论文部分内容阅读
为了更加准确地计算和预测航空管制员的工作负荷,利用雷达管制模拟试验获取的数据,分别采用线性回归、神经网络的非线性回归和基于神经网络的支持向量机方法,建立了基于扇区复杂性因素的管制员工作负荷实时计算模型。结果表明,这3种模型的绝对误差平均值分别为0.969、1.049、0.240;相对误差平均值分别为16.667%、17.979%、6.229%;均方根误差分别为0.186、0.206、0.114。另外,若采用5%作为基准精度,基于神经网络的支持向量机模型可以将相对误差控制在-0.5%~0.5%,表现出较强的误差控制能力。研究表明,可以采用扇区动态复杂性因素来计算管制员的工作负荷,相比线性回归、神经网络的非线性回归方法,基于神经网络的支持向量机方法对管制员工作负荷的计算有更高的精度。
In order to more accurately calculate and predict the workload of air traffic controllers, the data obtained by the radar control simulation experiment are used to establish the data structure based on the sector complexity, linear regression, nonlinear regression of neural network and support vector machine based on neural network respectively. Controller workload real-time computing model of sexual factors. The results show that the absolute error of the three models are 0.969, 1.049 and 0.240, respectively; the average relative errors are 16.667%, 17.979% and 6.229% respectively; and the root mean square error are 0.186, 0.206 and 0.141 respectively. In addition, if using 5% as the reference precision, the SVM model based on neural network can control the relative error between -0.5% and 0.5%, showing a strong error control ability. The research shows that the workload of controller can be calculated by using the sector dynamic complexity factor. Compared with the linear regression, the nonlinear regression method of neural network, the neural network-based SVM method can calculate the controller workload more The accuracy.