foundation-models/golden-batch-sentinel-data

---
license: cc-by-4.0
task_categories:
- time-series-forecasting
- tabular-classification
tags:
- bioprocess
- manufacturing
- anomaly-detection
- fault-detection
- batch-monitoring
- pharma
pretty_name: Golden Batch Sentinel Data
size_categories:
- 1M<n<10M
---

# Golden Batch Sentinel Data

Benchmark datasets for process monitoring and fault detection in batch manufacturing.

## Datasets

### IndPenSim (Industrial Penicillin Simulation)

A 100,000L fermentation simulation with 100 batches and rich multivariate signals.

- **Source**: [Mendeley Data](https://data.mendeley.com/datasets/pdnjz7zz5x/2)
- **Paper**: [Modern day monitoring and control challenges...](https://doi.org/10.1016/j.compchemeng.2018.05.019)
- **Batches**: 100 (90 normal, 10 faulty)
- **Variables**: 37 process variables (Raman spectra excluded for efficiency)
- **Time resolution**: 0.2 hours

**Files:**
- `indpensim/batches.parquet` - Main batch data
- `indpensim/statistics.parquet` - Batch statistics and fault labels

### Tennessee Eastman Process (TEP)

The most common benchmark for fault detection in multivariate industrial processes.

- **Source**: [Harvard Dataverse](https://dataverse.harvard.edu/dataset.xhtml?persistentId=doi:10.7910/DVN/6C3JR1)
- **Fault types**: 20 different fault scenarios
- **Variables**: 52 (41 measured + 11 manipulated)

**Files:**
- `tep/fault_free_train.parquet` - Normal operation (training)
- `tep/fault_free_test.parquet` - Normal operation (testing)
- `tep/faulty_train.parquet` - Faulty operation (training, all 20 faults)
- `tep/faulty_test.parquet` - Faulty operation (testing, all 20 faults)

## Usage

```python
from datasets import load_dataset

# Load IndPenSim
indpensim = load_dataset("foundation-models/golden-batch-sentinel-data", data_dir="indpensim")

# Load TEP
tep = load_dataset("foundation-models/golden-batch-sentinel-data", data_dir="tep")

# Or load specific files
import pandas as pd
from huggingface_hub import hf_hub_download

path = hf_hub_download(
repo_id="foundation-models/golden-batch-sentinel-data",
filename="indpensim/batches.parquet",
repo_type="dataset"
)
df = pd.read_parquet(path)
```

## License

The original datasets are provided under their respective licenses:
- IndPenSim: CC BY 4.0
- TEP: Public domain

This compilation is provided under CC BY 4.0.

## Citation

If you use this data, please cite the original papers:

```bibtex
@article{goldrick2019modern,
title={Modern day monitoring and control challenges outlined on an industrial-scale benchmark fermentation process},
author={Goldrick, Stephen and others},
journal={Computers \& Chemical Engineering},
year={2019}
}

@article{downs1993plant,
title={A plant-wide industrial process control problem},
author={Downs, James J and Vogel, Ernest F},
journal={Computers \& Chemical Engineering},
year={1993}
}
```

许可证：CC BY 4.0
任务类别：
- 时间序列预测
- 表格分类
标签：
- 生物过程
- 制造
- 异常检测
- 故障检测
- 批次监控
- 制药
数据集名称：黄金批次哨兵数据集（Golden Batch Sentinel Data）
数据规模：100万<n<1000万

# 黄金批次哨兵数据集（Golden Batch Sentinel Data）

面向批次制造过程监控与故障检测的基准数据集。

## 数据集详情

### 工业青霉素模拟数据集（IndPenSim，Industrial Penicillin Simulation）

该数据集为容积100,000升的发酵模拟数据集，包含100个批次样本与丰富的多变量信号。

- **数据来源**：[Mendeley数据集平台（Mendeley Data）](https://data.mendeley.com/datasets/pdnjz7zz5x/2)
- **关联论文**：[现代监控与控制挑战...](https://doi.org/10.1016/j.compchemeng.2018.05.019)
- **批次数量**：100个（90个正常批次，10个故障批次）
- **变量维度**：37个过程变量（为提升效率已移除拉曼光谱数据）
- **时间分辨率**：0.2小时

**数据文件：**
- `indpensim/batches.parquet` —— 主批次数据文件
- `indpensim/statistics.parquet` —— 批次统计数据与故障标签文件

### 田纳西Eastman过程数据集（Tennessee Eastman Process，TEP）

多变量工业过程故障检测领域最常用的基准数据集。

- **数据来源**：[哈佛数据文库（Harvard Dataverse）](https://dataverse.harvard.edu/dataset.xhtml?persistentId=doi:10.7910/DVN/6C3JR1)
- **故障类型**：20种不同的故障场景
- **变量维度**：52个变量（41个测量变量 + 11个操控变量）

**数据文件：**
- `tep/fault_free_train.parquet` —— 正常运行工况训练集
- `tep/fault_free_test.parquet` —— 正常运行工况测试集
- `tep/faulty_train.parquet` —— 故障运行工况训练集（包含全部20种故障）
- `tep/faulty_test.parquet` —— 故障运行工况测试集（包含全部20种故障）

## 使用方法

python
from datasets import load_dataset

# Load IndPenSim
indpensim = load_dataset("foundation-models/golden-batch-sentinel-data", data_dir="indpensim")

# Load TEP
tep = load_dataset("foundation-models/golden-batch-sentinel-data", data_dir="tep")

# Or load specific files
import pandas as pd
from huggingface_hub import hf_hub_download

path = hf_hub_download(
repo_id="foundation-models/golden-batch-sentinel-data",
filename="indpensim/batches.parquet",
repo_type="dataset"
)
df = pd.read_parquet(path)

## 许可证

本数据集收录的原始数据遵循各自的开源协议：
- 工业青霉素模拟数据集：CC BY 4.0
- 田纳西Eastman过程数据集：公共领域

本整合数据集遵循CC BY 4.0协议发布。

## 引用方式

若您使用本数据集，请引用以下原始论文：

bibtex
@article{goldrick2019modern,
title={Modern day monitoring and control challenges outlined on an industrial-scale benchmark fermentation process},
author={Goldrick, Stephen and others},
journal={Computers & Chemical Engineering},
year={2019}
}

@article{downs1993plant,
title={A plant-wide industrial process control problem},
author={Downs, James J and Vogel, Ernest F},
journal={Computers & Chemical Engineering},
year={1993}
}