Combined model based and data-driven diagnosis in a cloud

15 May 2017

A short video and demo of a conceptual, proof of concept, implementation combining model based and data-driven techniques in a cloud setting. The on-board system code is generated using the toolbox.

The use-case is diagnosis of an automotive engine and experimental data from an automotive engine is used in the evaluation. For more details, see the use-case.

The implementation of the system consists of three main parts:

  1. An on-board diagnosis system; A model-based diagnosis system written in C++. The core diagnosis code is automatically generated using the fault diagnosis toolbox at faultdiagnosistoolbox.github.io.

  2. A cloud system based on Django, Machine-learning in Python/scikit-learn, a Postgres database, and an NGINX web server

  3. A web interface written in Javascript using Polymer web components

An overview of the cloud system module is showed by the figure below.

Below is a video with a description of the system in operation, illustrated using two use-cases

Project responsible Erik Frisk (erik.frisk@liu.se)

Implementation by students @ Linköping University

  • Linus Ahlénius (Electrical Engineering)
  • Fredrik Björklund (Mechanical Engineering)
  • Sven Engström (Electrical Engineering)
  • Daniel Fahlén (Electrical Engineering)
  • Elina Fantenberg (Electrical Engineering)
  • Nils Larsén (Mechanical Engineering)
  • Oskar Lindahl (Electrical Engineering)
  • Lage Ragnarsson (Electrical Engineering)

Updated:

Comments