The new industrial landscape preaches the inclusion of automation, sensors and intelligence in manufacturing processes. The Internet of Things (IoT) came to aid in this digital transformation, allowing monitoring and control of all types of machines, robots and production lines.
This monitoring and control create a large amount of data, proportional to the number of IoT devices distributed in the factories. This data needs to be sent to another device or data center — through an internet connection — to be processed.
This data processing generates understanding and insights about how the manufacturing processes are going, the status of each machine, supply usage, stocks and other parameters. However, the process also increases data traffic on the company’s network, which increases latency, costs and the need for infrastructure improvements.
To solve the problems caused by the increase in data traffic that the Internet of Things brings, a new concept can be introduced: Edge Computing, which seeks to carry out data processing as close as possible to where the data is generated.
Edge Computing devices
Edge Computing is computing performed on devices that collect data — which are located on the edge of a process, rather than a data processing central station. In factories, an edge-computing device can range from a traditional computer with CPU, keyboard and monitor to an IoT device itself. The processing power needed in these devices depends on the data to be processed.
An IoT device that monitors multiple sensors doesn’t need too much processing capacity. It can focus on centralizing data and controlling smaller processes. When the device also becomes an edge computer, it needs a little more processing power to be able to process the data and even make decisions locally.
Even a smartphone can work as an edge processor — that is, it can process data in the place where it is collected — given the great processing power in current smartphones. With the many options for connection — Wi-Fi, Bluetooth, NFC etc. — available on an ordinary smartphone, it can connect directly to smart sensors, process the data, make decisions and send only what is needed to the cloud.
Edge Computing Benefits
In the traditional factory structure, the data network ends in an internal data center or an external cloud where the storage, management and analysis of the data collected by the sensors are carried out. This works well when only employees’ computers are connected to the servers.
However, as IoT devices are added to the network, the volume of data being trafficked and the complexity of the system increases. This is due to competition between them for communication resources.
Therefore, higher data traffic means higher latency in the system. Latency is the time it takes for communication to occur on the network. With more devices connected to the network, communication loses in speed. That is, higher latency causes slowness in the network, making it difficult for the devices connected to work.
In addition, if the company uses a cloud service provider or its servers stay on the premises, there is an increase in costs due to the higher use of bandwidth and resources.
By bringing processing closer to where data is generated, the network is less burdened and resources are better used. Latency decreases as the server is closer and connected to a smaller number of devices. Data sent from edge devices to the central server or cloud is already pre-processed. Additionally, only the necessary data travels in the network, generating less traffic.
Thus, applying edge computing in factories saves in bandwidth in the cloud and in the need for servers with high processing capacity. With processing spread over the factory, the different areas become more independent.
Therefore, failures or stops in one area do not affect processing in another area. Even a failure in the central server may not get in the way of internal processes, since edge processing ensures local storage for future sending, when everything returns to normal.
Another advantage is data security. Sensitive data may be limited only to the computer or device at the edge by traveling in a safe location, within the factory, or in an even smaller area. Sectioning and isolating the treatment of data also gains in robustness, since one system is isolated from another. A failure or shutdown in an edge system affects only that location, the other systems continue to work as usual.
Adding intelligence to Edge Computing
With the recent increase in the processing power of IoT devices, Edge Computing also gains the power of artificial intelligence. Artificial intelligence is the field that develops technologies focused on reproducing human skills such as data analysis, decision making and image recognition. In some cases, we can process the data and generate responses directly on the edge computer.
Thus, when, for example, a sensor is read, the value is stored and processed locally. The response to control a process is also executed locally, in an automated way, increasing decision making speed. This makes the system even smarter — because it has more information to inform future decisions — and only sends the status of the process to a central system.
One of the biggest beneficiaries of this evolution is image processing. Images coming from cameras can be processed locally, without the need to be sent to a central server to be processed. Vision systems, which identify objects, perform Optical Character Recognition (OCR) and other apps get faster and do not take up much network bandwidth when doing everything at the edge.
With more intelligence at the edge, apps that occupy large processing power can improve their performance, as in virtual reality used in the industry. Rendering virtual reality items over the image of the cameras can be done in less time with local processing.
The inclusion of IoT in factories with the digital revolution has created a huge amount of data to be processed, generating more traffic on industrial networks. Using Edge Computing, we bring processing, storage, analysis and decision making closer to where data is generated.
This means that the extra traffic generated by the new Industry 4.0 paradigm — which reaches central servers or in the cloud — is limited only to what is needed. What is sent to central servers is not a piece of data collected by a sensor, for example, but the result that has already being processed: an indicator or a status that can be used for a decision-making.
Thus, the availability of data locally can help reduce costs and provide agility in processes. Changes can be made directly at the place where the data is generated, by the people who know the processes the most, directly on the production line. This enables more precise and faster action on the data collected, allowing processes to become more agile and intelligent, opening new possibilities for improving the performance of results of all types of business.