Science and psychology can drive production efficiency
Industry 4.0 encompasses a wealth of technologies all promising a business advantage, but how useful are they really?
When you understand your processes properly you can dramatically improve outcomes. This is at the core of research at the Centre for Intelligent Autonomous Manufacturing Systems (i-AMS) within Queen’s University Belfast, Northern Ireland. This multidisciplinary research centre, added to the university’s industry-facing Northern Ireland Technology Centre (NITC), totals 50 years’ involvement in advanced manufacturing in Northern Ireland.
i-AMS brings together academic teams from engineering, computer science, applied mathematics and psychology to address the challenges companies are facing around Industry 4.0. The team draws upon expertise in technical areas including data science, machine learning/AI, manufacturing design, control, automation, robotics, factory simulation, virtual reality and augmented reality, and explores how these can be employed to reduce manufacturing costs, improve quality and increase productivity.
The centre has defined its research around three broad themes. The first is soft sensing, prognostics and factory simulations. This theme is essentially concerned with achieving a data driven bottom-up understanding of manufacturing processes and systems. We are exploring how to provide real-time assessment of machine performance, tool health and energy consumption, and how to achieve fleet-level information sharing and integration in order to deliver enhanced process monitoring and real-time fault detection – ultimately achieve zero-defect manufacturing. In simulation, we are interested in the integration of data-driven and first principles modelling of complex manufacturing systems to deliver comprehensive factory simulation models and digital twins.
The second theme is loosely defined as flexible automation, and is concerned with developing much more intelligent and adaptable automation systems. Specific areas of interest to i-AMS are how to design effective and safe multi-robot and human-robot collaborative working environments, and how to employ virtual and augmented reality technologies to enhance productivity.
The third theme, autonomous and intelligent decision-making, can be thought of as taking a top-down approach, where the goal is to develop AI systems that can integrate data from upstream supply chains, from manufacturing processes within the factory, and from products in the field in order to achieve optimal decision-making, with regard to tasks such as maintenance scheduling, energy and waste reduction, quality control and production planning.
The i-AMS centre has been setup to be relevant to manufacturing as a whole. We actively engage with sectors including semiconductor, aerospace, polymer processing, construction and agri-food to ensure our research is informed by industry needs. In the food industry, for example, which is a huge part of Northern Ireland’s economy, as a result of the current issues with Brexit and concerns about access to low-cost labour there is increasing interest in embracing Industry 4.0 technologies, and in particular automation, to remain competitive.
For data analytics and AI to be a viable proposition, key requirements are having the data available and in a form that is convenient for analysis. This can be a big barrier for companies to engage with manufacturing informatics if they do not already have the infrastructure in place to automatically log data and integrate all data sources in one place.
Ideally, there would be someone in-house who has the expertise to work with and update data analysis models and tools as processes and products change. This requires capital investment as well as capability within the team, which presents challenges of cost and return on investment, particularly for smaller companies that do not have the scale to justify a dedicated data analytics resource in-house.
Beyond that is the problem that people are not necessarily aware of what data analysis can bring them, hence an important part of our engagement with industry is increasing the awareness of the value of data analysis.
Zero defect manufacturing
All industries are striving to achieve zero defect manufacturing – that’s the ultimate goal. Today, certain sectors are very close to achieving that goal through the adoption of advanced process monitoring and data analytics techniques. The technologies we are focusing on will enable the user to gain a much greater real-time awareness of the state of their processes.
The better processes are understood, the quicker it is to recognise when and where things are starting to go wrong, and therefore the earlier one can intervene and take corrective action. The outcome of this is zero-defect manufacturing, as well as higher quality products and less waste.
An example of this technology is soft sensing. In manufacturing, you generally assess the quality of the final product by taking samples and measuring them offline to verify that they have been manufactured to the required standard. This can be an expensive and time-consuming process so typically you measure infrequently and hence only have a limited, and often delayed, awareness of changes in the processes that need to be corrected. Instead, by using advanced data analysis and machine learning techniques, we can look for relationships between process variables that are monitored continuously, such as temperatures, pressures and flow rates within production systems, and the output quality metrics of interest.
Where these relationships exist and sufficient historical data is available, we can build models to predict the quality metrics for every single item produced – not just the sampled ones. Using this approach, which is referred to as soft sensing or virtual sensing, enables continuous monitoring of a process and as a consequence, more accurate control of product quality to be achieved.
Failure prediction is a developing area in manufacturing informatics. Process health monitoring involves the same principle as soft sensing – analysing the data that is continuously being collected of a production machine to see if it contains signatures that reflect its state of health.
If such signatures can be identified, we can observe how they trend over time and develop models to predict the time to failure, or machine breakdown. With the ability to predict the time to failure of machines, much more resource-efficient predictive maintenance strategies can be adopted.
Traditional preventative maintenance dictates that if a component has a life expectancy of say three-to-six weeks, it will get replaced after two-and-a-half weeks to avoid any potential issues. But this means that a component with the potential to have lasted for six weeks is being thrown away with half its life remaining – highly inefficient from a sustainability and resource utilisation point of view.
By moving to a predictive model, maintenance schedules can be optimised to maximise resource utilisation while taking account of the cost and availability of maintenance teams.
Show me the money
A number of larger companies have fully embraced Industry 4.0 because they understand the value of data and automation. We have been working with a diverse number of manufacturing companies from bakeries and carpet-makers to polymer processing and semiconductor producers, all wanting to improve productively and reduce costs. We have a lot of traction but there are still barriers to moving forward because of cost issues.
The big hurdle for most companies is justifying the investments to their accountants as the payback period can be difficult to predict for data analytics projects, and be greater than the normal two-to-three-year return on investment (ROI) target for automation projects. This is especially the case for SMEs whichvmay not have the capital for large investments, and therefore have to make changes incrementally, delaying the ROI.
There are various funding initiatives available at the moment within the UK to encourage companies to engage in Industry 4.0 R&D, such as the Industrial Strategy Challenge Fund. We are keen to make sure Northern Ireland is competitive in securing some of this funding to support our local industry and have been engaging with the Manufacturing Technology Centre and the catapult network in the UK to establish initiatives, making sure we have a regional focal point for such investments going forward. Also, the Belfast Region City Deal, which was approved in the Autumn Budget 2018, contains plans for an Advanced Manufacturing Innovation Centre (AMIC) in Northern Ireland, with funding committed for buildings, equipment and demonstration environments.
The vision for AMIC is stated as to ‘operate at the interface between academia and industry - accelerating new technology developments through the innovation phase and ensuring that real industrial challenges based on market need are solved through collaboration with the best university research’.*
This is the ecosystem and environment we are developing in Northern Ireland to support our local businesses as they embraces the wave of innovation that is Industry 4.0.
*You can read more about funding under the Belfast Region City Deal here: bit.ly/2TKGBa9