Synthetic IoT Data Generation using HPC Infrastructure

Experts: Altan Koçyiğit, Erhan Eren, Kerem Nazlıel

Infrastructure Expert: Kerem Kayabay

Turkey Petroleum Refineries Corporation (TÜPRAŞ) operates four refineries in Turkey, controlling most of Turkey’s crude oil refining capacity. Many heterogeneous units in the field produce time-series Internet of Things (IoT) data. Among the factors that cause heterogeneity are various brands, tasks, degradation levels, and configurations. These units generate more detailed alarm data (e.g., frequency, action) or process data (e.g., key-value pairs) logged in lower granularity. The company wants to utilize these datasets for multiple purposes, such as alarm filtering and root-cause analysis. However, the industrial and heterogeneous nature of these units, varying granularities in datasets, and missing periods cause problems in data analysis. Within this project’s scope, TÜPRAŞ will try to generate synthetic IoT data leveraging HPC resources to mitigate issues caused by data quality.

Please click here to view the results of this study.