Exploiting Big Data and analytics to improve productivity in manufacturing
This article was first published in Manufacturing Global
By applying lean tools and techniques and empowering the shop floor we have seen real improvements in manufacturing performance. However, with the rise of the connected and integrated shop floor, we are now seeing a new revolution in factory productivity. Data is at the centre of that improvement, with our work with a range of companies in life sciences, automotive and consumer goods, showing that effective use of data can bring 10-15 per cent improvements in productivity.
A leading life sciences company which runs a batch manufacturing process was experiencing a range of bottlenecks that were affecting productivity and overall equipment efficiency. By extracting and analysing multiple data sources ranging from ERP, machine, environmental and maintenance we could identify where they could find the next level of improvement opportunities. In particular, the data was analysed to identify the characteristics and optimum machine settings of a perfect batch in one of the bottleneck machines. By applying the settings and characteristics of this perfect batch to all batches, the company achieved a 15 per cent improvement in scrap and 2 per cent improvement in uptime. They are now scaling up this approach across other machines on the site and looking at the potential of applying it to their operations worldwide.
Better data can also help the automotive sector address the problem that, when it launches a new car, it can take weeks or even months before quality and warranty problems in the field are identified centrally. This is a major challenge for OEMs and can also lead to costly product recalls (example size of recalls). By collecting and analysing data from vehicles as they are launched and using it immediately to start to assess failures and feed this back to engineering and manufacturing they can deal with these quality issues before they become a big problem. That significantly reduces the cost of quality and warranty in manufacturing and their repair requirements.
These examples show the potential of data to address specific bottle necks and points in the manufacturing process or the supply chain. However, as the technologies and their capabilities improve, the next level of data-driven transformation will come from the ability to analyse the factory eco-system and even the supply chain as a whole. To help them do this, companies are already starting to build digital twins of their manufacturing processes and supply chains based on the data and processes they are running. Once these digital twins are in place then machine learning and AI can be used to test and optimise the operation in the virtual world before applying it to the real factory environment. The advantage of using a digital twin is that it can gain an understanding of all of the variables which drive inefficiencies across the whole process and find ways to optimise that process rather than just looking at an individual machine.
So how can a company start to use data driven manufacturing to change what they do? For us the mantra is “Think big, start small, scale fast.” Thinking big is all about developing the vision of what data can do to transform the factory and supply chain, and creating a clear picture to show senior leaders where it will bring value. Then, by starting small, businesses can prove the value of data through initial use cases. That then opens up the opportunity to scale fast by accelerating the delivery of the use case by making sure the processes, technologies and applications are in place to implement them in an effective and agile way, and giving the data scientists the inputs they need to scale up the solutions.
The main challenges in getting these projects off the ground are firstly to ensure that the structures are place to extract the right data to address the challenges, understanding that this is often not the obvious data. For example, factory environmental data, humidity and temperature can have an impact on machine set up and performance, but may not have previously been considered important. So the business will need to build the data collection mechanisms, put sensors in place, and provide the support cloud IT infrastructures and, at the same time, make sure the data is clean and accurate.
Secondly, like any data driven improvement it does not just happen by hiring a great set of data analysts. What is needed is a combination of data analytics capabilities, knowledge and experience of the manufacturing processes and the shop floor and importantly the ability to change the way the operators and engineers work. That can require a real adjustment in mindset from engineers as they move from being the experts to being part of a process which trusts the data to deliver the results.
Using Big data, analytics and machine learning will bring the next big wave of performance improvement in manufacturing, and will build on the achievements we have already seen from lean. The results and use cases derived from these technologies can be delivered in a matter of weeks, so the time to generate a return on investment is short. However, like any other significant change in operations they will require new skills and changes to existing capabilities and the challenges of putting these in place should not be under estimated.
However, the potential of data is clear. Manufacturers are already seeing tangible benefits and there will be much more to come for those who are willing to embrace the opportunity and put the right processes, culture and investment in place.