Nine ways manufacturers can use data and AI to get ahead of the competition

Digital technology has long been used to assist manufacturing, and manufacturing businesses have often been at the forefront of technological change. In the 1980s and 1990s, almost every aspect of the manufacturing process came to be ‘Computer Aided’, resulting in a blizzard of acronyms including CAM, CAD, CAPP and CAE. However, it can harder to see precisely how the rise of technologies such as machine learning, computer vision and artificial intelligence will affect the sector, given the much more diverse and increasing hyper-specialised nature of the manufacturing sector.

Terms such as ‘Industry 4.0’ or ‘Smart manufacturing’ are used a kind of catch-all for the applications of many of these technologies to manufacturing operations, but the exact meaning or implications are often left vague. I will highlight what I think are the key data-driven technologies that will affect manufacturing businesses in the coming few decades and sketch out how and why I think they could be used.

1.       Digital Twins

Digital Twins have the potential to change manufacturing profoundly. While Digital Twins have been used extensively to build models of individual components (and sometimes processes), in order to perform tasks such as predictive maintenance, the potential of the idea is much greater when applied at a larger scale. The ability to construct digital models of whole systems or organisations will offer opportunities in risk management, logistics, process optimisation and that will be transformative.

2.       Generative design

Generative design has been around for some time. One of the more famous examples is the aerial of the NASA ST5 spacecraft, designed by a genetic algorithm in 2006. What has changed more recently is the use of very powerful deep learning approaches, such as generative adversarial techniques and reinforcement-based methods, to construct very complex designs. These tools are becoming increasingly seen in commercially available CAD software, and could change design profoundly, since they use techniques that explore a far wider range of permutations than are possible with traditional methods.

3.       Augmented Reality and Virtual Reality

It has been clear for several years that AR and VR techniques will become increasingly widely used by manufacturing. For several years EDF Energy has used VR to deliver key health and safety training for employees who operate and maintain their nuclear power plants.[1] They also use an ultra-high resolution VR simulation to train the maintenance engineers who operate the plants, reducing down time for maintenance.[2] In future, this could be used to allow fully immersive virtual design of manufacturing plants prior to breaking ground, minimising unanticipated obstacles. It could also be used to provide fully immersive product demonstrations to users on the other side of the world, even before any products are produced or shipped.

4.       Additive Manufacturing (3D printing)

3D (and 4D) printing allows manufacturers to deliver customised finishes to their product offerings, as well as enabling the construction of shapes that would be difficult or impossible by traditional techniques. Plastic still is the most frequently used substance in additive manufacturing, but far from the only option. It is easy to see how this will be a boon to housebuilders, with individualised finishes provided on-site at modest cost. But as sectors such as textiles are also experimenting with 3D printing, for clothing, footwear and home furnishing, as well as automotive and toys. Research is under way to expand the range of materials commonly available, and the range of substances is likely to increase in the future.

5.       Robotics 

Many highly automated manufacturing processes have been using robotics extensively for many decades. However, many processes that require skills such as manual dexterity are typically still human-centred. Advances in sensors and deep learning (largely reinforcement learning) have now advanced robotics to the point that this will probably be less true in future. UC Berkeley in fact demonstrated a robot that could learn to accomplish that most boring of human tasks, folding laundry, as long ago as 2011. More impressively, Boston Dynamics has recently unveiled robots that can accomplish tasks requiring fine-tuned, dynamic motor control and reflex actions, such as balancing on a ball.

6.       Machine Vision

Image processing technologies already have a central role in many industrial processes – in many cases, along with IoT devices it is foundational to the power of robotics. The use of advanced imaging devices backed by machine learning is central to Quality Control in many precision engineering process. This is likely to increase, and use of MV will be tightly linked to the increased use of AI in robotics.

7.       Robot Process Automation

RPA (not to be confused with robotics) is usually described as using software to mimic the operation of a human computer user. It is used to automate time-consuming, repetitive processes with minor variation, such as filling in forms, creating presentations or inputting and extracting data. By ensuring the free flow of data, RPA can also improve process management, for example by:

• Automatically ensuring that all relevant data is stored securely in a central location (with proper access controls – no more manually copying email attachments)

• Automatically updating a database or spreadsheet when information is entered elsewhere

• Automatically notifying all interested parties when an event occurs (no more miscommunications or emails flying around)

• Automatically generating tasks in process from a set of inputs (no more errors from manually following a flow chart!)

8.       Logistics and Supply Chain Analytics

Logistics and supply chain data can be another source of competitive advantage for manufacturers. By merging their internal SCM data with external open datasets, it is possible to conduct analytics to assess supply chain disruption risk, predict disruptions before they happen and conduct scenario analysis to assess long and short-term costs and potential unforeseen knock-on impacts on operations.

9.       Customer analytics

All this is in addition to less obvious ways in which manufacturers can use data. For example, like any company, manufacturers may use customer data or analysis of social media to optimise their marketing efforts or improve customer service. While this is perhaps more feasible for consumer-focused companies, many manufacturing companies have a diverse, often global, customer base, and could use customer analytics to profile their most profitable customers, understand the value generated by different customer segments, or up-sell and cross-sell existing customers through improved product recommendations.

[1] Lloyd Dean, Head of Digital and Innovation Learning, edfenergy.com, 18.06.18 (Virtual Reality Learning | Innovation Blog | EDF (edfenergy.com).

[2] VVProPrepa – The Reactor Building in one click. edf.fr Virtual reality and optimised nuclear power plant maintenance (edf.fr)