Predictive Operation & Maintenance Failure Application

This application uses machine learning trained on historical data to predict equipment status based on current sensor readings.

The application provides:

• Current operational status
• Failure probability
• Key drivers contributing to potential failure
• A comparison between trained data ranges and current sensor readings

The historical dataset must include a failure label as the last column, with values 0 (normal) or 1 (failure). A sketch of feature importance explaining the main failure drivers can be included by selecting the “Include feature importance” checkbox. 24-hour rolling mean features can be enabled using the “Use rolling features (24h)” checkbox to capture short-term trends in sensor behavior. This helps determine whether abnormal readings represent a sustained trend or temporary spikes.