Abstract

In recent years, with the increasing demand for intelligent society, intelligent photonics has developed rapidly. Machine learning (ML), as a subset of artificial intelligence (AI), has played an important role in the intelligent evolution of optical fiber sensors. Its impact extends beyond enhancing sensor performance by introducing innovative problem-solving approaches. Specifically, ML algorithms have become instrumental in signal demodulation and elevating the efficacy of discrete and distributed sensors, and have also greatly promoted the development of optical fiber speckle pattern processing. This paper presents the latest advancements in ML-based optical fiber sensors, outlines the problems faced by conventional demodulation methods and the common ML algorithms applied in optical fiber sensors, and emphasizes key applications. Additionally, this paper delves into the challenges and future development of this emerging research direction.

Introduction

Due to advancements in powerful computing tools, hardware, widespread use of cloud data, and the evolution of the Internet of Things, the accessibility of large datasets has significantly improved. This has promoted the emergence of many efficient machine learning (ML) algorithms. These algorithms, aided by continuous advancements in data analysis technology, have demonstrated effectiveness in training models and solving specific problems. If ML algorithms are complex and deep enough in terms of layers, then ML can become so-called deep learning (DL) [1]. Regardless of the name, ML seems to have a wide range of applications in modern society, because the achievements of artificial intelligence (AI) often exceed those achieved by manual control design.

In 2021, Nature Photonics further advanced its exploration of the hot topic of optics/photonics through a focused issue called “Machine learning of Light”. As an important type of microphysics, optics seems to have similarities in complexity with some of the “black box” processes in ML. In the Q&A session, David Pille and Aydogan Ozcan discussed the two primary drivers behind the current application of ML in photonics [2]. One direction is to use AI methods on hardware to design optical structures and devices with specific performance. Another direction is to use optical systems as fully optical/hybrid statistical inference models to achieve AI calculations.

Optics and photonics applied to ML computing offer unique advantages, allowing for ultrafast calculations at extreme frame rates and low energy consumption [3], [4], [5], [6]. For instance, Feldmann et al. [7] proposed a fully optical neural synaptic system. They use wavelength division multiplexing technology to implement a scalable circuit architecture for photonic neural networks (NNs) and use optical pulse signals to regulate the phase change materials that encapsulate neurons. This phase change material can regulate transmittance and reflectivity through a controllable laser. This system can perform both supervised and unsupervised learning. The work successfully demonstrates direct pattern recognition in the full optical system. This photonic NN has the potential to leverage the inherent high-speed and high bandwidth of optical systems, thereby enabling the direct processing of full optical communication and visual data. However, logic gates, computation, and NNs are all linear (passive) systems that cannot address nonlinear (active) issues. The lack of effective nonlinear optical processes as activation functions for many nodes poses challenges for all-optical implementation [8].

Conclusion and future perspectives

This paper reviews the application of the ML algorithm in optical fiber sensors. In recent years, the advent of ML has greatly impacted optical development. Compared with static demodulation algorithms and complex spatial optical paths, the ML algorithm offers better signal processing mechanisms for optics. ML algorithms enable automated processing, analysis, and recognition of spectrums, thereby enhancing the efficiency and accuracy of spectrum analysis. Using ML algorithms to classify and identify the spectrum of different substances, enabling qualitative and quantitative analysis of substances. By learning from known corresponding spectrums and building models, ML algorithms can predict and analyze unknown spectrums. ML algorithms can extract and recognize features from the spectrum, analyze the content and proportion of various components, and reveal the intrinsic nature of substances. With the continuous development of spectroscopy and ML algorithms, the application prospects of ML in spectroscopy will become increasingly widespread. This advancement will contribute to more efficient and accurate analytical methods in chemical analysis, environmental monitoring, food safety, and other fields, promoting progress in scientific research and expanding application fields.

