Open Access
Add to BibSonomy
Abstract
The refractive index (RI) measurment of the liquid often changes with the sensor's temperature. Accurate RI measurements require simultaneous measurements of the liquid temperature to calibrate the RI measurement. This study proposed what we believe to be a novel dual D-shaped plastic optical fiber sensor capable of simultaneously measuring both RI and temperature. A fluorescent material made of rhodamine B (RhB) is embedded in one of the D-shaped structures of the dual D-shaped fiber, which can be excited by a green laser to produce orange fluorescence. The fluorescence and the input laser of the fiber are superimposed at the output end of the fiber to form a specklegram containing information of two wavelengths. It was found through experiments that the effects of temperature and RI change on the red and green channels of the specklegrams are different, and the neural network can learn this feature to complete the measurement of both RI and temperature parameters. For RI sensing, the maximum error between the average predicted value and the true value of the test set is 0.0005. For temperature sensing, the maximum error between the average predicted value and the true value of the test set is 0.26°C. In addition, because the intensity change of the fluorescence varies linearly with RI and temperature, the sensor also has good stability.
© 2024 Optica Publishing Group under the terms of the Optica Open Access Publishing Agreement
1. Introduction
Optical fiber plays a crucial role in sensing applications, being widely used in temperature, pressure, magnetic field, and refractive index (RI) sensing, among others. RI measurement, in particular, has significant applications in medical diagnostics, food safety, environmental monitoring, and biometrics [1–3]. Traditional RI measurement methods involve the use of fiber Bragg gratings [4], fiber interference [5], and surface plasmon resonance [6] to observe changes in the output spectrum of the fiber for sensing. As changes in liquid temperature can cause changes in its refractive index, measuring changes in liquid temperature while measuring RI can result in more accurate sensing. Furthermore, it is often necessary to measure multiple parameters of a sample simultaneously in biometrics and industrial applications, making the design of a multifunctional, low-cost fiber multiplexed sensor quite necessary.
In previous research on fiber multiplexing technology, it was usually necessary to use fiber Bragg gratings and interferometers to achieve multiplexed sensing. Liao et al [7]. embedded a Bragg grating in a Mach-Zehnder interferometer to achieve simultaneous temperature and RI sensing. J. Shi et al [8] used a long-period grating (LPG) connected to a polarization-maintaining fiber in a Sagnac loop to simultaneously measure the RI and temperature. These sensing systems often require precision machining and are not easily integrated, only being testable in ideal environments. Compared to quartz fiber, plastic optical fiber (POF) has the advantages of a larger diameter, good flexibility, and ease of coupling and processing, and has been widely used in the field of fiber sensing in recent years [9–15]. Some researchers have taken advantage of the ease of customization of POF, using a simple mechanical die press print method [16] to etch long-period gratings on POF to improve the sensitivity of plastic fiber RI sensing. A sensor is proposed based on the surface plasmon resonance effect to measure temperature and RI at the same time. The sensor is made of two parallel D-shaped structures polished on POF, a PDMS coating on one polished area, while a 50 nm gold layer is deposited on the other surface using a plasma sputtering device. The two materials will produce different resonance peaks, and by monitoring the offset of the two peaks, temperature and RI can be measured simultaneously [17]. Based on this, Teng et al. proposed a new structure, by etching a V-shaped groove on the other side of the POF to enhance the SPR effect and narrow the wavelength width of the resonance peak [18]. Combining previous research, it is not difficult to find that to simultaneously complete POF and temperature measurement, it is often necessary to design a fiber structure that can complete multi-parameter sensing, and combine changes in the transmission spectrum to complete RI and temperature measurement. These methods often require the use of precision machining technology and a high-resolution spectrometer to complete, and the system cost is relatively high.
