In a medical application, the snap on finger sensor that monitors blood pressure and heart rate is generating a few hundred bytes of data per second -- a trivial amount. But combine that flow with that of other sensors, and you are able to have a clear picture of that patient's complete medical history as well as the medical literature detailing all analogous medical conditions. The flow of data can then report out an immediate alert to any rare or potentially life threatening situations that must be dealt with right away.
In both of these situations, the current ability to generate and collect data vastly outweighs the existent ability of today's systems to combine (perhaps thousands) of diverse sensor sources, evaluate real-time conditions, initiate immediate action and pass on the critical data and initial actions to be stored for further analysis and historical reference. This implies that a new area of "activity based processing" must be developed to meet this challenge that extends far beyond the intelligent control systems built to administer sensor systems today. This new systems must encompass an "instinctual processing node" which will be rule based and tightly coupled to the activities the industrial sensors are monitoring, which are programmed to act immediately -- as in the heart/blood pressure example -- then to store for later analysis patterns or other derivative data in the company infrastructure.
By focusing on these types of requirements and not specifying standards, the Industrial Internet Consortium's efforts will help the industry understand and tackle the challenges ahead.