Distributed optical fiber sensing that enables simultaneous vibration and temperature monitoring is demonstrated in this work, which is achieved by implementing polarization optical time-domain reflectometry (P-OTDR) and Raman optical time-domain reflectometry (ROTDR) respectively through space-division multiplexed (SDM) configuration in different cores of a multicore fiber (MCF). Optimized system setup is obtained in the proposed hybrid system, where only one laser source is used, and the generated pulse is shared by the two reflectometers. Owing to the reason that separate interrogation fiber cores are used by the two reflectometers, the proposed hybrid system allows for simultaneous measurements of spontaneous Raman and Rayleigh scattering signals, thus it no longer suffers from the incompatible pump power levels issue, which results from different pump power demands, and hinders the implementation of the hybrid system in single mode fiber due to the fiber nonlinear effects imposed restriction. In addition, in order to extend the sensing range, we also explore the employment of wavelet transform denoising technique in the improvement of signal-to-noise ratio of ROTDR measurement. Eventually, simultaneous distributed vibration and temperature sensing in 5.76 km sensing range with 3 m spatial resolution and 0.57 ◦C temperature uncertainty has been demonstrated. The proposed hybrid sensing system has the outstanding advantage of high reliability, since the cross-sensitivity issue is effectively mitigated, which shows great potential for pipeline monitoring in oil and gas industry.

Introduction

The intelligent monitoring of environmental variation in the fields of civil, mining, and energy exploration has attracted ever increasing attention recently, where the environmental parameters (e.g. vibration, temperature and strain, etc.) are utilized to perceive the structure integrity and health condition, so that the safety and reliability of industrial operations can be enhanced notably. Specifically, simultaneous vibration and temperature sensing are highly desired for long range and distributed pipeline monitoring in oil and gas industry [1], [2], which can be used for intrusion detection and leakage detection, respectively. Distributed optical fiber sensing (DOFS) technique based on the measurements of backscattering signals turns out to be one of the most promising solutions for this application, thanks to its outstanding capability of long sensing range (tens of kilometers) and high spatial resolution (meter-scale) [3]. However, it is difficult to achieve simultaneous and discriminative measurement between vibration and temperature by using a single distributed fiber sensing technique, because most of the DOFS techniques are suffering from the cross-sensitivity issue among multiple parameters, which will decrease dramatically the reliability of measurements in real field deployment. One solution of this problem is to use two sensing techniques simultaneously, where one is used to monitor vibration, meanwhile the other one can be utilized to sense temperature.

This paper outlines the second part of an experimental study to show the capabilities of distributed optical fiber sensors (DOFS) in their application to the structural health monitoring (SHM) of the shear performance of concrete structures. SHM seeks to obtain the shear crack characteristics of concrete elements: detection, localization and quantification of shear damage. The two first were discussed in the first part of the experimental study. In the present paper, the quantification is dealt with by proposing a method to obtain the mean shear crack width in concrete beams. The method is based on the experimental data obtained by a DOFS bonded to the concrete surface. First, the basis of the methodology are presented and, later on, experimentally checked by testing of three partially pre-stressed concrete (PPC) beams subjected to a shear test with increasing level of load. The DOFS were deployed in the web of the beams to conform a 2D grid mesh to measure the strain profile along two orthogonal directions. The experimental data was obtained using an OBR (Optical Backscattered Reflectometer) system with high spatial resolution and sensitivity that allow a complete mapping of the cracking pattern and to obtain the required data for the calculation of the crack width. The results show the feasibility of the proposed method in calculating the shear crack width when compared to the results from traditional instrumentation.

Introduction

One of the most important parameters to quantify the level of damage in a concrete structure is the presence of cracks. Cracking behaviour and control of crack width are significant factors in the design of concrete structural members. Nowadays, most formulations to evaluate and control the crack width have been originally developed for tensile and flexural cracks. In the specific case of shear stresses, there is not a specific methodology to evaluate the shear crack width in concrete members [1]. This is because the mechanism of diagonal cracking is not perpendicular to the vertical reinforcement and it is more complex than axial or bending cracking [2]. The few methodologies to evaluate shear cracks width in concrete members are based on several experimental and analytical studies [1,3,4]. However to obtain experimental data in these formulations during and after the test, several manual and observational activities have to be performed such as tracing the crack pattern with a marking pen, or measuring the cracks width progression by using visual techniques. It is important to consider the implications, since several measuring errors could appear in all of these activities. In this paper, a method to obtain the mean shear crack width in concrete structures using experimental data obtained with distributed optical fiber sensor (DOFS) connected to an optical backscattered reflectometer (OBR) is proposed.

