1. Introduction

Liquid sensing is indispensable in various industrial settings, including chemical processing, fuel storage, oil tanks and reservoirs, transportation systems, and power transformer, to maintain safety and optimize operations. It helps in averting disasters such as environmental leakage or spills and facilitates the efficient functioning of systems by preventing overflows and ensuring optimal storage levels [1]. Electrochemical liquid sensors are ubiquitously utilized across various applications. Nonetheless, they can present inherent safety challenges in specific environments, particularly when monitoring liquids that are conductive, corrosive, combustible, or even explosive in nature. Contrasted with traditional electrochemical sensors, optical fiber sensors emerge as optimal solutions in such adverse conditions due to their myriad benefits, including immune to electromagnetic field, resistance to chemical corrosion, small footprint, lightweight, high resolution, low power consumption, and capability of long distance monitoring [2].

In the realm of liquid detection, optical fiber sensors are typically employed in two distinct applications. The first pertains to the detection of pipeline oil leakages within the oil industry, predominantly facilitated by Raman-based distributed temperature sensing or Brillouin systems [3]. In this context, the optical fiber is aligned parallel to the pipeline, enabling distributed liquid leakage detection. This detection is indirectly achieved by gauging the temperature differential between the fluids and the surrounding environment. However, this method often necessitates a significant volume of liquid leakage and might exhibit hysteresis detection due to minimal temperature fluctuations. The second prevalent application involves liquid level sensing, wherein the sensor is directly immersed in the liquid, and detection is accomplished through cladding mode interaction. Numerous sensors have been developed based on fiber Bragg grating/tilted fiber Bragg grating [4–6], long period grating [7,8], Fabry-Perot interferometer [9], in-fiber interferometers [10,11], and multimode interference-based fiber structures, such as single mode-multimode-single mode fiber configurations [12–16]. While most of these sensors adeptly measure liquid levels, their spatial resolution and measurement range are contingent upon the density of the multiplexed sensors. Despite the potential of multimode interference-based fiber structure sensors for long-range, high-sensitivity liquid level measurements, they fall short in detecting liquid spills/drops across multiple locations by offering distance feedback. To surmount this limitation, there is a burgeoning interest in sensors capable of discerning multiple liquid locations with high spatial resolution. Contrary to point/discrete sensors, distributed optical fiber sensors afford spatially continuous measurement of requisite parameters along the designated optical fiber pathway [17–19]. Consequently, integrating the multimode interference sensing mechanism into the distributed sensing system emerges as a promising avenue for future developments.

In this paper, we introduce a single mode-graded index multimode-coreless fiber sensing architecture that directly interfaces with liquids. This fiber configuration is seamlessly integrated into an optical frequency domain reflectometry system, enhancing its distributed sensing capabilities. Within our proposed sensing paradigm, the coreless fiber region is exposed to the external medium, facilitating liquid detection through cladding mode interaction. The Rayleigh backscattered spatial domain signal intrinsic to the fiber structure is harnessed to pinpoint the precise locations of liquid droplets with high spatial resolution. We have successfully identified the positions of four sequential oil droplets and demonstrated the ability to estimate their dimensions. Our innovative sensing approach paves the way in both liquid droplet detection and liquid level measurement for industrial applications.

2. Structure design and measurement principle

A schematic diagram of the single mode-graded index multimode-coreless fiber structure is depicted in Fig. 1. Initially, this structure is spliced by integrating a segment of single mode fiber (Corning SMF-28e+) with a short section of multimode fiber. The primary role of this truncated multimode fiber is to expand the beam size emanating from the single mode fiber. In our sensor design, we employ a graded index multimode fiber (Thorlabs, GIF625, core: 62.5 µm, cladding: 125 µm, NA: 0.275) to achieve this beam expansion. Post-splicing, the graded index fiber is truncated to approximately a quarter-pitch length (240 µm). Given the refractive index profile inherent to the graded index fiber, it can function as a lens, translating a point source on the lens's entrance surface into infinity or collimating it. Subsequently, the graded index fiber's terminal side is spliced with a segment of coreless fiber (Thorlabs, FG125LA, diameter: 125 µm). As all three fiber sections possess a consistent diameter of 125 µm, sensor fabrication is straightforward through fusion splicing, obviating the need for meticulous considerations such as offset-fusing or in-fiber interferometry sensors during the fabrication phase.

 

Fig. 1. Schematic diagram of a single mode-graded index multimode-coreless fiber structure for Rayleigh backscattered liquid detection. SMF: single mode fiber, MMF: multimode fiber.

