Contexte Métier
In a manufacturing setting, IoT sensors monitor machine operating temperatures. Due to minor sensor inaccuracies, readings can be very close to a threshold. We need to classify temperatures into 'Normal', 'Warning', and 'Critical' states, ensuring that values just over a boundary are correctly categorized using a tolerance factor (fuzz).
Préparation des Données
Creation of a dataset with sensor temperature readings, including values that are exactly on, or slightly outside of, the defined range boundaries.
Copié !
| 1 | DATA casuser.SensorReadings; |
| 2 | INFILE DATALINES; |
| 3 | INPUT SensorID $ Temperature; |
| 4 | DATALINES; |
| 5 | S01 75.0 |
| 6 | S02 90.0 |
| 7 | S03 90.001 |
| 8 | S04 110.0 |
| 9 | S05 110.05 |
| 10 | S06 . |
| 11 | ; |
| 12 | RUN; |
Étapes de réalisation
1
Create a format library named 'SensorFormats'.
Copié !
| 1 | PROC CAS; |
| 2 | ACTION sessionProp.addFmtLib / |
| 3 | fmtLibName='SensorFormats', |
| 4 | replace=true; |
| 5 | RUN; |
| 6 | QUIT; |
2
Define a numeric format 'TempState' with a 'fuzz' factor of 0.1. This means values within 0.1 of a range's end will be included in that range.
Copié !
| 1 | PROC CAS; |
| 2 | ACTION sessionProp.addFormat / |
| 3 | fmtLibName='SensorFormats' |
| 4 | fmtName='TempState' |
| 5 | fuzz=0.1 |
| 6 | ranges={'low-90=Normal', '90<-110=Warning', '110<-high=Critical'}; |
| 7 | RUN; |
| 8 | QUIT; |
3
Apply the 'TempState' format and fetch the results to check how the fuzz factor affects categorization.
Copié !
| 1 | PROC casutil; |
| 2 | load DATA=casuser.SensorReadings outcaslib='casuser' casout='SensorReadings' replace; |
| 3 | QUIT; |
| 4 | PROC CAS; |
| 5 | TABLE.fetch / |
| 6 | TABLE='SensorReadings', |
| 7 | FORMAT=true, |
| 8 | vars={'SensorID', 'Temperature'}, |
| 9 | formatDef={name='Temperature', FORMAT='TempState'}; |
| 10 | RUN; |
| 11 | QUIT; |
Résultat Attendu
The fetched table should show the formatted 'Temperature' values. Sensor S02 (90.0) should be 'Normal'. Sensor S03 (90.001), which is just above 90, should be categorized as 'Warning' due to the fuzz factor. Sensor S05 (110.05) should be 'Critical'. The missing value for S06 should remain unformatted. This validates that the fuzz parameter correctly handles boundary conditions.