Processing big data and setting up the ideal architecture

There are several questions raised on the best practices in collecting and integrating big data in industries. Is it better to follow a centralized approach? Or would a decentralized system work better? Needless to say, a lot depends on the specific requirements on a factory floor but there are some basic guidelines that experts provide.

“In reality, manufacturers will use a combined approach to meet the variety of needs,” said Heiner Lang, Senior VP of Business Unit Automation and Electrification Solutions at Bosch Rexroth. “Real-time applications which require 100 percent availability when it comes to communications, such as motion control and safety, will continue to be wired into the machines. In addition, these machines will send pre-processed sensor data to edge servers in the factory hall for immediate processing, error analysis and maintenance analysis. This will soon be done wireless via 5G. Cloud-based big-data processing will be used for non-real time planning and optimization tasks. Within the cloud, users can use the advantages offered by artificial intelligence to increase productivity.”

Philipp Unterhalt, MD of HAHN Group and Interim CEO of Rethink Robotics, said that in a complex environment such as production lines, one of the prerequisites lies within the decision how to deal with the data. Some clients would rather have it centralized in order to enable full monitoring access in one place while others will see the benefit in decentralized data.

“The latter, many times, will be related to the security of the data,” Unterhalt said. “We observe how cybersecurity becomes a major success factor for IIoT as clients will only provide their data if they are assured no third party can access it. Such actions will probably enable clients to centralize data. We believe partnerships in OT security are a precondition to successful Industry 4.0 big data offerings. Last but not least some models of big data see value in edge computing, which is the ultimate decentralized model. Doing so means reducing the need to maneuver data from the sensors all the way to the centralized system.”

Architecture suggestions for better data sharing

On-premise architecture is popular in industries today, but going forward, cloud-based solutions will become more and more popular.
 

“While the traditional way of processing the data and sharing is very local (which is typical for industries who are concerned about privacy), the entire industry is moving into ‘over the cloud’ architecture,” Unterhalt said. “We see high value in offering a ‘non-specific’ cloud solution, from Azure to AWS (Amazon) and others, and we believe our clients should have the flexibility to use any offered cloud services. We are also looking at new standards such as ‘in-factory’ Wi-Fi connectivity architecture to enable swift processing/sharing of the data, eliminating the need for cabling.”

Sunil Mehta, GM of Automotive Business Development, Factory Automation and Industrial Division at Mitsubishi Electric India, is of the opinion that a typical architecture as shown above for data processing and sharing is based upon a centralized Programmable Logic Controller (PLC) communicating with various equipment on standard protocol over the industrial Ethernet network.

“In case of automotive industries, one gateway PLC can be considered for each shop like stamping, welding, painting, assembly, utilities, etc., and further these PLCs can communicate with each other or server directly,” Mehta said. “With the technology advancements, edge computing is also becoming popular where edge computing devices collect the data from various equipment and then pass on relevant, important data to the cloud or server after processing. This will reduce traffic on the network and real-time decision making is possible at the shop level to improve the quality and speed of processes.”