I- Introduction

II- Operation Principle and Sensor Design

III- Experimental Setup

IV- Results

V- Conclusions

References

Abstract

Liquid level sensing is nowadays a relevant issue in a broad range of applications, forcing the sensors performance and cost to be evaluated in parallel. This work proposes a fiber opticbased liquid level sensor system using a Fabry-Pérot interferometer (FPI) embedded into a polyurethane resin diaphragm. The FPI is based on microcavities generated upon catastrophic fuse effect, enabling the fiber recycling and sensors fabrication in a cost-effective way, compared to traditional methods. To enable the simultaneous temperature control, a Fiber Bragg Grating was used as thermal reference sensor to compensate the temperature cross-sensitivity. The sensor prototype was tested in a field application, using two different configurations, an open chamber configuration, where the diaphragm is in contact with the atmosphere, and a closed chamber configuration, revealing sensitivities of 4.4±0.1 pm/mm and 1.57 ± 0.04, respectively. These sensitivity values are within the figures of merit for diaphragm-based sensors recently reported.

INTRODUCTION

NOWADAYS, the monitoring of the liquid-air interface is required to infer level in a wide variety of applications such as, for instance, fuel supply systems, oil reservoirs, wastewater treatment plants [1]. Other relevant applications can be found in medical treatments, chemical processing and pharmaceutical development, demanding high resolution on liquid level detection [2]. Traditionally, liquid level sensing is mainly based on mechanic and electrical techniques, such as capacitance and radio frequency admittance [3], float type devices as magnetic floating gauge [4], ultrasonic and radar based sensing [3]. Despite that electrical-based sensors are widely employed, their applicability is limited when the sensing medium is conductive, potentially explosive and erosive [5], [6]. Optical fiber-based sensors have been considered an attractive prospect for liquid level monitoring, due to its unique advantages, such as electromagnetic immunity, electrical insulation, compatibility with remote sensing and sensor multiplexing. Moreover, these advantages render optical fiber appropriate for long term, reliable level measurements in special conditions [7]–[9]. Distinct operation principles are behind optical-based sensing, namely refractive index variation. These sensors measure the liquid level based on the immersed sensor length which lead to a specific effective surrounding refractive index and several implementation examples can be found: etched Fiber Bragg Grating (FBG) [6], etched chirped FBG [2] or long period grating (LPG) [10]. An additional technology is based on excessively tilted FBG that generates a fast and slow-axis due to its birefringence [11]. However, due to its length (102 m), these sensors are restricted to a reduced range liquid measurement.

Introduction

The structural safety of civil infrastructures, ocean platforms and aerospace structures has received increasing attention, because the failure of those important structures usually leads to large abundant of casualties and economical loss. To characterize the structural performance, structural health monitoring (SHM) technology has been recognized as one of the most effective and intelligent measures [18,19,1,23,22,10,7]. By the use of smart sensors and components, the real-time, long-term and continuous information of the in-situ structures can be provided for the damage identification, disaster forecasting and warming, and safety and life-time assessment [35,17,20,16,26]. Among these smart sensing elements, optical fiber based sensors are the most popular in civil engineering for the unique advantages of high sensitivity and precision, corrosion resistance, anti-electromagnetic interference, good stability, geometrical shape-versatility, absolute measurement and convenient integration of network [25,34,32,12]. For the brittle material properties of silica fiber, bare optical fiber is weak to resist the shear or torsion force in structural construction and operation. Especially for the embedded case, the packaging technique is the most critical factor to guarantee the survival and enhance the durability of optical fiber based sensors. However, the existence of the protective layer introduces the intermedium between the sensing fiber and the monitored structure, which makes the strain measured by the sensor not entirely represent the actual strain of host material [2]. The error between the measured strain and the actual strain is attributed to the strain loss in the transferring path. To eliminate the strain transfer error and improve the measurement accuracy of optical fiber based sensors, strain transfer theory has been developed to establish the quantitative relationship of strains between the host material and the optical fiber [13,33,9,28]. Considerable attempts have contributed to studying the strain transfer mechanism of optical fiber sensors. The earliest research started from the 1990s, and the strain relationship between the concrete and sensing fiber was studied with a polymer-to-glass modulus ratio of 1/200 given [21]. Host material with optical fiber sensor embedded was then simplified to infinite elastic cylinder model, and plane-strain theory was adopted to explore the strain transfer mechanism [24]. However, the simplified model in the two theories couldn’t be used to accurately determine the strain transfer relationships for various host materials with non-elastic behavior. In 1998, a systematic strain transfer theory of a three-layered structure embedded with optical fiber sensor was established [3]. To analyze the effects of a local interfacial slippage on the strain transfer ratio, a two-layered mechanical model consisted of host material and optical fiber was discussed [13]. For a multi-layered structure with various packaging layers, the unified strain transfer formula was conducted [33]. The improved strain transfer deduction of a three (multi)-layered sensing model by the use of simplified geometrical and physical functions was proposed [15]. Strain transfer analysis was also extended to special cases for considering the viscoelastic material properties of the monitored structure [30].

