水面机器人机舱设备预测维护系统研究与设计

摘要

水面机器人是继无人驾驶汽车、无人机后的工业界、学术界的又一关注热点,机舱作为水面机器人的控制和运行中枢,要求具备高度的智能化和自治化.基于边缘计算模式设计和实现了水面机器人机舱预测性维护系统,该系统具有机舱运行参数实时感知、故障诊断,估计剩余寿命,进而开展预测性维护的功能.论文给出基于边缘计算的预测性维护系统体系结构设计,以NVIDIA Tegra Parker深度学习Soc为核心的边缘网关架构设计,基于GRU门控循环单元神经网络学习算法的故障诊断、剩余寿命预测方法及预测性维护实现思路,最后在NASA的PCoE发动机退化模拟数据集上验证了系统设计的可行性及预测算法的有效性.%Unmanned surface vehicles are another focus of attention in the industrial and academic fields after unmanned vehicles and unmanned aircraft. The engine cabin, as the control and operation center of USVs, requires a high degree of intelligence and autonomy. Based on the edge computing model, the predictive maintenance system of unmanned engine cabin is designed and realized.The realized system has the function of real-time sensing,fault diagnosis,estimated residual life of the cabin operating parameters,and then the predictive maintenance function.It presents the architecture design of the predictive maintenance system based on the edge calculation, the edge gateway architecture design with NVIDIA Tegra Parker as the core, realization of fault diagnosis, residual life prediction and predictive maintenance based on Gated Recurrent Units neural network learning algorithm. Finally, the feasibility of the system design and the effectiveness of the prediction algorithm are validated on the PCoE NASA engine degradation simulation data set.