Graduate Program in Electrical Engineering, Federal University of Espirito Santo (UFES), Vitória 29075-910, ES, Brazil.
Physics Department & I3N, University of Aveiro, 3810-193 Aveiro, Portugal.
Sensors (Basel). 2021 Jul 3;21(13):4568. doi: 10.3390/s21134568.
This paper proposed a liquid level measurement and classification system based on a fiber Bragg grating (FBG) temperature sensor array. For the oil classification, the fluids were dichotomized into oil and nonoil, i.e., water and emulsion. Due to the low variability of the classes, the random forest (RF) algorithm was chosen for the classification. Three different fluids, namely water, mineral oil, and silicone oil (Kryo 51), were identified by three FBGs located at 21.5 cm, 10.5 cm, and 3 cm from the bottom. The fluids were heated by a Peltier device placed at the bottom of the beaker and maintained at a temperature of 318.15 K during the entire experiment. The fluid identification by the RF algorithm achieved an accuracy of 100%. An average root mean squared error (RMSE) of 0.2603 cm, with a maximum RMSE lower than 0.4 cm, was obtained in the fluid level measurement also using the RF algorithm. Thus, the proposed method is a feasible tool for fluid identification and level estimation under temperature variation conditions and provides important benefits in practical applications due to its easy assembly and straightforward operation.
本文提出了一种基于光纤布拉格光栅（FBG）温度传感器阵列的液位测量和分类系统。对于油类的分类，将流体分为油类和非油类，即水和乳液。由于类别变化较小，因此选择随机森林（RF）算法进行分类。通过位于底部 21.5cm、10.5cm 和 3cm 处的三个 FBG 识别了三种不同的流体，分别是水、矿物油和硅油（Kryo 51）。通过放置在烧杯底部的珀耳帖器件对流体进行加热，并在整个实验过程中将温度保持在 318.15K。RF 算法对流体的识别准确率达到了 100%。使用 RF 算法进行液位测量，还可以获得平均均方根误差（RMSE）为 0.2603cm，最大 RMSE 低于 0.4cm。因此，该方法是在温度变化条件下进行流体识别和液位估计的一种可行工具，由于其易于组装和操作简单，在实际应用中具有重要意义。
