This showcase demonstrates how Luma analyzes five different operational scenarios in a simplified starch processing operation. The following sections include a simple flow diagram, each of the five cases, and Luma's responses.
Luma uses the same prompts across all cases. It receives a detailed process description, example experience and lessons learned, and system prompts to perform its analysis.
<aside> ðŸ§
Feel free to share your feedback.
</aside>
Luma is in beta. She is constantly improving and may make mistakes.
Figure 1. Basic Flow Diagram defined above.
In each case below, Luma analyzes a different dataset from 22 transmitters (plus time) across the process.
Each dataset covers one hour, from 06:00 AM to 07:00 AM on October 10th, 2025.
Data points are recorded every 20 seconds, resulting in 180 continuous readings per transmitter. In total, each dataset contains 23×180 individual data points.
For realism and easy assessment, the data is visualized in different parameter trends; pressure, temperature, flow, and vibration.
Each case includes one potentially problematic drift in a parameter, identified as the Focus. Cases are designed so that no safeguarding alarms or switches are tripped, with details provided under Mechanism and Symptoms.
The goal is to show that Luma can detect even the smallest drifts before alarms activate. Each case includes a Conclusion from Luma and a link to the full unedited response.