基于多参数和特征层融合的化学品燃爆状态辨识方法

A Method of Hazardous Chemical Materials Deflagration Identification Based on Multi-Parameter and Feature-Level Fusion

作为化学品燃爆特性检测关键环节的燃爆状态辨识过程,目前仅以人眼或单一传感器辨识的手段进行,存在人工辨识受主观影响大、单一传感器适用范围窄的缺点,难以取得精确的辨识结果。提出引入多传感器、利用模糊神经网络系统(FNNS)对检测数据作特征层融合,以快速准确辨识燃爆状态的实验方法。结合模糊逻辑推理及RBF神经网络优势,在多层拓扑结构基础上建立了火焰光、热强度及其梯度与目标状态间特定的泛函映射。实验结果表明:该方法有效互补特征信息,提高检测装置的鲁棒性和状态辨识结果的置信度,错误率优于0.1%,使得燃爆特性的描述更加准确。

The deflagration identification on which eyes as well as single sensor were used is key work of deflagrating-characteristic detecting of chemical materials.But as artificial identification works subjectively and single sensor has certain limitation,the deflagration-identification result is not perfect.A new,quick and accurate deflagration-identification method is proposed,which is based on multi-sensor and feature-level fusion with fuzzy neural network system.As advantages of fuzzy logic consequence and RBF neural network are colligated,the functional map between the target state and flame information which contains the luminous intensity,the temperature and their gradients on the basis of multi-level topological structure is then built up.Experiments show that this method increases the robust of device and the confidence degree,of which the error rate is within 0.1%.Then the measurement of deflagrating characteristic is more precise.