Soft Vibration Sensor: Load prediction using Deep Neural Networks (vor Ort)
* Presenting author
Abstract:
Convolutional Neural Networks (CNN) have significantly contributed to increase the accuracy of regression-based methods. In the field of image recognition different deep learning networks based on CNN layers are standardized and can be adapted for different use cases now. It has been shown that CNN can be used as well as virtual sensors to approximate vibration responses at different locations to reduce the number of sensors or predict vibration signals in general. Furthermore, the usage of deep neural networks simplify, accelerate and improve load prediction at different system levels for vibration signals. Developed is a framework for a vibration soft sensor to predict loads and hence reduce the reliability testing and simulation effort. Preliminary applications of vibration soft sensors on automotive components will be presented.
