This thesis begins with the description of background ,aims and goals of the project. Afterwards, a brief explanation of the probability theory and some probabilistic distribution functions that are used in this work are given.Then the concept of structural reliability is explained. Subsequently the meaning of Monte-Carlo simulation technique in relation to this workis described.
Traffic load models given indifferent codes cannot be used directly for the design of long-span bridges, i.e.bridges with span lengths larger than 200 meters. For the determination of the characteristic load values for such bridges, the Swedish National Road Administration (Vli.gverket) conducted traffic load measurements in five different locations in Sweden.
All measurement data of course contain errors. Therefore, the collected data from one of the measurement series are investigated in this work. The investigation shows that among the collected data there are vehicles that are registered with unreasonable lengths or/and weights. Totally about 10% of the whole collected data are considered to be inaccurate and should be excluded before further analyses of the vehicle data. About 10%,of the whole collected data, are vehicle dataregistered with one axle. Because of the limited information, these vehicles could not be identified and therefore are not excluded.
The characteristic and design traffic load values for different queue lengths are calculated using vehicle data both before and after filtration of inaccurate data. The results hows that the influence of inaccurate data is insignificant. The largest difference between characteristic load values calculated before and after filtration of inaccurate data is 3,5%. Comparing the characteristic as well as the design traffic load values with corresponding values according to BRO 94 shows that the code values are considerably higher than the actual values.
In this work, the Monte-Carlo simulation technique is also used to generate vehicle data. The simulated vehicle data show good agreement with the collected data. Afterwards, the probability distribution functions from the simulated as well as the collected vehicle data are calculated and compared. The result shows that the probability distribution functions agree well specially for grater queue weight values.
Stockholm: KTH , 2000. , 119 p.