基于RBF神经网络的震伤人员快速评估模型

The Rapid Assessment of Wounded Personnel Based on RBF Neural Network Model under the Background Earthquake Disaster

震伤人员的快速评估对统筹配备医疗资源和进一步预测废墟受伤和失踪人数十分关键。为提高震后受伤人员评估结果的可靠性,通过对影响震伤的关键性影响因子的提取,采用能够有效处理模糊性和非线性指标的RBF神经网络模型对震后受伤人员进行快速评估。从分析震后造成人员受伤的影响因子入手,从承载体减抗风险能力、暴露性和敏感性三个维度提出震伤人员预测指标体系;在受伤人员预测方法上,考虑指标的小样本性、非线性和部分指标的模糊性特征,将模糊逻辑与神经网络方法结合起来,采用动态优化的径向基(RBF)神经网络方法,以提高评估模型的全局搜索和优化能力,避免常规BP神经网络较早陷入局部优化的不足;案例结果显示:与BP神经网络训练的绝对误差3.24%相比,RBF神经网络震伤人员评估模型的绝对误差能够降低至1.71%,精度提高47.2%,说明本文评估模型评估可靠性高、模型鲁棒性强,能够推广于震灾应急的管理决策之中。

This is a critical question to assess the wounded personnel under background earthquake disasters, and the question can effectively solve the allocation of medical resources, forecasting of debris injured and missing personnel. The paper takes the extraction of indexes of wounded personnel assessment and the RBF neural network model to improve the reliability of assessing result. Firstly, the paper puts forward some assessment of wounded personnel around the ability of resisting earthquake disasters, vulnerability and sensitivity. Secondly, the paper combines fuzzy logic and neural network method to solve the small sample, nonlinear and part fuzziness of indexes and the author put forwards the RBF neural network method that has the dynamic characteristics. The RBF neural network method can improve the global search ability and optimization ability and RBF neural network can avoid the shortcoming of BP neural network method. Thirdly, the case result showes that the RBF neural network is superior to the BP neural network. The RBF neural network can reduce the absolute error to 1.71% and increased the reliability to 47.2 %. So the RBF neural network model has the characteristic of high fault-tolerance and robustness. The RBF neural network model can be applied to earthquake emergency management decisions.