What comes to mind when you think of a factory? A dark Victorian mill? A Charlie Chaplin skit? Or a roboticized production line? Industrial manufacturing has consistently been a marker of progress, acting as a sign of the times as each generation has developed increasingly advanced technologies.
But although industrial companies have used digital technology for years to improve their processes, the power of connected sensors, data and artificial intelligence (AI) at all stages of an operation has yet to be realized on a grand scale. This article will focus on the intersection of Big Data, AI, and IoT devices in an industrial setting, continuing the series on these technologies and the ecosystem they support.
The industrial sector is an ideal proving ground for automation and optimization, due to the huge number of specific processes in any operation. These organizations also tend to be spread across various locations and have a huge chain of suppliers, distributors and users which their product or service may touch. This means that the processes followed at each stage – whether during production, maintenance, or distribution – can end up being fragmented between operations. Technology therefore plays a huge part in industrial environments, helping to log material usage, measure flow in utilities and oil pipelines, and bring operations together under one system.
Industry 4.0 refers to the next step in industrial technology, with robotics, computers and equipment becoming connected to the Internet of Things (IoT), and enhanced by machine learning algorithms. Advances in sensor technology and connectivity modules have allowed more equipment to be measured, monitored, and tracked between sites, and orchestrated from a central, remote location. With this accessibility, managers, executives and even data scientists can use that insight to improve the efficiency and productivity of the whole operation. Thanks to the rise of cloud computing and the consequent falling costs of data storage, a huge amount of data can now also be stored and fed into machine learning algorithms to help automate specific processes within an organization.
A holistic view
Bringing AI into industrial processes is not as easy as buying a new piece of equipment however. Due to the complex and interlinked nature of industrial processes, companies must have a solid understanding of what they want from AI in the first place. ‘Whether it comes from sensors along the production floor or connected devices out in the wild, ultimately you can’t do anything with that data without having a structured thought-process’ says Shekhar Vemuri, CTO at Clairvoyant. With a ‘strong foundational data strategy’ in place, companies can then look at the whole system end-to-end as data flows through an enterprise, as long as data itself is the focus. ‘If you still think of data as a secondary product of your operations, then your organization will keep struggling’ says Vemuri, ‘with data as the primary asset it becomes part of your business processes, and you can see how each bit of data relates to the other.’
With a holistic view of the production line, companies can use AI to get further value from that data or gain more insight into the equipment. ‘People are now looking at how to leverage industrial IoT sensor data to project things that may happen – whether predictive maintenance, line management or quality control,’ says Vemuri, ‘and things will start snowballing over the next 12-18 months because there’s growing knowledge there to make it happen.’ Vemuri is clear to point out however that technology is not everything: ‘while I would love to say it’s purely a technical problem, it’s a people and an organizational problem too’, so while setting out a solid goal of what you want to achieve is important, ‘you have to think big but act small, because otherwise it’s too abstract to go and grasp the value – set your goal but make baby steps towards it.’
Connecting the dots
Collecting all this data, however, relies entirely on a reliable connection. Various connectivity standards exist that cater to the industrial environment, optimized for sending data packets over long distances (LPWANs, or low-power wide-area networks), keeping a device in the field for as long as possible (embedded SIMs or eUICC), or maximizing the potential of existing network infrastructure (cellular LPWANs that use the LTE spectrum). Ethernet, or wired internet connection, is also widely used in local, stationary IoT applications, but is not suitable in many cases that need wireless connectivity. All of these connectivity types have their benefits and drawbacks, but many industrial companies run on ‘legacy technology which is simply not built for the new world,’ says Venkat Viswanathan, co-founder of LatentView Analytics. Because of this, uprooting an organization to adopt a new connectivity standard would be completely unfeasible in many cases.
Industrial companies may well choose cellular multi-network connectivity for the meantime then, and work on getting more of their processes automated before next-gen network technology becomes available. Edge computing alongside machine learning may provide part of the solution, as this allows more data to be qualified and more automated tasks assigned at the source, before being transmitted to the cloud for analysis, reducing the amount of data transmitted over the airwaves. Whilst in some cases this replicates legacy equipment from a communications perspective, edge computing improves latency and efficiency that true automation requires, and will remain compatible with newer systems where legacy systems are already becoming out of date. This also allows companies to evaluate their existing systems and processes using the reliability of cellular connectivity, and bring about the incremental change needed to achieve complete automation.
Piece by piece
Incremental change is the name of the game, as many industrial organizations are too spread out and fragmented to perform a complete overhaul and immediately benefit from cutting edge technology. In fact, ‘many of these processes are still completely manual’, according to Viswanathan. ‘If you can imagine a refinery and the various equipment there, they actually have people eyeballing the equipment, looking for a problem and making a note with paper and pen.’ While this is, of course, an extreme example, both Viswanathan and Vemuri agree that ‘the number one thing that companies need to focus on is sponsorship from top management.’
Bringing about a new wave of industrial progress with AI, Big Data and IoT will not happen overnight. To take advantage of the opportunity that these technologies bring requires a holistic strategy, strong leadership, and an understanding of how data flows through an organization. Many industrial companies rely on equipment that functions perfectly well but will not fit with new technologies, and the same is true of some of the executives at the top – they are perfectly happy with how things are, but will not be able to adopt new technologies without a change in mindset. The increasing number of leaders that do appreciate the benefits of Industry 4.0 however, need to remember that meaningful change on such a huge scale can only come in baby steps.
By Charles Towers-Clark
Last month, the FDA announced the 12 winners of it’s Low or No-Cost Tech-Enabled Traceability Challenge, which was part of the New Era for Smarter Food Safety initiative. The initiative aims at achieving end-to-end traceability throughout the supply chain.
Digital tools enabled by advanced analytics, artificial intelligence, and machine learning can help companies uncover the fastest and most effective path to abating the O&G industry’s greenhouse gas emissions.
The time to prepare for change is not when it hits. It’s before it hits, and during times of relative calm. Reacting to change in the moment keeps you forever on the defensive, and the consequences can be severe.