In part one of the Artificial Intelligence (AI) and Simulation series, Chief Technology Officer at CFMS, Ian Risk, discussed the benefits that AI and simulation technologies bring to manufacturers of high value products, calling on industry leaders to adopt them in unison to drive down costs, improve time-to-market and boost efficiency. In part two, Ian considers some of the challenges that AI and simulation bring to the digital engineering community and offers solutions for the manufacturers embracing these tools.
I’ve spent my entire career working around the digital engineering community, so I strongly believe there are clear opportunities and benefits for manufacturers to develop and adopt AI and simulation technologies together. However, I also believe there are a number of challenges inherent with the implementation of these two technologies. Important questions that need to be addressed by myself and others working in the digital manufacturing space include; how can we ensure that simulation is an accurate reflection of real-world processes? To what extent can key manufacturing decisions be based on simulated data? What measures should be put in place to ensure these technologies are accurate and reflect ethical practices carried out by a human workforce? What happens if something goes wrong? Ultimately these issues boil down to a question of trust.
Drawing on my experience with companies that are working to combat the challenges associated with the adoption of digital technologies, it is clear that the best pre-emptive solution is having a highly reliable set of data from which insight can be derived. Cultivating a strong dataset that captures information throughout the lifecycle of a product is the overarching requirement in order to gain insight holistically, avoiding a siloed approach to engineering or manufacturing. Ensuring a seamless flow of data through many tiers of a supply chain is one of the biggest challenges high value manufacturers face, therefore it is important to form the basis of a digital thread from the data a product generates.
At CFMS, we are uniquely positioned to work with manufacturers to create this digital thread and build the communication framework that allows a connected data flow and integrated view of an asset’s data. Simulated data can form the beginning of this process, but taking steps to capture data through the life cycle of a product ensures that these virtual models of systems, the so-called ‘digital twins’ are validated and improved, consequently building trust in them. A manufacturer able to harness a large and reliable set of data is already taking steps to prevent any barriers they might face with industrial digitalisation.
Ultimately, the complete supply chain will have to adapt and production lines will need to evolve to create the assured and coherent dataset on which autonomous decisions can be taken. This is going to take time and will take a concerted effort of all stakeholders to make this happen. National initiatives such as Made Smarter will be a vital mechanism in spreading awareness and demonstrating capability to bring about change.
In advance of this, CFMS is already working closely with manufacturers to create bespoke ‘digital backbones’ ensuring the data they capture on the production line is aligned to the needs of AI and simulation. As experts in the digital engineering field, a major challenge we experience is ensuring organisations understand what types of data they need to capture, at what frequency and how to transmit this cost effectively and, most importantly, the problem the organisation is trying to solve.
Simulation enables us to better understand real-world problems safely and efficiently by providing a strong initial dataset from which an engineer can make an informed decision. However, if the intent is to use an AI system to monitor quality or predict maintenance needs, this will only be as good as the data it has been trained on and simulation alone cannot provide this. A continuous flow of real-world, balanced data from a smart, connected factory is therefore a prerequisite to harness the potential of AI.
As members of the digital engineering community we must act responsibly and consider the ethical implications of using simulated data to train AI systems. Running simulations can be an incredibly effective way to collect data quickly and efficiently, saving money and time in the long run. However, if this data is derived solely from virtual simulations without a physical prototype, this brings into question the validity model and how much it can be trusted. All sources of data, virtual or real, must be assessed to make sure they are completely sound and free from bias. Data used to inform an AI system needs to be ethically sourced in order for a company to be compliant. An auditable trail needs to be created to pass certification and ensure manufacturers have followed due process throughout the lifecycle of a product. This is because the liability can never lie with the technology itself, but with the manufacturer or supplier that trained the system with the data. Rigorous testing needs to be carried out by all manufacturers before adopting AI and simulation to make sure the outcomes are not inadvertently biased or corrupted.
Looking to the future, it will be imperative for manufacturers of high value, complex products to not only embrace AI and simulation as a way to cut costs and drive efficiency, but to consider the challenges that adopting these technologies bring. The overwhelming consensus in the digital engineering community is that the better the data manufacturers can arm themselves with, the better enabled they will be to combat any challenges associated with adoption. As experts in the field of digital engineering, we are uniquely positioned to support manufacturers in this process helping them to create a strong data set and therefore a solid platform for both AI and simulation to thrive.