Download Full Size | PDF

Within our sensor configuration, the single mode fiber, due to its small core diameter (8 µm), is spliced to a pitch-length graded index fiber, thereby facilitating effective beam expansion over a larger area (Φ62.5 µm). A segment of coreless fiber is then spliced to the graded index fiber, serving a pivotal role in sensing by being directly exposed to the ambient medium. This coreless fiber is essentially a pure silica glass rod without an additional cladding layer. Light transmitting from the graded index fiber to the coreless fiber remains predominantly confined within the coreless fiber, given that air possesses a refractive index lower than that of the coreless fiber (1.44 at 1550 nm). When the sensor's coreless fiber segment is surrounded by a liquid with a refractive index marginally exceeding 1.44, the evanescent light at the coreless-liquid interface can penetrate into the liquid. Rayleigh scattering in the optical fiber arises from stochastic fluctuations in the fiber's index profile [20]. Given the pronounced refractive index fluctuation at the coreless-liquid boundary, an enhanced Rayleigh scattering reflectivity at the liquid's contact point is anticipated. Consequently, a Rayleigh backscattering-based optical frequency domain reflectometry can be synergized with the proposed structure, enabling straightforward tracking of the spatial domain signal along the coreless fiber for distributed liquid detection. It should be noted that distributed sensing is achieved by tracking the back-reflected light along the fiber structure. This sensing mechanism presents a notable difference compared to the conventional single mode-multimode-single mode fiber structure, which tracks the transmissive spectrum.

3. Experimental setup

The schematic of our experimental setup for distributed liquid detection is depicted in Fig. 2. We employ a tunable laser source (TLS, Keysight 81680A) as the interrogation system's light source. The TLS owns a tuning speed of 5 × 10³GHz/s (equivalent to 40 nm/s), a tuning range (Δν) of 2.5 × 10³GHz (or 20 nm), and an initial wavelength set at 1520 nm. A coupler (split ratio: 10/90) splits the light emitted from the laser into two distinct pathways. The first pathway directs light towards the auxiliary interferometer, specifically a Mach-Zehnder interferometer. This interferometer provides an external clock, which subsequently triggers the data acquisition card. This card samples the interference signal, ensuring equal optical frequency spacing, thereby mitigating the nonlinearity inherent in the TLS's frequency tuning. The second pathway is channeled to the main interferometer. Within this main interferometer, the fiber under test (FUT) integrates a standard single-mode fiber spanning 3.35 m with our proposed fiber structure. Within the FUT, the graded index fiber is 240 µm, aiding in beam expansion, whereas the coreless fiber, tailored for distributed liquid drop detection, has a length of 22 cm. To ensure the coreless fiber remains accessible to external liquids, its protective coating layer has been carefully stripped off.

 

Fig. 2. Experimental setup for distributed liquid drops measurement. TLS: tunable laser source, CPL: coupler, CIR: circulator, PC: polarization controller, FUT: fiber under test, BPD: balanced photodetector (Thorlabs, PDB450C), DAQ: data acquisition card.

Download Full Size | PDF

4. Results and discussion

In our proposed method, we leverage the spatial domain signal of Rayleigh backscattering for liquid sensing. Initially, we examined the Rayleigh backscattering signal of the fiber structure in the absence of any liquid. Figure 3 depicts the spatial domain signal of the proposed structure, obtained after applying a Fourier transform to the originally recorded data. As illustrated in Fig. 3, the structure's spatial characteristics are evident through its Rayleigh backscattering reflectivity. Given the known configuration of the structure, it is feasible to precisely identify reflection events on the Rayleigh backscattering trace. The two splicing positions, along with the coreless fiber endface, exhibit Fresnel reflections characterized by supreme reflectivity, thereby facilitating straightforward identification. Notably, the dual endfaces of the graded index multimode fiber are discernible as two proximate peaks, as highlighted in the magnified inset figure. The clear demonstration of two resolved endfaces at an interval of 240 µm underscores our system’s resolving capability. In our sensing system, we have a tuning range of 20 nm, which equates to a theoretical spatial resolution of 40 µm. This means our system is fully capable of resolving the multimode fiber section. We also deliberately retained a minuscule amount of coating material adjacent to the coreless fiber endface, which is conspicuously visible adjacent to the endface reflection peak. It merits attention that side lobe ghost peaks are discernible for each Fresnel reflection. These ghost peaks typically manifest symmetrically around the Fresnel reflection in optical frequency domain reflectometry [21]. It is possible to properly suppress the ghost peaks for a better spatial domain trace observation by using different processing approaches as proposed in Ref. [21]. While suppressing ghost peaks is not the main focus of this work, we have concentrated on the sensing realization for liquid detection. Armed with this comprehensive data from the initial Rayleigh backscattering spatial domain signal, we can seamlessly transition to liquid detection by monitoring the spatial domain Rayleigh backscattered reflectivity.