In recent years, with the advancement of image processing technology, fiber sensing based on specklegram recognition has sparked a research boom in the fields of bending [19,20], pressure [21], and magnetic field sensing [22], among others. As the evanescent wave of the fiber is affected by the external environment [23], the output light intensity of the D-shaped fiber will change with changes in the external RI, and the mode distribution of the output light will also change accordingly. Based on this principle, researchers have designed different fiber structures to make the transmission waveguide affected by the external RI [24–26]. By calculating the correlation between output specklegrams under different RI conditions, a connection between the RI and output specklegrams has been established. In addition, researchers have made a highly sensitive RI sensor by comparing the intensity of the output specklegram of D-shaped fibers under different RI [27]. Due to the fiber specklegram sensor (FSS) lack of multiplexing capability [28]. The research on RI sensors based on specklegram only discussed the impact of temperature on RI measurement, and cannot simultaneously complete RI and temperature sensing.
In previous research on FSS, researchers only focused on establishing a mapping relationship between changes in fiber output mode field and sensing parameters and ignored the changes in wavelength and intensity contained in the specklegram. Inspired by fiber fluorescence sensing, we etched two parallel D-shaped areas on a POF and applied a RhB-doped material to one D-shaped area. When the fiber transmission light passes through this area, the leaked light will excite RhB to produce fluorescence with a larger wavelength. RhB is a fluorescent material that is sensitive to temperature, and its excited fluorescence will decrease in intensity with temperature changes [29]. The other D-shaped area will cause changes in the intensity of the transmission light and fluorescent due to the changes in RI of the external environment. The neural network is used to analyze the output specklgram of the fiber, the temperature and RI can be measured simultaneously. In addition, due to the wavelength information and light intensity information added to the specklegram, the stability and repeatability of the FSS can be greatly improved.
2. Principle and method
2.1 Sensing structure manufacturing
The POF employed in this experiment is commercial step-index fiber, of which the core material of the fiber is PMMA with the RI of 1.49 and a diameter of 1.48 mm, the cladding material is a fluorinated polymer with a lower RI of 1.41 and a thickness of 0.01 mm. The fabrication process of the fluorescence probe is shown in Fig. 1. Step 1: Dissolve 0.02 grams of RhB into 20 milliliters of an ethanol solution. Stir thoroughly using a stirring rod to ensure complete dissolution. Subsequently, immerse Dow Corning 737 silica gel into the solution for 20 minutes. Step 2: Fabricate a double D-shaped fiber structure. The fabrication method of the D-shaped fiber is in accordance with the wheel polishing method [30]. Upon securing the fiber, initially utilize a coarse grinding wheel to grind the fiber, followed by a smoother grinding wheel for polishing. In this experiment, the depths of the two D-shaped structures are not identical. The deeper groove is filled with fluorescent material to ensure a greater amount of fluorescence can couple into the fiber from the D-shaped structure. The shallower groove does not require the addition of other materials. Step 3: Apply the soaked silica gel onto the deeper D-shaped structure and utilize a mold to remove the silica gel outside the fiber structure. Allow the fiber to air dry for one hour to complete the probe fabrication process.
2.2 Operational principle
During the transmission process in multimode fibers, interference occurs due to the different phase velocities of different modes. At the output end of the fiber, the speckles with uneven brightness can be collected through a camera. The intensity distribution of these speckles can be represented as:
Fig. 2(a) illustrates the sensing principle of the proposed sensor. The fluorescent material made of RhB, when excited by a green laser, undergoes energy-level transitions. The excitation wavelength shifts to red, producing orange fluorescence, which is coupled to the fiber through the D-shaped structure of the fiber. The excitation fluorescence intensity of RhB is sensitive to temperature changes. When the temperature rises, the excited fluorescence undergoes fluorescence quenching, resulting in a decrease in the fluorescence intensity collected at the fiber receiving end. The fluorescence output by fiber satisfies the condition of total reflection during transmission in the fiber, of which the reflection angle can be calculated by $\sin \theta = {{{n_1}} / {{n_2}}}$, where n1 is the RI of fiber and n2 is the RI of the external environment, As the value of n2 increases, θ correspondingly decreases, leading to a greater leakage of fluorescence. Consequently, the output intensity of the fluorescence also diminishes. In addition, as the RI of the external environment increases, the output intensity of green laser will decrease [23]. After the transmitted light passes through the double D-shaped structure, the specklegram output of the fiber can be represented by the following equation:
Fig. 2. (a) Schematic of the probe structure. (b) The sensing principle of the sensor. (c) The specklegrams output with different ${d_1}$.
