Industrial Data Summit 2019: Connected Products & AI

Industrial Data Summit 2019 was held in London’s Mary Ward House and saw attendees spend the day rotating between a series of 30-minute roundtable-based conversations, co-hosted by an industry expert and a world-class manufacturer.

This innovative format enabled attendees to take their pick from almost a dozen discussion topics, including: connected product innovation, digital skills, data security, predictive maintenance, and more.

Connected Product Innovation

My first of three afternoon roundtables was hosted by Suresh Daniel, data and architecture integration director at the world’s leading industrial thread business Coats, and Ruptesh Pattanayak, director of manufacturing solutions at Microsoft.

For all the talk about the connected factory and supply chains, how do organisations actually connect their various facets and processes in order to accelerate innovation and time to market?

Connected product innovation could be the answer. It accelerates product development by empowering manufacturers to design and validate products using the affordable, scalable power of the cloud, an approach which has several clear advantages, according to Microsoft’s Ruptesh:

“Products have already undergone a significant digital transformation,” noted Ruptesh, “leading to an incredible wealth of customer and operational data. This capability has dramatically increased an organisation’s granular ability to address specific customer needs with new products or solutions.”

At the same time, the manufacturing paradigm is changing drastically from a liner, disassociated model to a circular framework. Rather than having R&D at one end, separated from the customer at the other, a closed loop allows all parties to communicate with each other and for feedback/insights to be shared frequently, if not constantly.

The combination of these two shifts sees manufacturers under increased pressure to introduce new, better, smarter products at a much more rapid pace than before. However, those at the table noted how their current innovation processes were somewhat disjointed, complex and wasteful in terms of time, materials and labour.

Additionally, for many of the businesses represented, the bulk of their SKUs were either flagship products or highly bespoke, made to order items which created particular challenges for their businesses.

“When a particular flavour combination is your best-selling product and what you’re known for, how do you embrace change,” asked one food and beverage manufacturing director.

“Our clients demand customisation but that leads to greater operational complexity. It increases our number of finished SKUs and inventory of materials, and limits our intelligence regarding demand forecasting,” noted another participant.

Could a solution lie in giving an appropriate supply chain executive a seat at the table alongside manufacturing, engineering and sales?

Another issue raised was the support traditional manufacturing organisations need to collect the consumer data now available to them.

Compared to traditional focus groups, social media can deliver a far broader range of insights into customer experiences, product use, demand spikes and feedback analysis, but gathering this information at all, let alone in a time-frame short enough to leverage it to the greatest effect, is well beyond your prototypical manufacturer.

The same could be said for connected products, said Ruptesh. “Thousands of your products in use out in the field could be bringing terabytes of information into your organisation, but you need to have a strategy around where that data is ultimately going to be stored and used,” he concluded.

Artificial Intelligence (AI) versus Machine Learning 

Manufacturers are showing a greater  interest AI technologies as they try to make sense of rapidly increasing volumes and sources of data, not to mention incorporating more data-driven decision-making into their processes.

My second afternoon roundtable explored AI versus machine learning and was hosted by Sabine Mavin, senior software architect at multinational construction giant Laing O’Rourke, and Matt Armstrong-Barnes, chief technologist at Hewlett Packard Enterprise (HPE).

There are varying definitions, but the group’s consensus was that machine learning is a subset of AI that provides systems with the ability to automatically learn and improve from experience without being explicitly programmed. It focuses on the development of tools and algorithms that can access data and then explore, analyse and leverage this information themselves.

They can even learn from these experience and, through the connectivity of cloud, share that learning globally in a matter of seconds.

It wasn’t surprising to hear from attendees that accurate budgeting and cost-control, responsive planning and dynamic forecasting were areas where they could see AI having the most impact – and generating the greatest gains.

“You may want to improve the efficiency of your production process, build deeper customer relationships or improve your pricing accuracy,” noted one supply chain director, “but this can only be achieved through a more unified, accurate approach to forecasting, planning, simulation and scenario analysis. A 1% improvement here can generate substantial gains or cost-savings across the entire supply chain.”

Most businesses are at the ‘explore’ stage, unsure of where AI could benefit their organisation.

“If your dataset is vast, AI is very good at predicting and spotting patterns and would likely offer business value,” noted HPE’s Matt Armstrong. “A good starting point is either hook into  your operational technologies, or go over the top of them with something like video analytics for quality assurance.”

Employing someone to sit and monitor a dozen or more camera screens is a relatively low-skilled yet surprisingly labour-intensive tasks.. Naturally, that leads to get lapses in concentration and things being missed.

With the increase in cameras and the resulting number of screens, it’s just not effective to use human operators anymore. That’s where AI comes in. AI is a perfect candidate to undertake QA because the algorithms are  becoming ever-more sophisticated, and that’s driving dependable accuracy.

Systems are now able to say with high confidence whether a finished good meets the quality criteria or whether it must be rejected or reworked.

When selecting a potential ecosystem partner, Matt advised doing your research and choosing a technology vendor which shares your culture, has relevant industry or sector experience and can demonstrate previous successes.