 

Fig. 3. Rayleigh backscattering measurements for the proposed fiber structure. The dual endfaces of the graded index multimode fiber, as well as the endface of the coreless fiber, are discerned through distinct Fresnel reflection peaks. The inset provides a zoomed-in view, highlighting the two endfaces of the graded index fiber with an interval of 240 µm. Residual coating material proximate to the coreless fiber endface manifests as a discrete reflective peak.

Download Full Size | PDF

To elucidate the capability of liquid detection, we selected off-the-shelf vegetable oil as our test sample with a refractive index of approximately 1.47. Utilizing a plastic liquid transfer pipette, we deposited four successive oil droplets onto the coreless fiber segment. It's important to note that our experiment was conducted at room temperature without significant temperature fluctuations. The fiber sensor was strategically positioned on a planar metal substrate, maintaining a straight fiber layout. Subsequent to each oil deposition, data was promptly captured. Figure 4 presents the spatial domain results of Rayleigh backscattering in the presence of the oil droplets. In Fig. 4(a), the spatial domain depiction with a singular oil droplet is illustrated. A pronounced enhancement in Rayleigh backscattering is evident along the coreless fiber segment. The introduction of the oil instigates an initial Rayleigh backscattering reflectivity that is approximately 20 dB superior to the intrinsic reflectivity of the coreless fiber. While this value remains inferior to the Fresnel reflection magnitude, its distinct reflectivity renders it sufficiently discernible for liquid detection. Drawing insights from the spatial domain Rayleigh spectrum, we can also approximate the dimension of the oil droplet. As showcased in the magnified inset of Fig. 4(a), the oil droplet's dimension is gauged to be around 1.6 cm. It is worth mentioning that this estimated dimension is subject to temporal fluctuations, given the propensity of the oil droplet to traverse the optical fiber's axis on our flat metal substrate. Consequently, the deduced droplet dimension represents an initial estimate upon its deposition on the fiber.

 

Fig. 4. Distributed detection of oil droplets: monitoring the spatial domain signal of Rayleigh backscattering along the coreless segment. a, Spatial domain Rayleigh backscattering trace with a singular oil droplet on the coreless fiber, accompanied by an inset illustrating the estimated droplet size of 1.6 cm. b, c, d, Spatial domain Rayleigh backscattering traces corresponding to the presence of two, three, and four oil droplets on the coreless fiber, respectively.

Download Full Size | PDF

To verify the sensor's capability for distributed liquid detection, we released oil droplets using a plastic transfer pipet. These droplets were deposited onto the coreless fiber at intervals relative to the preceding droplet. Figure 4(b) demonstrates a spatial domain Rayleigh backscattering trace with two oil droplets. Within the delineated red dashed box, we can discern two distinct enhanced Rayleigh backscattering spectra along the coreless fiber segment. Interestingly, the reflectivity induced by the second oil droplet closely mirrors that observed when only a single droplet is present on the fiber. Conversely, the initial droplet's reflection begins to diminish in reflectivity, and the spatial domain spectrum broadens as the oil deforms along the fiber. As we persisted in introducing oil to the coreless fiber, analogous effects were evident with three and four droplets, as depicted in Fig. 4(c) and Fig. 4(d), respectively. The intervals between oil droplets measured from the first droplets are 1.2 cm, 1.6 cm, and 2.2 cm. It becomes evident that an increased number of oil droplets on the coreless fiber results in diminished reflectivity from the spatial domain Rayleigh backscattering, especially for droplet locations proximal to the fiber's termination. This reflectivity pattern can be attributed to the predominant light power consumption by initial reflection events, while subsequent reflections contribute less to the formation of the reflective spectrum. Moreover, as the oil droplet interacts with the optical fiber, it gradually spreads, increasing its contact area. This expansion results in a broader reflection Rayleigh backscattering trace and concomitant reduction in reflectivity. We monitored the variations in oil droplet size and summarized in Fig. 5(a). To validate our observations regarding reduced reflectivity, we further tracked the peak reflectivity fluctuations of the coreless fiber endface with increasing oil droplet counts. As plotted in Fig. 5(b), a discernible decline in endface Fresnel reflection is observed as oil droplets are progressively added along the coreless fiber segment. This trend further underscores the modulation of Rayleigh backscattering energy distribution through alterations in reflection events.

 

Fig. 5. a, Estimation of oil droplet size with increasing droplet count. b, Variation in peak value at the coreless fiber endface corresponding to the incremental addition of oil droplets along the sensor.

