EMBUTCAR

Comunicaciones en congresos

• Congreso: 8º Foro de Excelencia 4.0 de los clústeres de automoción de España” (22/03/2022); presentación de VW Navarra y AIN de la solución desarrollada en proyecto EMBUTCAR (videoconferencia).

Otros

• Premio a la iniciativa “Foro de Excelencia 4.0”, en el Congreso Nacional de Clústeres.

The Embutcar project fits into the industry 4.0 framework. Within that, predictive maintenance is one of the primary tools for increasing machine availability and productivity.

In the automotive body part die stamping industry enormous gearbox actuated type presses.

To prevent a failure from causing a serious production incident, predictive maintenance by vibration can be used, with which future failures in components can be detected weeks or even months in advance and repairs can be scheduled.

That practice is common in other kinds of machines, but it is hard to use in the die stamping industry, because the impact of the press makes it impossible to study ways of failure.

The goal of the Embutcar project was, precisely, to create a vibration monitoring system for deep drawing systems that makes it possible to diagnose the kinetic chain of the gearbox system.  To achieve that, the impacts had to be detected or eliminated automatically and the operating mode of the press had to be identified.

The final goal was to create a system that automatically diagnoses some ways of failure and provides notification so they can be repaired.

In the first (2018) and second (2019) phases, a series of tasks were developed that were grouped into three work packages.

• WP1. Development of a data capturing system, and design the new hardware needed and install it in a vibration monitoring pilot plant in a production press at the plant It consisted of:
  • Developing new MEM sensors
  • Developing a new acquisition system (CMS) with an AIN EoloCMS 3 de AIN device as the starting point and adding some other functions to it:
    • MEMs/AC sensor reading capability
    • Communication with a PLC on the press to synchronise vibration signal processing
    • Basic logic programming to be able to store and analyse data from two press operating conditions
  • A pilot press at the Volkswagen press room in Navarre
• WP2. Data processing, and developing algorithms and software for post-processing the vibration data. The algorithm detects the press impacts and different ways the press works
• WP3. Signal analysis, basically consisting of:
  • Studying the construction design of this kind of press
  • Studying the forced and failure frequencies of the components of the machine
  • Designing viable processing solutions for this kind of machine

In the third and final phase of the project (2020), activities were focused on developing techniques for automatic failure diagnosis and a notification system for them. The work packages pursued were:

• WP3. Signal analysis, which was completed with the definitive configuration.
• WP4: Failure diagnosis, which consisted of:
  • Classifying possible types of failures: Identifying 178 different ways of failure
  • Data cleaning, entering external data and classification. All the facility vibration records had to be cleaned and classified to train the AI algorithms.
  • Test of AI techniques to be implemented: The result of the comparative tests was to choose a Convolutional Neural Network (CNN) algorithm as an AI technique.
  • Programming the chosen AI technique, an auto-diagnosis algorithm was developed for each failure mode. That merged the CNN technique with decision-making by a vibration analyst.
• WP5: A system for notifying maintenance needs, which consisted of:
  • testing and selecting a notification system. Finally, communication using digital signals between EoloCMS and the PLC was chosen, as well as email notification to AIN.
  • Programming the Notification Systems: A new temporary bidirectional communications system was developed that increases data refreshing for auto diagnosis to 10 minutes, which drastically reduces network traffic.

The results of the project were basically as expected:

• The pilot plant and all the hardware/software for it that was developed is 100% operational, and is acquiring data as expected. The design can be used as a basis for other automotive plants.
• The processing algorithms that were designed are 100% functional, however not everything done to minimise the impact of deep drawing was viable.
• The preliminary auto-diagnosis algorithms were implemented in the pilot plant and they work properly. Those algorithms were the starting point for future algorithms for the wind energy sector.
• The failure notification system is 100% operational and it notifies AIN about defects.