How do you build a digital twin?
Digital twins are complex creations (read the previous blog “Just what is a digital twin – and how can it help me?”), so they are usually generated and operated by specialist companies as part of their analytics offering. These companies not only take care of the whole lifecycle of the digital twin, they also provide data collection platforms onboard as well as applications, dashboards, and tools that utilize the digital twin’s capabilities. So how does your shipping company go about getting one built?
Finding the right expertise
Your organization has probably already recognized the value of data and digitalization, but is lacking the internal competencies to exploit it fully. By outsourcing the data processing and analysis to an analytics company that specializes in data collection and visualization using digital twins, you can concentrate on putting the insights they provide into practice.
However, it’s important to remember that you need the right expertise in order to recognize the right partner to work with. Understanding the lifecycle of a digital twin will help you understand what capabilities you require from an analytics company.
Developing the toolkit
The lifecycle of a digital twin can be divided into three main stages: development, generation, and operation. Each stage requires a different skill set from employees and typically happens in a different part of the analytics company. The development stage includes developing the toolkit that can be used to generate unique digital twins for different assets. The toolkit also includes the capability to process data using commissioned digital twins.
Developing the toolkit is the most challenging part in a digital twin’s lifecycle, and is typically handled in the R&D department by highly specialized data scientists with a degree in applied mathematics, machine learning, statistics, or similar. Development work includes gathering information about the underlying physics of the asset, about the operating environment, and about any constraints under which the asset is operated. In addition, automatically collected data from the asset is needed to develop the machine learning methodologies required. The outcome is a mathematical model that can be used to simulate the asset’s behavior and performance.
Generating the digital twin
The generation stage is where the tools developed in the previous stage are used to create a unique digital twin for each physical asset. First, the digital twin is adjusted to match its physical counterpart. This process ensures that the data collected from the vessel fulfills the quality requirements of the digital twin and that the mathematical model created is of good quality.
In an established analytics company, digital twins are generated by specialized engineers with sufficient understanding of statistical methods and the necessary programming skills to troubleshoot any problems that may arise.
Operating the digital twin
Once the initial version of the digital twin has been generated, it can be run in production. This means that the digital twin continuously enriches the data feed coming from the ship to the cloud, augmenting the raw data with insight provided by the modeling. For example, a digital twin of an engine can continuously provide insights on current performance and pinpoint non-optimal performance. Running the digital twin in production happens as part of the data processing chain in the analytics company’s cloud service. The quality of insight provided by the digital twin is continuously monitored and any necessary corrective actions taken. Monitoring may reveal that a new version of the digital twin needs to be generated.
Some of the more sophisticated digital twins can also learn to self-adjust to gradual changes in the physical asset. For example, a digital twin of a vessel’s hull and propeller may adjust to changes in performance over time, and be able to quantify these changes.