D5.2 Big data mining and analytics tools II

A public software deliverable, led by Fraunhofer, featuring the Big data mining and analytics tools; this updated D5.1
 

The Task 5.1 Multi-Level and Cross-Domain Big Data Analysis and Management can be split into two parts.

On one side, the collection processing and management of data, which is the Big Data Analysis task. On the other side, the visualization of this data in order for humans to be able to interpret it easier and faster, which is the Visual Analytics task. The Big Data Analytics task was mainly implemented by the LinkSmart® IoT Learning Agent (LA).  

Most of the new feature development had happed on the period between M5 to M15 and described in D5.1. This deliverable describes the period from M15 to M29, in which most of the work had been on the implementation of the pilots using the feature of the LA developed previously.

The main task of the LA is in the BSL-2 use case, in which the COMPOSITIONIIMS should collect sufficient data to predict a future failure and notify the adequate employee. To do so, the BMS connects to BSL subsystems and serializes and transmits the information to the COMPOSITION cloud. Then the LA collects and processes the data for executing the learning process. The learning process is the management of the data and machine learning model in order to be able to create live reliable predictions and continuous learning.

This process is orchestrated following the CEML methodology described in D5.1. In this process, the data is processed, cached and delivered to the DLT for creating predictions and train the model. With the predictions sent by the DLT,the LA propagates it to the Visual Analytics and other COMPOSITION subsystems such as DSS and SFT.  

Finally, the LA had been tested beyond the normal data loads faced in COMPOSITION pilots. This had been done to test the Big Data capacity of the LA and its scalability capacity. The LA has shown that is able to process effectively a big amount of data and scale in case the data exceeds the capacity of a single LA capacity.

A Visual Analytics tool has been implemented in order to enhance the decision support offered by COMPOSITION IIMS. The tool provides to the end-user the capability to visualize, analyze and explore industrial data derived from multiple sources. Furthermore, coordinated views of the data are supported in order to enable multifaceted perception and discovery of hidden subtleties in it. The interactive Visual Analytics tool imports data from Data Analytics tools of the project and based on this data it applies visualization techniques and presents the output to the users as graphical representations.  The tool completes the UIs of the project alongside with DSS and Marketplace interfaces and supports advanced visualizations that were not able to support by the aforementioned tools.