DRDO has been actively involved in the design and development of various types of UGVs for different applications. The first Indian tracked unmanned ground vehicle, named as MUNTRA (Mission Unmanned TRAcked) with the state-of-art technologies was designed and developed by the Combat Vehicles Research & Development Establishment (CVRDE), Avadi, Chennai. CVRDE has developed the teleoperated MUNTRA series of tracked UGVs on the BMP II class of armoured, amphibious vehicles for specific operational missions of unmanned surveillance, NBC reconnaissance and mine detection. For one of the surveillance requirements of possible threats, CVRDE had established a fiber optic based communication link for a standalone, unmanned surveillance system (USS). Various land and sea based threats can be detected, tracked and classified using the sensor suite of the USS. Visual confirmation of the threats can also be achieved by the synchronous slaving of the electro-optic sight to the radar. This paper describes the performance analysis of fiber optic network in terms of BER, throughput, latency, jitter using Ethernet analyzers. BER is defined as the rate at which errors occur in a transmission system. This can be directly translated into the number of errors that occur in a string of a stated number of bits. The definition of bit error rate can be defined as: BER = Number of errors / Total number of bits sent. If the medium between the transmitter and receiver is good and the signal to noise ratio is high, then the bit error rate will be very small, possibly insignificant, and having no noticeable effect on the overall system. However if noise is present, then there is a chance that the BER will need to be considered. When data is transmitted over a data link, there is a possibility of errors being introduced into the system and the integrity of the system may be compromised. As a result, it is necessary to assess the performance of the system. Unlike many other forms of assessment, bit error rate, BER assesses the full end to end performance of a system including the transmitter, receiver and the medium between the two. In this way, BER enables the actual performance of a system in operation to be tested. For fiber optic systems, bit errors mainly result from imperfections in the components of the link. These include the optical driver, receiver, connectors, splices, bends, cracks and the intrinsic properties of the fiber itself. Bit errors may also be introduced as a result of optical dispersion and attenuation that may be present. Noise may also be introduced in the optical receiver itself. Typically these may be photodiodes and amplifiers which need to respond to very small changes and as a result there may be high noise levels present. Another contributory factor for bit errors is any phase jitter that may be present in the system as this can alter the sampling of the data.

Romania has one of fastest internet connection in the world, being the first one in Europe and on the 5th place in the world with the fastest internet connection with a pick of 67 Mbps, according to Akamai Report published for December 2014. The status of fixed internet connections in Romania is observed by ANCOM as follows in Figure no. 1. According to Romanian telecom regulator body ANCOM at the end of 2016 more than half of 4,4 million fixed broadband connections enable speeds above 100 Mbps. The average fixed broadband internet download speed in Romania went up by 78% last year, compared to 2015, to 95 Mbps while the average download speed for mobile connections also increased by 35% to 26 Mpbs. This very good situation of internet speed in Romania is due to fiber optic networks developments. The actual leader with the biggest fiber optic network, RCS&RDS, decided since 2000 year to have its communications network based on fiber optic cables. Their fight with incumbent operator, Telekom Romania, for growing-up the FTTH networks in the last 5 years made us to question ourselves during this article: where is going the future of fiber optic networks in Romania beyond the physical numbers of FTTH connections or Past Households/Connected Households? We will go in this paper through legal framework, new operators, technologies for fiber optic networks and challenges of actual fiber optic networks for future use and we will end with some conclusions.