3. Experiment and discussion
3.1 Experiment and data acquisition
The experimental setup is illustrated in Fig. 3. The laser outputs green light with a wavelength of 520 nm. A plastic optical fiber probe is secured within a quartz petri dish using two clamps. A heating stage is positioned beneath the petri dish, and a temperature probe is placed within the dish to monitor the temperature of the liquid. The output light from the fiber first passes through a 520 nm wavelength filter with a blocking rate of 97%, and is then collected by a CCD and transmitted to a personal computer. In the experiment, the temperature of the liquid varies between 30°C and 55°C, with a sampling interval of 1°C. The RI of the liquid varies between 1.334 and 1.400 by controlling the ratio of water to sucrose. The RI of prepared solution is measurement by refractometer (LingHeng LH-D50 H, China). During the RI variation process, specklegram collection is performed at refractive indices of 1.334, 1.34, 1.35, 1.36, 1.37, 1.38, 1.39, and 1.40, with each RI group corresponding to 26 temperature variations. A total of 1000 images are collected for each group, resulting in a total of 208 k specklegrams collected.
Since there are many kinds of specklegram collected in the experiment, the specklegrams collected at a temperature interval of 5°C are shown in this paper, as shown in Fig. 4(a). Since the laser input to the fiber is green light, and the excited fluorescence of the fluorescent material is orange light, the intensity of the red channel of the speckle is only affected by the fluorescence intensity, while the intensity of the green channel is equal to the sum of the input light and the green channel of the fluorescence. By comparing the specklegrams, it is found that the fluorescence intensity decreases with the rise of temperature and decreases with the increase of RI. The physical phenomenon is consistent with the analysis above. When the RI of the liquid is 1.334, the normalized intensity of the red and green channels of the specklegram varies with temperature, as shown in Fig. 4(b). The straight lines in the figure are the linear fitting lines of the intensity changes of the red and green channels with temperature, and their function expressions are y = 1.05-0.002x, y = 1.01-0.0004x, and the fitting coefficients R2 are 0.989 and 0.941, respectively. When the temperature of the measured liquid is stable at 45°C, the normalized intensity of the red and green channels varies with the temperature of the liquid, as shown in Fig. 4(c). The fitting line expressions of the red and green channel intensities with RI changes are y = 4.41-2.55x, y = 1.443-0.334x, respectively. The slopes of the four fitting lines are the sensitivity of the sensing parameters to the channel intensity changes. By comparing Fig. 4(b) and Fig. 4(c), it can be found that within the RI range given by the experiment, the change of RI has a greater impact on the output fluorescence intensity than the change of temperature. The effects of temperature and RI on the intensity of the red and green channels of the specklgrams are different. If mathematical fitting methods are used for sensing, there will be problems of large errors and complex calculations. Neural networks can learn the subtle differences between images by extracting image features, and the different sensitivities of the red and green channel intensities to the two parameters provide a basis for neural network generalization.
Fig. 4. (a) The specklegram collected in different temperature and RI. (b) The changed of normalized intensity of specklegram red channel with RI. (c) Under the same RI, the change of normalized intensity of specklegram red channel with the temperature curve.
