基于路面纹理特征的抗滑性能估计方法研究

发布时间：2018-05-16 00:28

本文选题：路面纹理 + 抗滑　；参考：《武汉工程大学》2014年硕士论文

【摘要】：路面的抗滑性能对道路交通安全非常重要。而路面的抗滑性能是由路面纹理决定的，并且路面沥青混合料的级配造成的宏观和微观构造直观表现为路面纹理。因此，本文基于路面纹理的抗滑性能估计方法展开研究。为了能更快更准的估计路面抗滑性能，本文将路面纹理与抗滑性能通过人工神经网络来建立联系。路面纹理利用自制的纹理检测台采集，该平台由激光传感器、STM32F417芯片和上位机软件构成。然后处理采集的数据，最后对该数据进行分析研究。最后利用摆式摩擦仪测得的摆值和检测平台采集的路面纹理高程值作为神经网络训练参数，来建立两者复杂的非线性关系，从而估计路面抗滑性能。然后本文将采集的路面纹理数据利用MATLAB进行三维重建，测得数据对于恢复路面纹理略显不足，因此利用插值法优化数据，然后再进行路面纹理恢复，最终恢复的路面纹理较为理想。最后，，本文利用BP神经网络建立抗滑估计模型，该模型可以提高效率，优于手工计算，且还具有通过对以往经验的学习和对新方案进行合理的归纳来完善自身性能的能力。模型建立后，利用人工神经网络预测值与实测实际值进行比较，从而来检验模型的可行性。经检验后发现改模型预测准确率不高且收敛较慢，因此还需寻找一个训练参数，同时利用遗传算法优化网络的初始权值，解决收敛速度慢的缺陷。经过模型修正后，神经网络预测误差控制在5%以内，且收敛速度提高了一倍。
[Abstract]:The anti skid performance of the pavement is very important for road traffic safety. The anti skid performance of the pavement is determined by the pavement texture, and the macro and micro structure caused by the gradation of the asphalt mixture of the pavement is directly represented as the pavement texture. Therefore, this paper is based on the method of estimating the anti skid performance of the pavement texture.
In order to estimate the anti skid performance of the pavement faster and more accurately, the road texture and anti skid performance are established through artificial neural network. The pavement texture is collected by the self-made texture detection platform. The platform is composed of laser sensors, STM32F417 chips and upper computer software. Then the data are processed and the data are divided. At last, using the pendulum value measured by the pendulum friction tester and the pavement texture Gao Chengzhi collected by the testing platform as the training parameters of the neural network, the complex nonlinear relation between the two is established to estimate the anti skid performance of the pavement. Then the data of the collected road surface texture data are reconstructed by MATLAB, and the data are measured for the recovery path. The surface texture is slightly insufficient, so the interpolation method is used to optimize the data, and then the pavement texture is restored. The texture of the road surface eventually recovered is ideal.
Finally, this paper uses the BP neural network to establish the anti slip estimation model, which can improve the efficiency, better than the manual calculation, and also has the ability to improve its performance by learning from the past experience and making a reasonable induction to the new scheme. After the model, the model is compared with the actual actual value of the artificial neural network. After checking the feasibility of the model, it is found that the prediction accuracy of the modified model is not high and the convergence is slow. Therefore, it is necessary to find a training parameter and optimize the initial weight of the network by using genetic algorithm to solve the slow convergence speed. After the model correction, the prediction error of the neural network is within 5%, and the convergence speed is proposed. It's twice as high.

【学位授予单位】：武汉工程大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：U416.217

【参考文献】