The assessment of power output (PO) has become increasingly popular within the last decade for both, amateur and professional road cyclists. Hence, a wide variety of power meters is available on the market. Most of the power meters are related to components of the drivetrain such as pedals, cranks or hubs. Watt et al. (2002) presented an approach to obtain PO through the bicycle frame strain and proved this concept in laboratory-based roller tests. By means of simple linear regression (SLR) as well as multiple linear regression (MLR) they were able to predict PO with a medium to high probability. However, two major limitations need to be considered. First, Watt et al. (2002) were looking only at three specific POs (150, 200, 250W) to feed their models and consequently a linear trend within that small PO range could be expected. Second, tests were run in a laboratory and hence neglected the noise that will be visible when running the straingauged bicycle in the field due to dynamic loads acting on the frame while steering, cornering, etc. The use of artificial neural networks (ANNs) seems to be a promising opportunity to overcome those limitations. Hence, this paper focuses on the following research question: Can strain gauges integrated into a bicycle frame be used to estimate the PO in road cycling?
© Copyright 2018 Sportinformatik XII. 12. Symposium der dvs-Sektion "Sportinformatik und Sporttechnologie" vom 5.-7. September 2018 in Garching. Abstracts.. Veröffentlicht von Feldhaus, Ed. Czwalina. Alle Rechte vorbehalten.