We explore how the combination of identification technologies, data capture and analysis can underpin supply chain transformation. By linking computer networks to sensors almost any information about an individual product or component can be obtained in real time – from its temperature to when, how and where it was made. But the value of data is in how it’s used. Worldwide traceability of components and finished goods can be used to spot product failures, support new service offerings or meet legislative requirements. By analysing the information, identifying patterns and creating algorithms we can support the development of smart, flexible and responsive manufacturing systems that are more productive, more efficient and better at meeting customer needs.

Our research into digital manufacturing can be clustered into three main – and connected - strands: data capture, data analytics and management, and intelligent systems.


Getting better data

If data analytics are going to be useful, we need to make sure that we capture good quality information that will be susceptible to analysis. But the quality of the data is often problematic.

It tends to arrive unstructured, in different formats, at different timescales and riddled with errors.

Data capture has been one of the primary focuses of DIAL research, initially through the development of RFID technology and increasingly through sensors to extract fine-grained traceability data from industrial environments. In 2000 DIAL became the Cambridge partner in the Auto-ID Labs, a group of labs from seven of the world’s leading research universities which pioneered research in RFID. As a result of this work, we have taken part in several EU projects looking at how item-level data improves food supply chain and manufacturing traceability, asset condition monitoring, and recycling operations.

We are currently working on an ‘intelligent data’ concept to help manage uncertain and volatile information. For example, when pickers in a warehouse misplace products, data in the system becomes increasingly disconnected from reality, leading pickers to the wrong locations when items are required. Using a combination of real-time shelf data and algorithms, the quality of data in the system is tracked to help pickers avoid or discover and correct these misalignments. We are also looking at tools and techniques for cleaning, combining data from multiple sources, and evaluating data sets for purchase.


So much data – but what does it all mean?

Data analytics is the science of studying data to uncover and interpret hidden patterns and trends. Through data and analytics we aim to unify the whole manufacturing system, including the extended supply chain and associated logistics, as well, as external factors such as weather and market conditions. This will enable manufacturers to get a much clearer picture of how they interact with their external environment, improve their efficiency and achieve competitive advantage.

As supply chains become increasingly complex, involving many different companies often scattered around the globe, it has become more and more difficult for everyone in the supply chain to keep track of products and components. This creates risk. But a manufacturer often does not know about a disruption until it hits them or one of their tier 1 suppliers. While RFID and sensors can be helpful in improving visibility along the supply chain, they only work if everyone in the supply chain is using them.


Sharing knowledge with partners about real-time problems across the inbound and outbound supply chain is one thing but there are other external factors which can disrupt supply – and for which publicly available data exists. We are developing a new analytical tool which combines data from, for example, social media and newsfeeds about things like weather and traffic conditions, with internally available supply chain delivery data. By combining this data we can find correlations that can help predict disruptions so that the factory can be continuously adapted to mitigate predicted changes.


‘Big data’ is one of the pillars of digital manufacturing but it is not a panacea.


Applying analytical methods to manufacturing is often a ‘bespoke’ task which needs detailed knowledge of the context in which it is being used in order to be effective. This means new methods need to be developed from scratch for each application. But these industry-specific analytics are much more successful at identifying dependencies in supply chains than black-box, machine-learning approaches.


Our aim, ultimately, is to connect every step of the manufacturing process across all phases of its lifecycle, and across its geographic and industrial boundaries, to create self-adapting, resilient manufacturing systems.


To achieve this, we need to understand which analytics approaches work best under what circumstances.

Making systems smarter

Our early work in RFID led to the pioneering idea that data from individual objects could be identified and exploited using object identification and internet technologies. Over time that concept has, of course, evolved into the ‘Internet of Things’. During the last two decades, DIAL researchers have developed a number of IoT systems that allow companies to monitor their products in use. The EU PROMISE project, for example, was an IoT system that enabled products to be recycled back into production at the end of their lives and resulted in one of the first proof-of-concepts supporting the circular economy. Before that DIAL had been involved in the development of modular ‘intelligent’ manufacturing operations which challenged conventional ‘command-response’ approaches to controlling operations to more flexible strategies where machines ‘talked’ to each other to determine the best control approach. We worked with an AI technology – intelligent software agents – to give machines the thinking power we wanted to achieve.

We have also designed an IoT system that allows engineering assets to monitor their condition, talk to each other to create batch orders, and negotiate with selected suppliers to order parts autonomously. The system eliminated communication bottlenecks that arise from manually placed standard orders. We are currently pushing the boundaries of the IoT concept by developing a ‘Social Network of Things’. This enables machines to report their ‘status’ into a common data-sharing platform analogous to a social network. By doing this we can create a single view of how the whole factory – or production network - is running. Algorithms that

run on the social network platform look at these updates and determine the best maintenance plan for the factory or network as a whole, rather than maintaining every machine on a reactive basis.

How we capture and manage data and how we turn it into actions through algorithms – these are at the heart of what we do at DIAL. In recent years, we have seen our pioneering work tend towards the mainstream. Most large manufacturers have already embarked on the digital journey or take it for granted that they should. But our work is by no means done. Putting current thinking into practice is challenging enough but with technologies – and business models – changing so fast, we need to be exploring the next generation of digital thinking.

IfM Review Issue 6 Articles
Digital Manufacturing at the IfM
the digitalisation of manufacturing economies

The digitalisation of manufacturing economies

How bright is your digital future?

A social network of things

Getting smart with digital

The importance of standards in a digital world

The importance of standards in a digital world

Manufacturing leadership in the age of digital disruption

Manufacturing leadership in the age of digital disruption

A disruptive influence