­Physics-Based Model Provides More In-Depth Analytics

The global supply chain was massively disrupted during the COVID-19 pandemic, which has prompted countries and trade organisations, such as the EU, to try to bring supply chains closer to home to safeguard their industries in case of other interruptions. More disruptions seem likely with Russian invasion of Ukraine and the USA’s trade war with China has increased doubt that the global supply chain will ever recover its pre-pandemic levels.

As well as a more robust supply chain, a further benefit of onshoring is that it creates well paid jobs at home. However, the downside to this is that the initial reason for migrating the jobs, was to move to countries with lower labour costs. Consumers have become used to the cheap goods that come from overseas manufacture, and with an ongoing cost of living crisis caused by rising interest rates and inflation, people will not be willing to pay more for products, no matter where they are manufactured. With no cheap workforce, and customers unwilling to spend more, manufacturers are left with no other option than to increase the efficiency of their processes.

Apart from government offered incentives to relocate, which help in the short term, manufacturing companies are only now considering the move back to developed countries as automation is available to help create the efficiencies that they need to minimize the difference in labour costs. Industry 4.0 has also been developed as a way to further improve efficiencies in operation. By monitoring the process in real time, an Industry 4.0 implementation can show the performance of the system in real time and also allow for planned maintenance. Industry 4.0 has been taken a stage further by the digital twin concept, where a digital model of the system can be used to predict the performance of a system when changes are made.

The digital twin concept is great in theory, but getting it to work is a bit more difficult. Most current ways of implementation rely on statistical modelling, which is based on historical data. Statistical modelling is great for identifying patterns in large quantities of data and therefore gives a good overview of the whole system and the outcome of any process, but it doesn’t give provide the details that can be vital to interpreting those results. One company that has tried a different approach is Industrial Analytics. Instead of relying on purely statistical data, the company combines it with a physics-based hybrid model that can enable accurate digital twins that represent dynamic equipment. The hybrid machine learning model is base on first principles, giving a deeper insight into what is happening in the underlying system and allowing data interpretation down to the component level. This method makes it possible to understand the root causes of any malfunction in the system. The solution can be implemented in the cloud, edge or a hybrid of both, and has been adopted in companies working in the oil, gas, chemicals and utilities industries.

Industrial analytics was purchased by Infineon last year. Infineon will be both a provider and a consumer of Industrial Analytics technology. The company will continue to provide the software to other manufacturers and assist with its integration, as well as implementing the solution into its own offerings, both to increase the efficiency of its own processes, and to provide analysis that will improve the company’s products. For example, it can be used to analyze how IGBT modules degrade over time to more accurately predict useful lifetimes, or predict how they will react to different events.