Digital Twin with Python (VTR UGE 21), day 5 of 5

Visualization and Capstone Development

Prof. Dr. Utku Kose, Süleyman Demirel University

Operations lab

Part A runs the assembled twin in an operations panel with faults you inject. Part B separates verification from validation. Part C estimates the model error behind a validation result.

Part A: The operations panel

The panel runs the pack and its twin for two hours in the browser. Inject a fault at a chosen time and watch the twin: It compares every reading with its own physics prediction, a model temperature that no reading ever corrects. When the mean of their differences over the last five minutes exceeds 1 °C in either direction, the twin declares the loss of trust, switches the fan to its fail-safe setting, full speed, and reports the change in the event log. The log can be exported as JSON lines, one record per line in the JavaScript Object Notation, the format of the notebook.





Open in ColabContinue in Colab, section 7: the assembled twin on a healthy and a faulted pack.
Part B: Verification or validation?
Decide which question each check answers.
Open in ColabContinue in Colab, section 5: verification by step halving and validation against noisy data.
Part C: The model error behind a validation result
The error between twin and sensor is measured by the root mean square error (RMSE). With a sensor noise of standard deviation sigma, the error of the model itself is estimated as the square root of RMSE squared minus sigma squared. All values are in °C.
Open in ColabContinue in Colab, section 8: the scorecard of your own asset with a worn fan.