Two prominent applications of high-accuracy time and frequency transfer can be identified: the operation of a global navigation satellite system (GNSS) and of telecommunication networks. Each of them has ambitious requirements regarding accuracy, availability, and security. Mobile telecommunication networks are operated according to the Long Term Evolution-Advanced (LTE-A) standard and are going to be prepared for future fifth generation (5G) standards. Mobile time-division duplex (TDD) operation, new features for increased spectrum efficiency like enhanced inter-cell interference cancellation (eICIC), future new mobile location-based services, and single-frequency network-based multi- and broadcast applications (MBSFN) services need not only frequency syntonization, but also time synchronization. In order to reach the required network synchronization quality [1], dedicated synchronization chains are implemented, structured as a hierarchical and layered synchronization network. Each network equipment draws its synchronization signal from a superior hierarchy element located closer to the primary synchronization source(s). An example of such a network is shown in Fig. 1. It comprises the network production part, responsible for routine operation of the network, with the requirement of continuous 24/7 operation, and a primary clock supervision part. At the network core level, the highest accuracy synchronization equipment including a number of primary reference time clock (PRTC) functions [2] is used, which is responsible for passing down the network reference time along the hierarchy. In order to increase the network synchronization stability and minimize GNSS related risks, enhanced PRTCs (ePRTCs) [3] and coherent network PRTCs (cnPRTCs) [4] combine GNSS receivers with atomic (cesium) clocks. This approach has recently been proposed for standardization by the respective committees of the International Telecommunication Union — Telecommunication Sector (ITU-T). Over a few core locations, ePRTCs are going to be distributed geographically in the network. The ITU-T specified maximum absolute time error (max|TE|) allowed for ordinary PRTC function is 100 ns [2], whereas the related value for ePRTC is more tightly set at 30 ns [3]. In addition, stability specifications for dynamic time error, expressed as maximum time interval error (MTIE) and time deviation (TDEV), apply. For 24/7 synchronization of the network production part, technologies such as synchronous Ethernet (SyncE) and Precision Time Protocol (PTP) with full timing support from the network (PTP-FTS) according to the ITU-T Recommendation G.826x series for frequency and G.827x for time synchronization, are able to ensure the required level of accuracy at the end application.

The main methods available for measuring slurry flow are: (1) the differential pressure method; (2) the electromagnetic method; (3) the Doppler ultrasonic method; (4) the volume method; and (5) the Coriolis mass method. A differential pressure flow meter (1) can be used for measuring the flow rate of various slurry fluids. The Venturi tube is generally used to ensure that the throttle device does not deposit solid particles. It is important to take care to avoid abrasion and issues with paste material jamming. The electromagnetic method (2) can measure pulp, coal slurry, slurry and other high-conductivity liquids, but this sensor cannot be installed in a horizontal pipe. The Doppler ultrasonic method (3) may introduce large measurement errors because of the heavier flow on the bottom of the tube. The volume method (4) is the most traditional measurement approach, but its efficiency is considered to be too low. The Coriolis mass method (5) can be used to measure the flow of liquid containing a small amount of solid particles; however, when the quantity of solids increases, and the solid particles are large, then the issue of anti-blocking must be considered(Wang and Zhang, 2014; Wang et al., 2015; Binzoni and Martelli, 2015; Andrzej et al., 2015; Tajima et al., 2014). Paste filling is an important method for mine filling, and flow measurement is, therefore, the first problem to be solved. Optical fiber sensing technology is introduced into the flow measurement to provide improved accuracy and stability. Optical properties such as intensity, phase, and frequency of the light transmitted in the optical fiber are controlled, and appropriate optical detection methods can then convert it into electrical signals. The resulting flow measurement instrument is highly complex and both the flow meter measurement principles and the measurement method have their advantages and disadvantages. Optical fiber Doppler sensors have rapidly developed in recent years to measure flow with high precision, fast dynamic response, good space differentiating ability as well as many other advantages. Consequently, it is becoming one of the main methods of flow measurement. Based on the scattering principle, the laser Doppler and the optical mixing principle, an optical fiber Doppler sensing system is designed, including the optical detection circuit, voltage amplification circuit and low-frequency filtering circuit.

