针对长距离分布式拉曼测温系统(R-OTDR)中数据量大影响系统实时性的问题, 提出通过并行运算实现高次累加平均算法提高系统信噪比和系统实时性能的方法。采用中央处理器(CPU)与图形处理器(GPU)协同合作的方式来提高系统的数据处理速度。传感数据由CPU控制数据采集端读取, 然后使用累加平均算法对传感数据进行去噪。在统一计算设备架构(CUDA)中, 通过调用GPU上的kernel函数对累加平均算法次数进行多线程分配, 以10线程的模式进行并行运算以提高数据处理速度。实验结果表明, 在10km传感距离下, 相比于原系统30000次累加平均算法, 采用50000次累加平均算法使系统测量误差由±1.1℃降至±0.5℃, 并且采用CPU和GPU协同合作的方式使50000次累加平均算法的运算时间由3890ms降至1.472ms, 提升了系统实时性能。

Aiming at the problem that the real-time performance of the system is reduced due to the large amount of data in the long-distance distributed Raman temperature measurement system, the system data processing speed is improved by taking advantage of the Image Processor (GPU) in parallel computing and combines the Central Processing Unit(CPU) and GPU. Among them, under the Unified Computing Device Architecture (CUDA), a high-order cumulative averaging algorithm is used to improve the system signal-to-noise ratio, and parallel computing is used to improve the real-time performance of the system. The CPU side reads the data on the data acquisition end, and then transmits it to the GPU to perform 50,000 cumulative average denoising on the scattered signal, and improves the data processing speed by rationally allocating threads. The experimental results show that compared with the CPU, the averaging processing of data on the GPU can increase the data processing speed from 3890ms to 1.472ms. The system detects a distance of 10km and a temperature error of ±0.5℃, which enables real-time accurate measurement. In summary, the data processing scheme used can be applied to improve system real time and reduce system measurement error of the distributed Raman temperature measurement system.