In production environments, monitoring the vibration of a machine or parts thereof can yield important information about the condition of the machine. The most common recommendation for vibration-based condition monitoring is to place a vibration sensor on each part of interest. These vibration sensors usually output preprocessed data, for example the root mean square value of a predefined time window.

    We propose to use machine learning to simultaneously monitor several vibration-emitting machine parts using only one single vibration sensor. Due to the superposition of multiple vibrations, this is not feasible using preprocessed sensor data. Our approach therefore consists of a one-dimensional convolutional neural network, which uses the raw vibration signal as input to classify the status of the monitored machine parts.

    As a first proof of concept, we monitored the status of three different motors. Using our approach, we were able to detect, which motors were running at a given time with high accuracy. We were able to significantly improve the classification results by transforming the raw data into the frequency domain. Our results are promising and show, that monitoring several vibration-emitting machine parts at the same time using only one vibration sensor is feasible.


    Access

    Download


    Export, share and cite



    Title :

    Enabling Single-Sensor Simultaneous Condition Monitoring of Several Vibration-Emitting Machine Parts Using Neural Networks


    Additional title:

    ARENA2036



    Conference:

    Stuttgart Conference on Automotive Production ; 2022 ; Stuttgart, Germany November 16, 2022 - November 18, 2022



    Publication date :

    2023-06-05


    Size :

    11 pages





    Type of media :

    Article/Chapter (Book)


    Type of material :

    Electronic Resource


    Language :

    English





    Condition Monitoring of Rotary Machine Parts

    Jain, Prakhar / Doshi, Siddhali | Springer Verlag | 2024


    Multicoordinate vibration sensor for cargo condition monitoring

    Vavilov, V. / Ismagilov, F. / Yakupov, A. | IEEE | 2016


    Using neural networks for monitoring UAV condition

    Gavrilov, A. I. / Zhiltsov, A. I. / Parfentiev, K. V. | American Institute of Physics | 2019


    Condition monitoring of railway single-throw equipment employing neural networks

    Fararooy, S. / Allan, J. / Abed, S. K. et al. | British Library Conference Proceedings | 1996


    110001 Using neural networks for monitoring UAV condition

    Gavrilov, A. I. / Zhiltsov, A. I. / Parfentiev, K. V. | TIBKAT | 2019