Data_1

The PdM_telemetry Dataset (D_1) is a synthetic dataset designed to support predictive maintenance (PdM) research for IIoT (Industrial Internet of Things) devices by providing sensor-based telemetry data. This dataset initially comprises 97,210 records and 30 features, including a binary target feature, 'failure', which indicates whether a device will fail within the next 24 hours. The remaining features, such as device operational metrics and error counts, serve as predictors. Data preprocessing steps involved creating additional features from timestamp values and consolidating error count indicators to streamline analysis. Due to class imbalance, with only 0.70% of samples classified as failures, the dataset was balanced using the SMOTE algorithm to achieve equal representation of failure and non-failure cases. Further enhancement was performed by generating 38,085 synthetic samples using the Gretel.ai platform, enriching the dataset's variability and relevance for machine learning model training. This dataset is intended for developing and evaluating predictive models for IIoT device failure, contributing to more accurate and timely maintenance interventions.

PdM_telemetry 数据集（D_1）为一项合成数据集，旨在通过提供基于传感器的遥测数据，支持工业物联网（IIoT）设备预测性维护（PdM）研究。该数据集最初包含 97,210 条记录和 30 个特征，其中包括二元目标特征 '故障'，用以指示设备是否将在接下来的 24 小时内发生故障。其余特征，如设备运行指标和错误计数，则作为预测变量。数据预处理步骤包括从时间戳值中创建额外特征以及整合错误计数指标以简化分析。鉴于类别不平衡，仅有 0.70% 的样本被分类为故障，因此采用 SMOTE 算法对数据集进行平衡，以实现故障与非故障案例的等量代表。此外，通过使用 Gretel.ai 平台生成 38,085 个合成样本，进一步增强了数据集的变异性，并提升了其与机器学习模型训练的相关性。本数据集旨在开发与评估 IIoT 设备故障的预测模型，有助于提高维护干预的准确性和及时性。