Base de datos Time-Series para la gestión de datos sensoriales en el mantenimiento predictivo de equipos de bombeo hidráulico en la producción de crudo.

Abstract

Industrial monitoring systems create large amounts of sensor data with precise timestamps, high frequency, and the need to store information for a long time. This project presents a Time Series Database (TSDB) architecture using PostgreSQL and the TimescaleDB extension to store, process, and analyze data from ABB sensors installed in hydraulic pumping motors. These sensors collect 22 mechanical, thermal, and electrical variables every 15 minutes, producing more than 64,000 records per month. The proposed model separates raw data using the raw and iot schemas, which helps with efficient ingestion, normalization, and organization of sensor information. TimescaleDB hypertables offer automatic partitioning and optimized indexing, allowing fast queries even on large historical datasets. Compression policies, retention rules, and duplication control reduce storage costs. The extraction, transformation, and loading (ETL) pipeline created in this project automates timestamp normalization, variable cataloging, and incremental loading. It ensures that each measurement is stored only once and validated correctly. The results show that a time series database is a scalable and reliable solution for managing industrial sensor data, providing query responses in milliseconds.