The ongoing trend of digitalisation and datafication is transforming industries globally, driven by the promise of improved products at decreased costs and developed through better understanding of people, markets, processes, as well as the power of foresight and automation.
Some sectors such as media, entertainment, communications, insurance, retail and travel are almost fully digitalised and quite far ahead of the game, or in other words, they have all goods and services accessible via digital means. Banking is well on its way, but others, especially the legal and manufacturing sectors are lagging far behind.
Manufacturing has undergone pockets of digitalisation, but the bulk of the industry has been late to transform. This is likely due to a number of reasons: the physical nature of the processes involved requiring the design of cyber-physical interfaces, expensive tools that have a relatively long lifespan, and the culture still being based on an attitude of consistency and tradition. Combined together, or even as separate entities, they are all the enemy of innovation.
The continued success stories of machine learning (ML) and artificial intelligence (AI) are, however, beginning to trickle down and change this somewhat gloomy picture. The list of tasks where machines can go a step further than human experts is growing, and now includes Chess, the Chinese board game ‘Go’, and a raft of video games from Pong to Doom. These have the propensity to recognise numerical digits and images of objects and even translate text.
More recently highly applied examples have arisen, such as optimising the power consumption of a large-scale data centre (a technology that in principle is applicable to factories), the design and manufacture of hypercomplex structures using algorithms, additive manufacturing, as well as the planning of Computer Numerical Controlled (CNC) milling routines. These examples are beginning to make good on the early and unproven declarations of the incalculable value of machine learning. In parallel, the Internet of Things (IoT) movement has begun to aid in this increased digital momentum and is helping make the notion of cyber-physical systems a reality.
Through all of these state-of-the-art digital technologies, the barriers for the digitalisation and datafication of manufacturing are being deconstructed, paving the way for the application of technologies like machine learning and artificial intelligence.
With lower costs of sensors, data storage and computation power – manufacturing data has become significantly more abundant and accessible. IoT technologies and sensors deployed across the factory floor means that the capability of advanced manufacturing facilities in capturing and sharing data is pervasive. However, the real value will only be created from the insights derived from such data.
That said, we often talk about the role of artificial intelligence and its applications, but at the same time neglect to think more crucially about how we use AI effectively. Just like Maslow’s hierarchy of needs, you need a solid foundation for your data before it can become effective with regards to AI and machine learning as can be referenced by the AI Hierarchy of Needs.
It denotes that in an advanced manufacturing environment, it is imperative to ensure that data can be integrated from various silos, creating a single source of truth across data sets that will enhance decision making across multiple geographic locations. This in turn will allow for more accurate analysis and predictive services using autonomous systems, providing insights into demand, supply, quality and yield of production and use performance. This can be utilised in areas such as Enterprise Resource Planning (ERP) systems, Customer Relationship Management (CRM) Systems, Product Lifecycle Management systems (PLM) and Warranty Systems.
Bringing the data together and combining it with external data sources such as weather, traffic and demographic data, is essential to providing a comprehensive analytical layer that supports informed decision making and can ultimately alleviate the challenges of data complexity and volume that have emerged in recent years.
Assuming we can get this data element right, the potential use cases in manufacturing are extensive.
In Engineering, virtual simulation is now embedded within any product development process and has become a key capability of most manufacturing businesses. Applications such as Computational Fluid Dynamics (CFD) and Finite Element Analysis (FEA) require code, computing power and iterations to converge on a solution. AI thus provides an opportunity to reduce design cycles and produce more complex yet low-cost workable designs than ever before.
Today personalisation is the order of the day and has given rise to ‘batch-size-one’ products that take the form of highly personalised/specialised finished goods in bio-medicine, clothing, furniture, FMCG and highly specialist component production that will give rise to the development of massively distributed small-scale manufacturing and processing units.
We are also seeing the rise of digital twins that can be used for monitoring and diagnostics to better understand anomalies that can be used to detect manufacturing issues before they even happen. Although anomaly detection has been used for many years in the manufacturing space, digital twins help to provide early warning predictions being acted out
within the virtual model of the twin. This can be done by embedding virtual sensors that can create additional data, and increase accuracy. It is then possible to utilise machine learning algorithms alongside these digital replicas of the physical equipment to help optimise processes.
We all agree that processes have to adapt and quickly as manufacturing companies face the continuous challenge of rapid delivery of high-quality products in the face of increasing complexity. However, that most certainly can’t come at the expense of quality. Through associated process data, the Zero Defects Manufacturing approach has come to light and achieves increased efficiency by combining multiple sensor signals along the manufacturing process with a view to optimising parameters relevant to machine performance and product quality.
Manufacturing has to step up to the plate and adopt these leading-edge digital technologies that will help propel it from the back of the pack to a leader amongst industries. If not, it risks becoming a case study in late adoption rather than a glowing success story.