Machine Scrap Prediction Analytics
A Power BI report on machine data that shows when scrap goes up and which operating conditions come with it.
The problem
Machines log a lot: temperatures, pressures, vibration, warnings. When a line produces more scrap, the reason is usually somewhere in that data, but nobody has time to dig through it. Standard reports show how much scrap there was, not when it is likely to happen.
What I built
- Data preparation. Combined several machine data sources into one model of machines, log messages, log levels and time.
- Features. Picked the signals that matter for quality: temperature, pressure, vibration and the machines’ own warnings and errors.
- Prediction. A model that estimates the scrap probability from the current operating conditions.
- Report. An interactive Power BI report with filters for machine, log level, temperature and time, KPI tiles for scrap, warnings, average temperature and pressure, and a time series of scrap against temperature and vibration.
What it shows
- Scrap, warnings and productivity per machine
- How scrap moves with temperature and vibration over the day
- Scrap broken down by log level (info, warning, error)
- Drill-down to a single machine, period or condition
Try it
The live report is embedded below.
Stack
Power BI, with big-data management techniques for preparing the machine data.