# STIR refinery Flow System Health Monitoring — Data

来源：https://www.kaggle.com/competitions/fault-classification-in-industrial-oil-gas-flow/data

本地 Meta Kaggle 快照；不是本次实时页面。

# Dataset Description

## Training Unit A (Meter A)
**File:** `unit_A.csv`  
**Rows:** 87  
**Columns:** 37  

**Columns:**  
`feature_1` … `feature_36`, `fault_state`

**Fault State Distribution:**
- `Fault State 1`: 35 (40.2%)  
- `Fault State 2`: 52 (59.8%)

---

## Training Unit B (Meter B)
**File:** `unit_B.csv`  
**Rows:** 92  
**Columns:** 52  

**Columns:**  
`feature_1` … `feature_51`, `fault_state`

**Fault State Distribution:**
- `Fault State 1`: 19 (20.7%)  
- `Fault State 2`: 24 (26.1%)  
- `Fault State 3`: 49 (53.3%)

---

## Training Unit C (Meter C)
**File:** `unit_C.csv`  
**Rows:** 181  
**Columns:** 44  

**Columns:**  
`feature_1` … `feature_43`, `fault_state`

**Fault State Distribution:**
- `Fault State 1`: 54 (29.8%)  
- `Fault State 2`: 23 (12.7%)  
- `Fault State 3`: 54 (29.8%)  
- `Fault State 4`: 50 (27.6%)

---

## Test Set (Unlabeled)
**File:** `test.csv`  
**Rows:** 180  
**Columns:** 44  

**Columns:**  
`id`, `feature_1` … `feature_43`

---

## Sample Submission Template
**File:** `sample_submission.csv`  
**Rows:** 180  
**Columns:** 2  

**Columns:**  
`id`, `fault_state`

---

# Overall Statistics
- **Total training samples:** 360  
- **Total test samples:** 180  
- **Total dataset samples:** 540