Hello Everyone,
I am happy to answer any questions regarding the evaluation methods of Aerotune, which I developed. First of all, the idea of Aerotune is explained very briefly:
The basis for the evaluation method are algorithms that we have obtained from simulations in which the cyclist is simulated by a mathematical model. Through the use of these simulation models, we can use properties such as Athletes data (including power, weight, CdA value as a function of the yaw angle, Crr value) and track data (including slope, direction) in variation as desired. The simulated results are compared with real measured values to continuously improve the models. Disturbances and boundary parameters (for example wind speeds in all dimensions divided into finite elements, temperature, air pressure and humidity) are also taken into account in the models. In addition, disturbances can be given to the individual measured quantities in order to investigate the influence of measurement errors and thus to be able to determine the accuracy of our calculation method. To validate these models, we have developed a standardized test procedure known as aeroTEST, which allows every athlete to optimize their aerodynamics by themselves and interact with other users in our social platform.
Clearly a theoretical model continues to be flawed since it does not reflect the reality. However, it can determine the parameters presently known and we are interested in with high accuracy. Over the last 10 years my teams has built up profound knowledge and I claim that an aeroTEST is possible with a high level of measurement accuracy despite varying wind speeds with professional algorithms and a standardized procedure.
For practical use, an indication of the measurement error is indispensable. Therefore, we use the Mathematical method of least squares as the solution in our model. With this we can also determine the measurement errors for the fitted parameters. For the comparison of the calculated measurement errors with the real statistically determined measurement errors, we conducted a small non-representative study in which 20 tests were performed by different athletes (positions, power meters, bicycle type etc.). For this purpose, we have compared the statistical measurement error (95% interval) with the calculated measurement error, so that the systematic model errors are corrected here. Thus, a measurement error can be stated for each test, so that the quality of the measurement process can also be evaluated.
For the solution of the model, some special features are to be mentioned, as for example we can exactly determine the height profile. Since the route is driven several times, more altitude data is collected and therefore the altitude profile is approached accordingly with a higher degree of certainty. For this we use the “multivariate adaptive regression spline” (MARS) algorithm. The advantage is that the height profile “automated” is divided into different areas and the sections are approximated with regressions. For example, the height profile can consist of several functions of different degrees, which together offer the best fitting.
Currently, only the data from the day of measurement is used, in perspective, all data collected and obtained on the track is used.
To the questions of Mr. Chung, the following is to explain:
The procedure of the aeroTEST is different from most procedures. Using the app (functions as well without it); several tests of a setup (constant seating position and equipment) are driven. Only if two tests (each round trip) are within their calculated measuring tolerance a new setup can be started. This is how we try to set a quality standard (which is the minimum). The user can decide to increase this by conducting more tests of the setup. To determine the Crr value, a third test (out- and inbound) is required, which is run at approximately half the power used in the first two tests. The model is determined to take a constant Crr value over the three tests and fits it. This allows more data to be used in determining the parameters, gaining in results that are more accurate. I want to mention that we are not yet satisfied with the results of the Crr value. If there is too much wind turbulence on the test day or bad measuring equipment (e.g. lack of Speed Sensor, Quality of Power meter), the Crr value converges only at the limits.
Here are more investigations necessary.
Accordingly, points 1) and 2) from Mr. Chung would have to be adjusted to be a set-up, respectively, so that the aeroTEST can be checked according to Mr.
Chung’s idea.
To conclude, I believe that the athlete’s expectations regarding the accuracy of the CdA value are very high. In general, these expectations are not sustainable. It seems to us that the aerodynamics measurement has been completely theorized. Nobody wants to try a system without knowing that it measures correctly. However, it is unclear to most athletes which measurement accuracy is practically achievable and which factors have a strong influence on it. In addition, I think for most counts the idea that no training is “sacrificed” for aerodynamic optimization.
Various studies show that, for example, an average measurement accuracy of approx. + -1% can be achieved in the wind tunnel. With a CdA value of 0.25 m² or 25 aeroPOINTS as in our example, this results in a range of 0.247 to 0.253 m² (24.7 to 25.3), meaning that at 45 kph this would be a measuring tolerance of approx. ± 4 W. If, for example, results are published in which the CdA value is given on 4 digits without measuring errors, then in our view this can simply be wrong and leads to false expectations for the athletes. In addition to that, the information of results from the professionals, who can only post small improvements this, can easily result in a wrong perception, of both the quality of the measurement method used and the expectation of the personal CdA Value. Therefore, a picture builds up that only small improvements are possible and must be measured with high precision. We have gained quite a different experience here. In fact, we are also skeptical about the indication of the CdA value in m², as in general all values are smaller than 1 and therefore suggest a small value anyway.
Moreover, in this context, it is necessary to think about the measuring equipment. The power meter, for example, is one of the most important sensors for a good measurement. We found that for many athletes it is not clear how high the measurement error of their power meter in reality is. De facto, this usually hardly interests an athlete, but when it comes to the CdA value, better accuracies are expected than the available power meters they use can provide.
In this context, I can only agree with Mr. Chung that there is still a significant need for clarification here.
From our point of view, aerodynamic testing should be part of the training in order to build a better understanding and to improve continuously. One single day cannot create the perfect aerodynamics; there are just too many factors to take into consideration. Testing aerodynamics, however, can make a lot of fun and has successful results that it will become an addiction. So many things can be optimized. For example, here are 5 hand positions, 5 helmets and 5 one-piece suits, which would theoretically have to be tested in each variation, so 125 setups. That is why we should start testing today to understand aerodynamics and add a new dimension to our sport.
Sebastian Schluricke
CEO of Aerotune
