Moore’s Law puts unprecedented compute and storage capacity at the edge. This combined with modern algorithms will spur the rise of the uber sensor. In 2019, a huge opportunity exists for organizations to develop applications at the edge using streaming data for use cases such as condition-based maintenance for mining machinery, autonomous vehicles, and much more. Organizations can improve performance and save money on data movement by executing applications at the edge and sending results back to the core.
Condition-Based Maintenance
In mining, self-driving vehicles, including mine cars and ore trucks, help streamline operations and reduce costs. Using sensors to monitor the health of machinery, companies can shift to a condition-based maintenance model.
Condition-based maintenance means maintaining equipment when there is an actual need through predictive analytics instead of relying on a regular maintenance schedule or repairing equipment only when it breaks down.
Small edge clusters on a worksite or manufacturing site ingest time series data coming from drills, valves, and trucks. Once the raw or aggregated data is transferred to the central data platform, it is enriched with the work orders, alarms, metadata, and more coming from traditional SAP or asset management systems. This central repository can be used to provide various dashboards to mining and process engineers to not only visualize and explore data coming from various work sites, but also ability to correlate it with data from other edge devices. Combined this gives analysts and the data science team the information to build predictive models, which can then be deployed at the edge.
Yield Management
Yield management is another area where sensors at the edge can dramatically improve business outcomes. In agriculture, analysis of drone and satellite imagery can help determine crop health and improve yield, but building video image collection and classification infrastructures can be tricky.
First, remote farms rarely have reliable, high-speed network connectivity. There is either insufficient network bandwidth to transmit full-fidelity video for remote classification or the connectivity isn’t reliable enough to trust that classification will happen when it needs to.
Second, distinguishing between objects that look similar – a healthy crop compared to a crop drying out from insufficient water – means off-the-shelf image classification models won’t work, and specialized models need to be trained.
Data collection and processing infrastructure need to be architected to classify images as close as possible to the point of collection. Edge processing combines several related, but distributed collection points for classification and processing at the edge. Edge data is aggregated and sent to a central place like the public cloud for further model training and more advanced analytics.
Maximizing IoT at the Edge
There are countless more use cases where computing at the edge not only provides a competitive advantage, but may be the only way to analyze data in real time. In 2019, edge sensors will drive data transformation initiatives for organizations looking to improve IoT outcomes.
