为解决现有隧道火灾报警器检测比较单一、测量受限而导致系统误报的问题，采用多传感器梯度数据融合模型对隧道火灾进行监测。在传感器与融合中心之间设计1个中间站，将温度传感器、火焰传感器、烟雾传感器采集到的数据信息上传至中间站进行预处理，采用相关性函数对传感器支持度较低的数据进行删除。并通过最小二乘法在中间站对来自同类传感器的多源数据进行局部融合，得到最优的融合数据。再利用D-S证据论算法将这些最优融合数据传送至融合中心进行全局融合，得到最终的融合值，通过对输出结果的判断来确定当前隧道内火灾的发生情况。最终的结果不依赖于某一个传感器或监测点数据。通过对现有的隧道火灾探测器运行的部分数据进行实验仿真，实验结果表明，相对于传统方法而言，基于多传感器数据融合的隧道火灾监测报警系统能够更加准确地识别出隧道内火灾发生情况。

 The existing tunnel fire alarms are relatively single in detection and limited during measurement, which leads to false alarms in the system. As a result, a multisensor gradient data fusion model is employed to monitor tunnel fires. An intermediate station is designed between the sensor and the fusion center, the data information collected by the temperature, flame, and smoke sensors are uploaded to the intermediate station for preprocessing, and the correlation function is used to delete the data with low sensor support. Then, the multisource data from the same sensor is locally fused by the least squares method at the intermediate station, and the optimal fusion data is obtained. Moreover, the DempsterShafer evidence theory algorithm is used to transmit these optimal fusion data to the fusion center for global fusion to obtain the final fusion value, and the current occurrence of fire in the tunnel is determined by judging the output results. The final results do not depend on a single sensor or monitoring point data. Through the experimental simulation of some data of the existing tunnel fire detectors, the experimental results show that, compared with the traditional method, the tunnel fire monitoring and alarm system based on multisensor data fusion can more accurately identify the fire occurrence in the tunnel.