Blake Burnette from IOT-eq showed how machine learning can be combined with internet of things edge devices to manage sand delivery during a frac job. This can be quite a complex process as customers have many different types of blenders. The sand hopper may run dry or overflow as the service company speeds up or slows down the blender. Many variables affect the process, frac job design, gate heights of the silo and equipment wear. A good problem set for machine learning. Burnette compared traditional PID-controlled sand delivery with machine learning. ML potentially offers a streamlined solution especially when used as a weighting factor to the PID controller. The problem ML is that it may ‘learn’ bad habits over time and the process may take too long. Enter IOT-eq’s ‘Toy Box’, a hardware and software combination that provides IoT connectivity to the cloud and a suite of machine learning dashboards that connect multiple sensors, cameras and operators to investigate and optimize ML solutions.
Mark Thompson from Swim.ai believes that edge computing is data-driven computing. The problem is that ‘databases don’t work at the edge’. A database stores information but does not have ‘agency’, the ability to act. On the other hand, real-world data is perishable, constantly changing and does not fit neatly into a database table. Enter the intelligent edge, that analyzes, learns and predicts from streaming data, on-the-fly, building a stateful digital twin model of the real-world directly from data. This is the technology that is used to control traffic lights and to route vehicles through a city without stops. Such edge-based systems can provide high processing bandwidth at a fraction of the costs of cloud-based central controllers.
© Oil IT Journal - all rights reserved.