3.2 Deep learning algorithms
Residual networks have been widely used in previous FSS sensing [31]. The addition of residual modules during the training process of the network model can effectively avoid problems such as model degradation and overfitting. The neural network model used in this experiment is based on Resnet18, as shown in Fig. 5. The specklegrams are scaled to an image of size 224 × 224 and input into the network model in three RGB channels. Considering the large size of the input image, a convolution kernel of size 7 × 7 is first used to traverse the image, and then after the maximum pooling layer, the image goes through four feature extraction processes. A complete feature extraction process consists of two convolution modules with a convolution kernel size of 3 × 3 and two residual modules. By continuously increasing the number of convolution kernels, the high-dimensional features of the image are extracted and then transformed into feature values by the global pooling layer into the fully connected layer. The output layer has two parameters, namely the value of RI and temperature. The dashed part of the residual module indicates that the feature dimensions extracted after convolution are different, and a convolution calculation with a convolution kernel size of 1 × 1 is required to ensure that the dimensions of the two modules added are the same. Before the data is input to the network, the specklegram data is divided into three parts, of which 60% is used as the training set, 20% is used as the validation set, and 20% is used for the test set. The neural network is implemented using the Keras framework with the RTX 3070 GPU. In this experiment, a Stochastic Gradient Descent (SGD) optimizer with a learning rate of 0.001 was utilized to update the network parameters. Throughout the training process, the network iterated 100 times, consuming a total of 23.5 hours. The time required to predict all values in the test set is 117 seconds.
3.3 Refractive index and temperature regression prediction
The prediction results of the residual network for RI and temperature are shown in Fig. 6(a) and Fig. 6(b), respectively. The red points in the figure represent the average prediction of each group of specklegrams, the length of the error bar represents the standard deviation of the prediction result, and the red line is the fitting line of the prediction value. The fitting coefficients R2 of RI and temperature are both 0.99. For RI sensing, the accuracy is 99.8% within an error range of 0.008, the maximum error between the average predicted value and the true value of the test set is 0.0005. Within an error range of 2°C, the accuracy of temperature prediction is 96.51%, and the maximum error between the average predicted value and the true value of the test set is 0.26°C. This shows that Resnet18 can complete the regression task of two parameters by analyzing specklegrams. The prediction error of temperature is greater than that of RI. This is because for the measurement range of this experiment, the change of RI has a greater impact on the output fluorescence intensity than the temperature, and the change of RI has a greater impact on the specklegram, so the accuracy of RI is higher.
For FSS, most sensors lack generalization ability. That is because the fiber core diameter with a larger normalized frequency is large, the transmission modes are too many, and when the sensing parameters change, the change of specklegram is difficult to quantify, and the limited training set cannot learn the change law of specklegram in the whole sensing measurement range [32]. However, for the sensing structure designed in this paper, the change of fluorescence intensity and external RI is a linear negative correlation. Compared with the specklegram change caused by mode crosstalk, the change of intensity information of different wavelength speckles is easier to be observed. In the well-trained neural network model, the change of color intensity should have a greater impact on the output result than the change of mode, and the change of fluorescence intensity can improve the generalization ability of the FSS sensor. To verify this conjecture, we collected 200 output specklegrams of the fiber at refractive indices of 1.345, 1.355, 1.365, 1.375, 1.385 and 1.395, respectively, under the condition of keeping the liquid temperature at 30°C, and used these six groups of specklegrams as the test set for testing. Figure 7(a) shows the change law of the intensity of the red channel of the specklegram when the RI increases from 1.34 with a step size of 0.005. The red points in the figure represent the specklegrams of the training set, and the turquoise points represent the specklegrams in the test set. The inset images in the figure are the original specklegrams and the intensity distribution maps of their red and green channels. The average intensity values of the red channel of the specklegrams in the test set are all distributed in the middle of the average intensity values of the red channel of the specklegrams in the training set. When the test set is input into the well-trained neural network model, the average value of temperature prediction is 30.19°C, and the standard deviation is 1.126. The prediction result of RI is shown in Fig. 7(b). The red points in the figure represent the average prediction value, and the length of the error bar represents the standard deviation of the prediction result. The predicted values of temperature and RI for each point are shown above and below the black line. The black line is the function graph of y = x, which is used to compare with the prediction result. Within an error range of 0.008, the prediction accuracy of the test set is 98.5%, the maximum error between the average predicted value and the true value of the test set is 0.0018. This shows that the neural network has learned the relationship between fluorescence intensity and sensing parameters, and the network has good generalization ability.
