Physics-informed turbulence intensity infusion: A new hybrid approach for marine current turbine rotor blade fault detection
Physics-informed turbulence intensity infusion: A new hybrid approach for marine current turbine rotor blade fault detection
复制标题
DOI：
10.1016/j.oceaneng.2022.111299
复制
发表时间：
2022-06
影响因子：
5
通讯作者：
Brittny Freeman;Yufei Tang;Yu Huang;James H. VanZwieten
中科院分区：
文献类型：
--
作者：
Brittny Freeman;Yufei Tang;Yu Huang;James H. VanZwieten
洋流涡轮机（OCT）将地球洋流中的动能转化为电能。然而，用于收集这种能量的OCT技术仍处于发展的早期阶段，这是由于技术和经济方面的挑战，这些挑战来自于与有限的地理位置和恶劣的操作环境相关的高操作和维护成本。为了缓解船舶发电可靠性问题，本文提出了一种新的物理引导转子叶片不平衡故障检测框架，该框架将从汽轮机电力信号中获取的非侵入性故障特征与环境条件数据相结合，提高了故障检测能力。这两个数据源的结合为物理信息神经网络的发展铺平了道路，确保我们的框架所做的分类与oct的流体运动学转子动力学在科学上是一致的。我们的框架的有效性在内部高保真数值模拟平台产生的模拟数据上得到了验证，该平台包括时间和空间动态海洋操作环境模型。测试结果表明，1类错误率为5.00%，2类错误率为2.92%。
Ocean current turbines (OCT) convert the kinetic energy housed within the earth’s ocean currents into electricity. However, OCT technologies used to harvest this energy are still at an early stage in development due to technical and economic challenges stemming from the high operation and maintenance costs associated with limited geographical location access and harsh operating environments. In an effort to alleviate reliability concerns associated with marine electricity generation, this paper proposes a novel physics-guided rotor blade imbalance fault detection framework that combines non-intrusively acquired fault features obtained from the turbine’s electrical power signal with environmental condition data to enhance the fault detection capabilities. The combination of these two data sources paved the way for the development of a physics-informed neural network that ensures the classifications made by our framework are scientifically consistent with the underlining hydro-kinematic rotor dynamics of the OCT. The effectiveness of our framework is validated on simulation data produced by an in-house high-fidelity numerical simulation platform that includes temporally and spatially dynamic oceanic operating environment models. Test results demonstrate a Type-I error rate of 5.00% and a Type-II error rate of 2.92%.