PROMECE - Big/Smart Data - Report of data analytics techniques for predictive maintenance in the industrial sector

Instituto Tecnológico de Informática has a non-economic Activities Plan (PROMECE) whose general objective is to strengthen the research lines in which the Institute works, within the scope of Information and Communication Technologies (ICT).

Within the Big/Smart Data R&D Line, we work on techniques to analyse large amounts of data, in order to obtain predictive models to detect anomalies in several scenarios. Another important topic is the visualization techniques to ease the analytics results transmission to the final users.

This document presents a handful of data analytics techniques for predictive maintenance in the industrial sector. First, we study the sensoring environment of the pilot, based on a real manufacturing plant. Then, we study the overall architecture for capturing the data from sensors, and analysing the signals to perform an initial EDA (Exploratory Data Analysis). The EDA allows us to identify the most representative variables and check the algorithms that are best suited for the objective of predicting a failure in machinery. Finally, a CEP system is proposed as the best solution to implement a rule-based system for supporting decision making, and some mockups are included, depicting how a generic tool for data visualization could look like.