Pre-verification of facilities with digital twins
Shortening the testing time for pre-FAT and SAT
Expansion of application of digital twins from 2020
Repeated regression tests, saving time
사진 확대 Jonas Larsen, a Novo Nordisk operational technology (OT) expert, and Poul Kristensen, a project binder customer relations director (left), are conducting the session. Reporter Park Sung-bae
A method is being introduced in which artificial intelligence (AI) reads the test procedure and operates it on a digital twin first before introducing new facilities into pharmaceutical factories. Novo Nordisk is reducing the time it takes to verify facilities by establishing a system that automatically repeats tests in a virtual environment prior to verification using actual production facilities.
Novo Nordisk unveiled an example of field acquisition test (SAT) automation using digital twins and AI at Lockwell Automation’s “Automation Fair 2026” in Boston on the 17th (local time). The SAT is a process of on-site verification to ensure that newly installed facilities work according to customer requirements and specifications.
Jonas Larsen, a Novo Nordisk operational technology (OT) expert, and Poul Kristensen, a project binder, participated in the presentation.
Novo Nordisk has been expanding its use of digital twins since 2020 at its Hileröd production facility in Denmark. Initially, the connection between machines and automated systems or other core systems was tested in virtual environments. In 2024, the digital twin was used for facility verification, and this year, it has advanced to the stage of automating pre-launch tests by combining AI.
The key is to turn human-written test procedures into a form that AI can actually execute. For example, if you enter the sentence “Press the emergency stop button after the pallet passes through the light curtain and check whether the reset button blinks,” the AI divides it into conditions, actions, and results to be checked. After that, the electrical document of the facility and the information of the controller (PLC) are compared to find the necessary signals and run the test on Lockwell Automation’s ‘Emulate 3D’ (Emulate 3D).
However, it does not leave all judgment to AI. This is because several facility signals can be connected to one expression. In the actual test, six candidates were searched for the expression “light curtain,” but it was difficult to determine which device was related to the palette by AI alone. In this case, the “Human-in-the-Loop” method is applied in which the engineer checks the correct signal in the middle.
사진 확대 Case data for automated field acquisition test (SAT) using digital twins and AI revealed by Novo Nordisk. ChatGPT
It also confirmed its effectiveness. According to the announcement of the session, the test, which took 2 minutes and 57 seconds manually, was reduced to 1 minute and 10 seconds in the automation system, reducing the test time by about 60%. The test run speed improved from 60% to up to 240%, depending on the conditions.
In addition, it was suggested that the test once made can be saved and repeated under the same conditions. This is because there is no need to create a new test in the regression test that is verified again after the facility change, so the more it is repeated, the greater the time saving effect. “The biggest advantage is that you can continue to repeat the same test without human intervention,” Christensen said in the session.
It also emphasized the speed of expansion after its construction. “It took months to create the framework itself, but it’s basically a matter of minutes to connect data points once the framework is in place,” Christensen said in a Q&A.
Ultimately, the goal is to reduce the time required to verify facilities by stopping actual factories. If you do most of the factory acquisition tests (FATs) and SATs in advance on digital twins, you can find errors that can occur when installing or changing actual production facilities in advance.
Another advantage of the digital twin is that the test results remain a record and can be tracked before and after the change. Digital twins are expanding beyond design tools to AI-based verification methods in manufacturing sites where facility verification and record management are important, such as the pharmaceutical industry.
[Boston Reporter Park Sungbae]