Calibrating Digital Twins With Operational Data Optimizes Efficiency
Data feedback loops create a closed-loop digital twin (CLDT) driving increased operational efficiency and improved design iteration progress.
Digital twins are virtual replicas of physical assets in live operations. The virtual twin residing on a cloud computing platform uses the real-time operational characteristics fed to the cloud computing infrastructure. This is done with the help of IIoT devices and low latency networks. We have discussed digital twins in detail in the articles given below. Read the articles to brush up on the concept of digital twins.
- Digital Twinning and its Use in SCADA Systems
- Digital Twinning: Leaders in the Emerging Technology
- Is Reality Capture the Next Step in Digital Twinning?
In this article, we discuss the closed-loop digital twin (CLDT) concept and how it can be used to drive efficiency.
Closed-loop digital twins (CLDT)
The true potential of digital twins can be extracted only with closed-loop digital twins. Let us understand the key differences between open-loop digital twins and closed-loop digital twins.
Open-loop digital twins receive data from real-world machines and run the virtual replica of them. The insights gained from running the virtual twin are not used to improve the operations of the original physical machine. The information flow is only in one direction, from the physical machine to the digital twin. The virtual models only act as digital replicas and do not give any feedback to the machines in the plant.
CLDTs also receive data from real-world machines and run the virtual replica of them. But the insights and analysis obtained from virtual twins are used to improve the operations of the physical machine. This is termed ’feedback’ and is used to improve the operations of the machine. Data flows in two directions for CLDTs. From the physical machine to the digital twin and vice-versa. Almost all instances of digital twins with practical applications are designed as CLDT. This feedback loop is represented in the image given below.
Figure 1. Contrasting data flow paths between digital twins without and with a feedback loop. Author’s image
CLDT to improve operational efficiency
Machines that are replicated with CLDT are part of larger plant operations. The insights generated from the CLDT are used to improve machine operations. Production and operations managers work to deliver and maintain optimal output from the machines in the plant. These metrics are tracked by key performance indicators (KPIs), but KPIs do not always move in the same direction. For example, when production time is minimized, the energy used might increase; process changes nearly always result in cascading effects.
The CLDT can be used to perform simulations with real-time operational data. Simulations can be done to optimize multiple KPIs. A large number of simulations can be performed on the virtual twin and the generated results can be analyzed. The various KPIs can be calibrated to test various configurations. The output of all simulations can be used to determine the settings that will deliver optimal operation before real-world machines are adjusted.
Figure 2. Simulation of automation cells can be used to predict design configurations as well as gain real-time feedback for KPI calibration. Image used courtesy of Siemens
Calibrated CLDT helps to find the best configuration of KPIs that suit the overall requirement of a factory, plant, or organization. When situations change, new calibration specs must also be calculated and adjusted, usually a time-consuming manual process. Calibrating the CLDT can be scaled and automated with the help of machine learning (ML), greatly boosting efficiency.
ML to calibrate CLDT
Machine learning uses data to build models. Complex statistical analysis can be performed with these models to obtain a desired output. Machine learning models can be created with the operational data of digital twins. The calibration of CLDT can also be built as a machine learning model, in turn automating the optimization of KPIs This makes the process faster with minimal human intervention. The desired KPI outputs can be specified and the machine learning model will take care of figuring out the configurations that realize the specified output.
Machine learning also can be employed to perform simulations on the fly. Continuous analysis can be performed with the help of ML. Any changes in parameters can be instantly relayed to the plant to change the machine settings. CLDT with ML can also be used to test prolonged operations in mere minutes. The power of ML can ingest a vast amount of historical and live data to model long-term machine operations in minutes. This capability is incredibly useful in simulating various scenarios that can improve operational efficiency.
Digital twins for design
Digital twins can also be used for improvements in the design phase. Consider a jet engine manufacturer that runs digital twins. Any operational changes required to improve efficiency can be applied on the fly. But digital twins can also help identify physical designs that prove to be a constraint. These cannot be changed within one design iteration of the engine, but it can certainly be addressed in the next.
Figure 3. CLDT allows software and hardware to work simultaneously, resulting in more robust designs and operations. Image used courtesy of Siemens
Digital twins are great assets for finding ways to improve machine designs. Another added benefit is that a new design can be tested with digital twins. Real-world characteristics can be applied to the new design to test its efficacy. This ensures that the new design is optimal in every possible way. Since testing is done with real-time data, there are minimal chances of failure or errors when the machine is commissioned.
Digital twins also make it easier to run experiments on new designs. Machine learning algorithms are a great aid in this regard. Thousands of experiments can be automated with ML and digital twins. This gives a scale of experimentation that was not possible before. This, along with the ability to use real-time operational data, helps digital twins deliver the most efficient designs in a cost-effective manner.
Digital twins are a powerful technology that can improve many facets of manufacturing and operations. Closed-loop digital twins are required to make use of the complete potential of the technology. CLDT combined with machine learning opens many possibilities that were not accessible before. The extensive experimentation possible helps to deliver operational improvements to machines in operation. It can also be helpful in delivering improvements in the design phase of new machines. Overall, calibrated digital twins will help to change paradigms in the new industrial era.