Fig. 7. (a) The average intensity of the red channel changed with the RI. (b)The result of the predicted RI with untrained specklegrams.
3.4 Repeatability and stability of the sensing system
The change of specklegram is very susceptible to external environmental influences, such as slight vibration, micro deformation of optical fiber, etc. Therefore, it is very meaningful to evaluate the repeatability and stability of the sensor for its practical application. We conducted 4 repeated experiments on 8 RI values within 32 minutes, collected 32 groups of specklegrams for testing, collected 200 specklegrams in each group, and a total of 6400 specklegrams are collected and input into the well-trained neural network for prediction. The test results are shown in Fig. 8(a), where the red points in the figure represent the average RI prediction of each group of specklegrams. Figure 8(b) is the change diagram of the average intensity of the red channel of the specklegrams collected in the 4 repeated experiments, and the table in the figure is the standard deviation of the intensity fluctuation of the red channel of the specklegrams corresponding to each RI. The fluorescence intensity is relatively stable in the RI repeatability test, and the network can well recognize the change of specklegrams information and make accurate predictions.
Fig. 8. (a) The predicted RI value changed with time. (b) The average intensity of the red channel changed with the RI.
The stability of the specklegram can be determined by calculating the change in the specklegram using zero normalized cross-correlation (ZNCC) [32]. To prevent changes in the RI caused by liquid evaporation, the probe is placed in the air for specklegram stability testing. During the test, the room temperature is stable at 24°C, the camera works continuously for 12 hours with a sampling interval of 1 minute, and a total of 720 specklegrams are collected. Taking the first specklegram as a reference, we calculated the ZNCC values of all specklegrams relative to the reference specklegram, as shown in Fig. 9. Over time, the ZNCC showed a slight decrease, but it could always stay above 0.995, indicating that the probe has good stability.
4. Conclusion
In this paper, a double-D shaped optical fiber sensor that can simultaneously measure the RI and temperature of liquids is proposed. This sensor has the advantages of low cost, small size, and easy fabrication. RhB, as a fluorescent material, is embedded in the D-shaped structure of the fiber. The fluorescence intensity output by the fiber decreases with the increase of the external RI and temperature. However, the changes in temperature and RI have different effects on the intensity of the red and green channels of the speckle. Based on this feature, a neural network can perform dual-parameter regression prediction on a specklegram. For RI sensing, the prediction accuracy of the network reached 99.8% within an error of 0.008, the maximum error between the average predicted value and the true value of the test set is 0.0005. For temperature sensing, the recognition accuracy of the network reached 96.51% within an error range of 2°C, the maximum error between the average predicted value and the true value of the test set is 0.26°C. In addition, the trained neural network model has good generalization ability for the specklegrams generated under this system. It is worth noting that the temperature measurement range of this sensor probe is not limited to the temperature range measured in the experiment. Between 25°C and 90°C, the excitation fluorescence intensity of RhB will decrease as the temperature rises [29]. However, the POF will deform when the temperature is higher than 75°C, so the working temperature range of this sensor is between 25°C and 75°C. In practical applications, the measurement interval can be designed within this range according to the needs and the corresponding specklegram can be collected for training. Different fluorescent materials have different physical properties. If the fluorescent material in the D-shaped structure is replaced, it is believed that this probe structure can also have good application prospects in other biological measurement fields such as PH or glucose.
Funding
Key Technologies Research and Development Program (2021YFB2800902); National Natural Science Foundation of China (61875008).
Disclosures
The authors declare that there are no conflicts of interest related to this paper.
Data availability
Data underlying the results presented in this paper are not publicly available at this time but may be obtained from the authors upon reasonable request.
