Machine learning is a powerful technology that allows computers to learn from data and self-improve. This form of artificial intelligence could have transformative effects across a range of industry sectors, including manufacturing.
On 18 July, leaders in machine learning and leaders in manufacturing met at the Royal Society to explore what the advancement of machine learning technologies might mean for the UK manufacturing sector. The workshop was the first in a series of events exploring opportunities presented by machine learning in different industry sectors, the barriers to its uptake and how these barriers might be addressed.
This blog post gives a few highlights from the discussion.
What could machine learning do for manufacturers?
Machine learning algorithms can detect patterns in data, and use these to make predictions. In manufacturing, it can support functions such as:
- Automation: machine learning systems are not always given an embodied form, but they can be used in robotics, creating systems that can undertake complex tasks as part of production lines.
- Optimisation: the analysis provided by machine learning systems can be used to optimise processes, products or prices, and increase productivity as a result.
- Predictive maintenance: the patterns detected by machine learning systems can be used to predict which bits of equipment or installations are going to fail, before this failure occurs, so that maintenance efforts can be directed accordingly.
- Personalisation: machine learning algorithms can analyse data about customers and their preferences, and use this to create personalised products.
What needs to be done for the sector to make use of this technology?
Our workshop identified key barriers in data, awareness, skills and standards, which are holding back progress in using machine learning in the manufacturing sector.
As we’ve discussed on this blog before, data is crucial for developing machine learning algorithms. Further machine learning activity could be catalysed by companies collecting useful and usable data, and by making this data available. This requires developing infrastructure for data collection and data sharing, though commercial pressures mean that simply releasing data is not always desirable. The latter of these, as the Royal Academy of Engineering’s and Institution of Engineering and Technology’s report on Connecting Data noted, can be addressed by methods which do not rely on making data open.
A mutual lack of awareness and understanding might also be holding back the widespread adoption of machine learning methods in manufacturing. The manufacturing sector has not been an obvious destination for people working in machine learning, who might not see the types of challenges that companies in this sector are working on. Meanwhile, manufacturing companies can struggle to identify how machine learning might help deliver products or services, with machine learning tools appearing inaccessible or unavailable. Increasing access to tools or expertise, or creating opportunities for people to meet, could help bring these two communities together.
Along with a number of sectors, manufacturing suffers from a relative scarcity of people with machine learning and data science skills. Skilled people are needed to handle and curate data and to develop machine learning methods for its analysis. As these machines become a routine part of work, people working in manufacturing will also need the skills to enable productive human-machine collaboration.
Some machine learning algorithms are ‘black boxes’, whose results are accurate, but whose methods are difficult to interpret. This raises questions about the standards needed to deploy machine learning systems: how can a machine learning process be certified? Would these algorithms need to show their working, or that they had been sufficiently trained? Could ISO standards be established? Being able to certify these processes could aid uptake and spread of this technology.
Could this be the start of a new story for UK manufacturing?
Countries around the world are developing new narratives for their manufacturing sectors, which reflect the challenges and opportunities for manufacturing in the 21st century and which help communicate a vision for the sector. We see this in Germany’s Industrie 4.0, Japan’s Robot Strategy and in the emphasis on advanced manufacturing in Singapore’s knowledge-based economy.
The UK has strengths in machine learning – in both academia and in business – and a long tradition of manufacturing. Could these underpin a new story for the UK manufacturing sector?
The Royal Society is currently carrying out a project on machine learning, which is considering the potential of this technology, and the challenges that come with it. You can read more about machine learning in this interview with Marcus du Sautoy.