Skip to main content

The Internet of Things (IoT) is creating a new structure of awareness – a cybernetic one – upon physical processes. Industries of different kinds are expected to join soon this revolution, leading to the so-called Factories of the Future or Industry 4.0. Our considered IoT-based industrial cyber-physical system (CPS) works in three generic steps:
1) Large data acquisition / dissemination: A physical process is monitored by sensors that pre-process the (assumed large) collected data and send the processed information to an intelligent node (e.g. aggregator, central controller);

2) Big data fusion: The intelligent node uses artificial intelligence (e.g. machine learning, data clustering, pattern recognition, neural networks) to convert the received ("big") data to useful information to guide short-term operational decisions related to the physical process;

3) Big data analytics: The physical process together with the acquisition and fusion steps can be virtualized, building then a cyber-physical process, whose dynamic performance can be analysed and optimized through visualization (if human intervention is available) or artificial intelligence (if the decisions are automatic) or a combination thereof.

We will focus on how to optimize the prediction, detection and respective interventions of rare events in industrial processes based on these three steps. Our proposed general framework, which relies on an IoT network, aims at ultra-reliable detection / prevention of rare events related to a pre-determined industrial physical process (modelled by a particular signal). The framework will be process-independent, but the actual solution will be designed case-by-case. We will consider the CPS working as a complex system so that these three steps, which operate with relative autonomy, are strongly interrelated. For example, the way the sensors measure the signal related to the physical process will affect what is the best data fusion algorithm, which in turn will generate a certain awareness of the physical process that will form the basis of the proposed data analytics procedure. As proof-of-concept, our approach will be applied to predictive maintenance in an automotive industrial plant from SEAT in Spain, in the Nokia base-station factory at Oulu and in the LUT laboratory of control engineering and digital systems.

Call Topic: Big data and process modelling for smart industry (BDSI), Call 2017
Start date: (36 months)
Funding support: 1 070 293 €

Project partners

  • Lappeenranta University of Technology - Finland
  • University of Oulu -Finland
  • Centre Tecnològic Telecomunicacions Catalunya - Spain
  • Sociedad Española de Automóviles de Turismo - Spain
  • Research and Education Laboratory in Information Technologies (Athens Information Technology) - Greece
  • Trinity College Dublin - Ireland

Main results

The main scientific results are:

  • General framework to approach the design of application-driven data acquisition, transmission, fusion and analytics towards a specific goal, which in FIREMAN is rare-event detection in industrial settings.
  • Demonstration of the efficacy of event-driven data acquisition to support fault-detection in industrial scenarios.
  • Fundamental studies in data mining that can provide high accuracy with explainable outcomes.
  • Theoretical contributions in data imputation methods to recover missing samples in datasets.
  • Extensive analysis of new techniques in advanced communication systems for machine-type communications (including beyond 5G and LoRa)

Details can be found in the deliverables and related publications.

Key exploitable results

The main results that could serve as the basis for future exploitation are:

  • Framework to select the most suitable IoT platform for different end-applications, which could be automated and further developed.
  • Explainable data mining algorithm for rare event detection that could be the basis of new software tools and expert systems.
  • Event-driven acquisition method that could be programmed in commercial products to pre-process data to improve the detection of outliers and decrease the communication burden.
  • Modifications of LoRa communication (not yet available in commercial products) to improve its reliability for massive connectivity.
  • New methods for data imputation that are not widely available in commercial products.

More results related to exploitation will be proposed when the first experiments in the testbeds start. The initial exploitation plan can be found here.