Reference
1. T. T. Vu Nu, N. H. T. Tran, E. Nam, et al., “Blood-based immunoassay of tau proteins for early diagnosis of Alzheimer’s disease using surface plasmon resonance fiber sensors,” RSC Adv. 8(14), 7855–7862 (2018). [CrossRef]
2. M. Batumalay, Z. Harith, H. A. Rafaie, et al., “Tapered plastic optical fiber coated with ZnO nanostructures for the measurement of uric acid concentrations and changes in relative humidity,” Sens. Actuators, A 210, 190–196 (2014). [CrossRef]
3. P. Wang, F. Gu, L. Zhang, et al., “Polymer microfiber rings for high-sensitivity optical humidity sensing,” Appl. Opt. 50(31), G7 (2011). [CrossRef]
4. P. Xue, J. Zheng, Y. Xu, et al., “Study of transmission characteristics for a long-period grating on plastic optical fibers and its refractive index sensing,” Sens. Rev. 42(3), 326–331 (2022). [CrossRef]
5. F. Yu, P. Xue, and J. Zheng, “Enhancement of Refractive Index Sensitivity by Bending a Core-Offset In-Line Fiber Mach–Zehnder Interferometer,” IEEE Sensors J. 19(9), 3328–3334 (2019). [CrossRef]
6. S. Zhang, Y. Guo, T. Cheng, et al., “Surface plasmon resonance sensor based on a D-shaped photonic crystal fiber for high and low refractive index detection,” Optik 212, 164697 (2020). [CrossRef]
7. C. R. Liao, Y. Wang, D. N. Wang, et al., “Fiber In-Line Mach–Zehnder Interferometer Embedded in FBG for Simultaneous Refractive Index and Temperature Measurement,” IEEE Photon. Technol. Lett. 22(22), 1686–1688 (2010). [CrossRef]
8. J. Shi, G. Su, D. Xu, et al., “A Dual-Parameter Sensor Using a Long-Period Grating Concatenated With Polarization Maintaining Fiber in Sagnac Loop,” IEEE Sensors J. 16(11), 4326–4330 (2016). [CrossRef]
9. S. Lu, C. Wang, and Z. Tan, “Multiple side-coupled images recognition in plastic optical fibers based on deep learning,” Opt. Commun. 545, 129709 (2023). [CrossRef]
10. H. Qu, Z. Chen, S. Gao, et al., “Femtosecond laser line-by-line tilted Bragg grating inscription in single-mode step-index TOPAS/ZEONEX polymer optical fiber,” Opt. Lett. 48(6), 1438 (2023). [CrossRef]
11. A. Leal-Junior, A. Theodosiou, C. Díaz, et al., “Fiber Bragg Gratings in CYTOP Fibers Embedded in a 3D-Printed Flexible Support for Assessment of Human–Robot Interaction Forces,” Materials 11(11), 2305 (2018). [CrossRef]
12. A. G. Leal-Junior, A. Frizera, C. Marques, et al., “Polymer optical fiber strain gauge for human-robot interaction forces assessment on an active knee orthosis,” Opt. Fiber Technol. 41, 205–211 (2018). [CrossRef]
13. A. Pospori, C. A. F. Marques, D. Sáez-Rodríguez, et al., “Thermal and chemical treatment of polymer optical fiber Bragg grating sensors for enhanced mechanical sensitivity,” Opt. Fiber Technol. 36, 68–74 (2017). [CrossRef]
14. G. Lopes, N. Cennamo, L. Zeni, et al., “Innovative optical pH sensors for the aquaculture sector: Comprehensive characterization of a cost-effective solution,” Opt. Laser Technol. 171, 110355 (2024). [CrossRef]
15. A. G. Leal-Junior, A. Frizera, L. M. Avellar, et al., “Polymer Optical Fiber for In-Shoe Monitoring of Ground Reaction Forces During the Gait,” IEEE Sensors J. 18(6), 2362–2368 (2018). [CrossRef]
