Aspects of modelling performance in competitive cycling

• Main Author: Cangley, Patrick
• Published: University of Brighton 2012
• Subjects: 796.62
• Online Access: https://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.589939

The aim of this thesis was to design, construct and validate a model to be used for enhancing the performance of competitive cyclists in road time trials. Modelling can be an effective tool for identifying methods to enhance performance in sports with a high mechanical component such as cycling. The thesis questioned whether an effective road cycling model could be built. Existing models were analysed and found to have insufficient predictive accuracy to make them effective under general time trial conditions. It was hypothesised that an effective and generalised model could be developed. A computer simulation model was constructed that extended the functionality of existing models. The three-dimensional model combined the bicycle, rider and environment in a single parameterised system which simulated road cycling at high frequency. Three model components were validated against published benchmark studies. Firstly, a pedalling model was compared to an experimental benchmark study. Modelled vertical pedal force normalised root mean squared error (NRMSE) was 9.5% and horizontal pedal force NRMSE was 8.8% when compared to the benchmark. Both these values were below the 10% error level which a literature analysis indicated as the limit for validity. Modelled crank torque NRMSE was 4.9% and the modelled crank torque profile matched the benchmark profile with an R2 value of 0.974. A literature analysis indicated R2>0.95 was required for validity. Secondly, bicycle self-stability was evaluated against a benchmark model by comparing the eigenvalues for weave and capsize mode. Weave mode error level of 9.3 % was less than the 10% error considered the upper limit for validity. Capsize mode error could not be evaluated as the modelled profile did not cross zero. Thirdly, modelled rear tyre cornering stiffness was qualitatively compared with the results of an experimental study. The experimental study reported mean cornering stiffness of 60N/deg at 3 degrees slip angle, 10 degrees camber and 330N vertical load. This compared well with a model simulation which generated mean cornering stiffness of 62N/deg at 3 degrees slip angle, 4 degrees camber and 338N load. Experimental validation comprised a field case study and a controlled field time trial using 14 experienced cyclists. In the former study, modelled completion time was 1 % less than actual time. In the latter study, model prediction over a 4 km time trial course was found to be within 1.4±1.5 % of the actual time (p=0.008). The validated model was used to test potential performance enhancement strategies. A strategy of power variation in response to gradient changes had been previously proposed, but never experimentally confirmed. The thesis model predicted a 4% time advantage for a variable power strategy compared to a constant power strategy. This was confirmed experimentally in field trials when 20 cyclists obtained a significant (p<0.00l) time advantage of2.9±1.9 %. The model also predicted a 1.2% time advantage if power was varied in head/tail wind conditions on an out-and-back time trial course. A 2% time advantage was obtained in field trials but was not statistically significant (p=0.06). A final investigation examined the sensitivity of model prediction to variances in assumptions and initial conditions. An important sensitivity was the aerodynamic coefficient which could cause time differences of up to 6%. Tyre forces were also found to be a critical factor in the accuracy of model prediction. The thesis investigation confirmed the hypothesis that an effective and generalised model could be built and used to predict performance in road time trials.