Environmental monitoring is essential in order to deal efficiently, safely, and effectively with resources, such as petroleum, natural gas, cultivated land, and others. To this end, various green technologies have been developed.1 However, understanding and prediction of environmental changes remain challenges. Thus, unexpected changes that cause or become an ecological or human-health threat need to be monitored in the temporal and spatial domains.2 Monitoring technologies have been focused on analysis of the chemical and physical properties of environments.3 Electrochemical sensors have been developed as monitoring sensors in various studies according to analytes4-6 or sensing principles.7-9 They have some advantages such as low-power operation, high selectivity, and the linear relationship between output signals and analyte concentration.10 Electrochemical sensors generally contain the electrodes coated with a catalyst that promotes either oxidation or reduction of certain chemical species.11 Even though they contain such high-selectivity catalysts, the study of catalyst species with high durability under extreme conditions of harsh environments remains challenging.12 Other types of electronic sensors, for example electromechanical sensors13 and thermoelectric sensors,14 have also been studied as monitoring sensors in harsh environments. A micro-electro-mechanical sensor has been extensively studied owing to its small size15 and low production cost.16 However, those sensors have a short life-time in harsh environments due to the failure resulting from creep and plastic deformation, corrosion, and electrical disturbance.17 Especially as environmental monitoring sensors are required to be running for long periods in harsh conditions, it is necessary to develop sensors that are durable in harsh environments for long operation times. The optical fiber sensor (OFS) is a promising sensor for harsh environments. Compared with the electronic sensors, the OFS uses optical cables that have low heat loss and high data bandwidth over electric cables.18 This allows connectivity over hundreds of thousands of kilometers via massive optical cables. Massive information technology resources are connected continent-to-continent19 thanks to submarine cables laid around the world as shown in Fig. 1.20 Optical fiber cables offer reliability even underneath the ocean for long durations, as optical fiber has unique characteristics such as immunity to electromagnetic interferences and chemical corrosion, and low heat loss.21 This indicates that the OFS can also generate and transmit measurement data with high sensitivity even in harsh ambient conditions, such as in downholes for oil recovery22 and in nuclear power plants.23 Moreover, the OFS is compatible with existing fiber-optic communication systems, which means environmental monitoring using the OFS can be remotely operated in a network operating system.24 The light weight, small size, and high flexibility of optical fibers play an important role in environmental monitoring sensors as well.

LOCOMOTION plays a crucial role on the capacity to perform activities of daily living, personal functional abilities, and independent development in the community. However, clinical conditions such as stroke, spinal cord injury, Parkinson disease and loss of a member by amputation can affect the human locomotion with varying degrees of severity [1], [2]. To overcome or attenuate human gait disorders, devices such as prostheses [3], exoskeletons [4], and smart walkers [5] have been employed for assistance and rehabilitation purposes. Lower limb exoskeletons are being used for gait assistance and rehabilitation therapies. Robotic exoskeletons are successfully employed for rehabilitation exercises due to its higher repeatability and the quantitative feedback of the patient recovery [6]. Furthermore, the embedded sensors can also be used for continuous monitoring and evaluation of patient progress with predefined objectives and allowing the customization of treatment with increasing levels of personalization and difficulties. This customization is achieved through the dynamic interaction between the exoskeleton and the patient, which happens with the transition between passive, active-assisted, and active-resisted movements of the robotic device controller [7]. Considering the human safety, it is desirable that the exoskeleton actuators have low output impedance [7]. Otherwise, there is a risk of accident with the patient [7]. Furthermore, the actuators must have a bandwidth close to the one of human movement in order to perform the desired movements of the physiotherapy section. A straightforward manner to achieve these requirements is by placing an elastic element between the load (human limb) and the actuator. This is the series elastic actuator (SEA) principle proposed in [8]. Although the spring placed between the load and the actuator can significantly reduce the system output impedance, it also reduces the system bandwidth [9]. Since the human movement occurs in low frequencies [10], SEAs have been successfully employed as exoskeletons actuators [11], [12], [7], among others. Another advantage of SEAs is the possibility of estimating the output torque from spring deflection. For this reason, sensors for spring deflection can be employed instead of the torque sensors, which may simplify the actuator instrumentation. However, sensors used to measure these deflections are generally encoders [7] and potentiometers [13] that require mechanical supports precisely assembled due its sensitivity to misalignments. This may result in a less compact system. Additionally, potentiometers may also result in noisy measurements.