Download Full Size | PDF

In the design of our proposed sensor, one might consider the use of a large core step-index multimode fiber as a substitute for the quarter-pitch length graded index fiber. While a large core multimode fiber can ostensibly serve a similar function to the graded index fiber in broadening the light beam emanating from the single mode fiber, the quarter-pitch length graded index fiber assumes a more nuanced role than only beam expansion. Owing to its intrinsic self-imaging property, this graded index fiber is pivotal in our Rayleigh backscattering-based sensing mechanism. This is because it facilitates the collective refocusing of the Rayleigh backscattered light back into the single mode fiber, thereby efficiently unveiling the spatial domain information along the coreless fiber. In our presented work, our emphasis is on evaluating the distributed liquid detection capability by introducing oil droplets rather than immersing the entire sensor in the liquid. Regarding liquid level measurement capabilities, we anticipate that our proposed sensing scheme will meet the requisite standards. In a liquid level context, the sensor's coreless segment can be submerged in the liquid, allowing for the tracking of the augmented Rayleigh backscattering reflection at the air-liquid interface.

It is crucial to extend the measurable distance on a larger scale. Especially in some industrial applications, there might be a need to install fibers spanning several kilometers or even tens of kilometers. Typically, in optical frequency domain reflectometry sensing system for parameters like strain or temperature, sensing is achieved by tracking the Rayleigh spectrum shift using optical domain information; it isn't directly observable from the spatial domain information. Thus, the sensing capability is primarily limited by the tuning range of the tunable laser source. However, in our demonstrated sensing mechanism, liquid detection is realized by directly tracking the Rayleigh backscattering reflection from the spatial domain data. This means we can still track reflections resulting from the liquid even if the fiber under test is extremely long, for instance, in the range of one hundred kilometers. Additionally, we must acknowledge that temperature variations would affect the sensor's performance. Generally, our sensing does not rely on tracking the Rayleigh spectrum. However, temperature variations would lead to changes in the refractive index of the liquid under test. As a result, the peak reflectivity might vary based on different temperature conditions. Nevertheless, the location of the peak appearance would remain consistent, even as reflectivity changes with temperature fluctuations.

The results obtained from our experiments show the potential of leveraging Rayleigh backscattering in the spatial domain for liquid sensing. The distinct reflectivity patterns observed with varying oil droplet counts on the coreless fiber highlight the sensor's sensitivity and capability for distributed liquid detection. In addition, it is worth mentioning that a liquid with a refractive index exceeding that of the coreless fiber is essential to meet the requirements of the sensing mechanism. Upon reviewing various oils used in daily life or specific industrial contexts, we observed that the majority have a refractive index greater than 1.46, with many surpassing 1.50. Thus, we acknowledge that our proposed sensor may not be suitable for all liquid types. Still, given the choice of coreless fiber, a significant number of liquid types can be detected, making our sensor viable for many applications. Given the pressing need and inherent challenges of multi-interface liquid level assessments in the oil and gas sectors—where water, oil, emulsion, and foam must be distinguished [22], our proposed sensing approach appears apt for such application scenarios. The observed modulation of Rayleigh backscattering energy distribution through alterations in reflection events provides insights into the intricate interplay between the liquid and the fiber sensor. Comparing our findings with existing literatures, our method offers a more nuanced understanding of the spatial domain information along the coreless fiber, especially with the use of the quarter-pitch length graded index fiber. This not only enhances the sensor's efficiency but also broadens its potential applications. Future studies could delve deeper into optimizing the sensor design for specific industrial applications, exploring its performance under varying environmental conditions, and testing its efficacy with different liquids of varying refractive indices.

5. Conclusion

Liquid sensing remains a cornerstone in a myriad of industrial applications, ensuring both safety and operational efficiency. While the ubiquity of electrochemical liquid sensors is acknowledged, their inherent safety challenges, especially in environments with hazardous liquids, cannot be overlooked. This research has underscored the potential of optical fiber sensors, particularly the single mode-graded index multimode-coreless fiber sensing structure, as a safer and more efficient alternative. Our proposed structure, which directly interfaces with liquids and is integrated into an optical frequency domain reflectometry system, has demonstrated high precision in detecting liquid droplets. The successful identification and size estimation of four consecutive oil droplets attest to the efficacy of this approach. Furthermore, the sensor's ability to provide spatially continuous measurements offers a significant advantage over traditional point sensors, addressing the current challenges faced in detecting liquid spills across multiple locations. In essence, this research not only showcases the advancements in liquid sensing technology but also paves the way for industrial liquid detection applications. As industries continue to prioritize safety and efficiency, we believe the adoption of such innovative sensing mechanisms will undoubtedly become paramount.