Within four years, more than 30 per cent of businesses and organizations will include edge computing in their cloud deployments to address bandwidth bottlenecks, reduce latency, and process data for decision support in real-time. Edge computing accomplishes this by bringing the businesses’ computational processes closer to the data sources, increasing the speed of these actions. Additionally, even if a single node is unreachable, the service still should be accessible to users.
What’s more, 69 percent of organizations say that prioritizing edge-based analytics will improve their ability to meet the Internet of Things (IoT) objectives for specific use cases.
Industries including manufacturing, water and wastewater, utilities and building, are implementing hybrid strategies to enable real-time analytics, such as machine anomaly detection and diagnostics, quality analytics, energy analytics and overall equipment effectiveness (OEE). First, anomaly detection on the edge leverages machine learning to monitor machine health, detect anomalous data from sensors, and reduce the time it takes to get critical information. Advanced notice of anomalous machine behavior gives maintenance employees time to prevent breakdowns before they occur, saving the business time, money and resources.
Second, running quality analytics on the edge enables faster decision-making, which is important for many industries. This type of data analysis on the edge is important for businesses that use real-time data to improve productivity or require solutions that scale over time or reside in a fast-paced environment full of unexpected changes. Edge computing gives you access to analytics and actionable insight on edge, right where the data is generated.
Third, energy analytics on the edge has allowed utility companies to get real-time data at remote energy production facilities, such as wind turbine farms or solar farms. It is not practical for remote equipment at these locations to quickly transmit data to and from the cloud because it slows the data analytics process.
Lastly, OEE measures how well a manufacturing operation is utilized compared to its full potential, measuring the percentage of manufacturing time that is truly productive. This includes measuring the speed at which the parts are produced (the performance), the quality of the parts being manufactured (the quality), and the number of interruptions to the manufacturing process (the availability). A perfect score of 100 percent indicates that all the manufactured parts are good, they were produced at maximum speed, and they were produced without interruption. Measuring these aspects is a best practice for any manufacturing operation.
Centralized Cloud Analytics Stumble in Critical Manufacturing Areas
Many enterprises have adopted cloud-first strategies. They have married their workflows to cloud platforms to connect low-cost, elastic global infrastructure with rich device data. Initially, this approach allowed these organizations to accelerate the deployment of connected products and industrial internet efforts. However, as they scale their digital transformation efforts, cloud-only approaches limit growth because of delays to transmit data from devices to the cloud and transmit analytics from the cloud back to devices. IoT use cases on the manufacturing floor often have unique, real-time data analysis needs. It is not always practical, economical, or even lawful to move, store and analyze IoT data on a core cloud infrastructure. Many manufacturing professionals recognize these limitations.
Edge computing solutions, which converge hardware and software into increasingly smaller devices that run smarter analytics onboard, enable real-time decisions and insights. Momentum for edge IoT solution deployment is increasing at a faster rate in the manufacturing, utility, and building use cases.
Edge computing often incorporates machine learning (ML) and artificial intelligence (AI) technologies. These techniques make the calculations performed on the edge even more efficient. That way, the system does not require the help of human operators to identify data irregularities that may point to a potential problem developing with a machine or system. AI can flag anomalies in an actionable way so that machine breakdown can be prevented. Another use case for AI on the edge is the detection of defective parts in a manufacturing operation.
However, the accuracy of the model’s AI uses for these purposes may degrade over time, so this is where ML becomes important. ML is incorporated into the process to create a closed-loop in which the computer contains supervisory programming that observes the accuracy of the AI model over time by analyzing data drifts within it.
An edge computing platform is designed to integrate both wireless and wired sensors into the IoT network. It compiles, validates, quality checks, and processes the data. It efficiently uses bandwidth to store and forward data, creating a sensor data lake. The data collected is also used to nimbly detect anomalies in machine operation on the edge.
Figure 1: Enabling high value use cases
This helps you stay on top of your operational goals, efficiency objectives, and machine health status, all in a simple package. This technology detects an anomaly and provides machine operators with the tools to determine the cause. Drill-down widgets and rule-based alerts couple with ML technology to enable easy machine diagnostics. Key performance indicator (KPI) calculations and machine fault mode diagnoses take the raw data collected by the system and turn it into actionable intelligence. Rather than allowing you to get lost in the sea of big data, this technology pinpoints the important nuggets of information and presents it to you in an easy-to-understand manner. This allows operators to quickly fix the issue and get critical processes running again.
Overall, these features reduce operational downtime, repair costs, and labour costs, while increasing energy efficiency and production output.
