Oil property sensing array based on a general regression neural network
Oil property sensing array based on a general regression neural network
复制标题
DOI：
10.1016/j.triboint.2021.107221
复制
发表时间：
2021-08-09
影响因子：
6.2
通讯作者：
Zhe, Jiang
中科院分区：
文献类型：
--
作者：
Jiao, Dian;Urban, Aaron;Zhe, Jiang
对多种润滑剂性能的在线监测对于保持和延长高速旋转和往复式机械的健康至关重要，这些机械用于国家的许多关键行业，包括航空航天、制造和能源。已有许多工作致力于开发以测量润滑油的特定化学/物理性质为重点的传感器。这些特性传感器的一个长期挑战是输出重叠问题(交叉敏感性)，这意味着它们无法提供准确的测量结果。本文介绍了一种基于广义回归神经网络(GRNN)的电容式油性传感器阵列，用于测量润滑油中的酸、碱和水含量。结果表明，GRNN可以从重叠的传感器阵列的响应中准确地识别出单个油品的性质，具有较高的精度和速度。
Online monitoring of multiple lubricant properties is critical in maintaining and extending the health of highspeed rotating and reciprocating machinery used in many of the nation's key industries including aerospace, manufacturing, and energy. There have been many efforts on the development of sensors focused on measuring specific chemical/physical properties of lubricant oil. One long-standing challenge for these property sensors is the overlapping output problem (cross-sensitivity), meaning they cannot provide accurate measurements. Here we demonstrated a capacitive oil property sensor array based on a new general regression neural network (GRNN) for measuring acid, base, and water content in lubricant oil. Results showed that the GRNN can pinpoint individual oil properties from the overlapped sensor array's responses with high accuracy and speed.