Equipment Maintenance Gap Analysis

Comparison of flagged failure records versus clean records across machine type, failure mix, and the sensor shifts that most clearly separate the two groups.

Records assessed
7,500
96.64% clean records
Flagged failures
252
3.36% of all records
Highest-risk machine type
L
3.83% failure rate
Largest sensor gap
+34.26%
Tool wear in failures
Gap summary

Flagged failure records are a small minority of the population, but the separation between clean and failure states is concentrated in a few signals rather than spread evenly across all sensors. Type L machines contribute both the highest failure count and the highest failure rate, making them the clearest operational hotspot in this sample.

Sensor means show the sharpest lifts in Tool wear and Torque, while rotational speed trends lower in flagged rows and the temperature fields remain close to clean-state baselines. The distribution views below show that the failure group consistently shifts toward higher wear and higher torque instead of simply containing isolated outliers.

A data-quality discrepancy also appears inside the flagged set: 7 records are marked as failures in the target flag while still carrying a No Failure subtype label. That mismatch is included in the failure-mix view as an unspecified label bucket.

Type L: 171 flagged records
Torque mean gap: +26.99%
Tool wear mean gap: +34.26%
7 flagged rows lack a failure subtype
Primary discrepancies to watch
  • Failure concentration: Type L runs at a higher failure rate than both M and H, indicating the strongest risk concentration by machine category.
  • Mechanical stress signals: Torque and tool wear rise sharply in flagged records, making them the clearest separators between failure and clean states.
  • Subtype reporting gap: A small portion of flagged rows still uses the No Failure label, so target status and subtype assignment are not perfectly aligned.
  • Temperatures are stable: Air and process temperatures move only marginally, suggesting they are weaker discriminators in this dataset.
Failure footprint
Dual donut: overall status and failure subtype composition
Failure rate by machine type
Bars show flagged counts; line shows failure rate
Sensor mean gap between failure and clean states
Percent difference relative to the clean-state mean
Torque distribution shift
Relative-frequency curves make the clean and failure shapes directly comparable
Tool wear distribution shift
Failure records cluster at substantially higher wear levels
Sensor comparison table
Sensor Clean mean Failure mean % difference