16. P. Xue, F. Yu, Y. Cao, et al., “Refractive Index Sensing Based on a Long Period Grating Imprinted on a Multimode Plastic Optical Fiber,” IEEE Sensors J. 19(17), 7434–7439 (2019). [CrossRef]
17. L. Liu, J. Zheng, S. Deng, et al., “Parallel Polished Plastic Optical Fiber-Based SPR Sensor for Simultaneous Measurement of RI and Temperature,” IEEE Trans. Instrum. Meas. 70, 1–8 (2021). [CrossRef]
18. T. Chuanxin, P. Shao, R. Min, et al., “Simultaneous measurement of refractive index and temperature based on a side-polish and V-groove plastic optical fiber SPR sensor,” Opt. Lett. 48(2), 235 (2023). [CrossRef]
19. S. Lu, Z. Tan, G. Li, et al., “A Sensitized Plastic Fiber Sensor for Multi-Point Bending Measurement Based on Deep Learning,” IEEE Photonics J. 13(5), 1–7 (2021). [CrossRef]
20. H. Gao and H. Hu, “Spatially-resolved bending recognition based on a learning-empowered fiber specklegram sensor,” Opt. Express 31(5), 7671 (2023). [CrossRef]
21. Y. Liu, X. Zou, G. Li, et al., “Reflective Optical Tactile Sensor Based on Fiber Specklegram Analysis With the Capability of Contact Position Identification,” J. Lightwave Technol. 41(8), 2540–2546 (2023). [CrossRef]
22. R. Zhu, S. Wan, Y. Xiong, et al., “Magnetic Field Sensing Based on Multimode Fiber Specklegrams,” J. Lightwave Technol. 39(11), 3614–3619 (2021). [CrossRef]
23. F. De-Jun, Z. Mao-Sen, G. Liu, et al., “D-Shaped Plastic Optical Fiber Sensor for Testing Refractive Index,” IEEE Sensors J. 14(5), 1673–1676 (2014). [CrossRef]
24. G. Mu, Y. Liu, Q. Qin, et al., “Refractive Index Sensing Based on the Analysis of D-Shaped Multimode Fiber Specklegrams,” IEEE Photon. Technol. Lett. 32(8), 485–488 (2020). [CrossRef]
25. M. A. Zain, H. Karimi-Alavijeh, P. Moallem, et al., “A High-Sensitive Fiber Specklegram Refractive Index Sensor With Microfiber Adjustable Sensing Area,” IEEE Sensors J. 23(14), 15570–15577 (2023). [CrossRef]
26. E. Fujiwara, L. E. Da Silva, T. D. Cabral, et al., “Optical Fiber Specklegram Chemical Sensor Based on a Concatenated Multimode Fiber Structure,” J. Lightwave Technol. 37(19), 5041–5047 (2019). [CrossRef]
27. A. Theodosiou, “Adaptive Refractive Index Measurements via Polymer Optical Fiber Speckle Pattern Analysis,” IEEE Sensors J. 24(1), 287–291 (2024). [CrossRef]
28. S. Lu, Z. Tan, W. Ji, et al., “A spatial domain multiplexing technology for fiber specklegram sensor,” Opt. Fiber Technol. 81, 103505 (2023). [CrossRef]
29. F. González-Martínez, O. González-Cortez, R. Pimentel-Domínguez, et al., “Composite polymer membranes for laser-induced fluorescence thermometry,” Opt. Mater. Express 8(10), 3072 (2018). [CrossRef]
30. J. Zhao, S. Cao, C. Liao, et al., “Surface plasmon resonance refractive sensor based on silver-coated side-polished fiber,” Sens. Actuators, B 230, 206–211 (2016). [CrossRef]
31. X. Wang, Y. Yang, S. Li, et al., “A reflective multimode fiber vector bending sensor based on specklegram,” Opt. Laser Technol. 170, 110235 (2024). [CrossRef]
32. G. Li, Y. Liu, Q. Qin, et al., “Fiber specklegram torsion sensor based on residual network,” Opt. Fiber Technol. 80, 103446 (2023). [CrossRef]
