A sensor network is an array of sensors that are deployed either directly inside the element to be assessed or very close to it. Optical fiber sensor networks represent a significant improvement over traditional sensors networks ensuring a wide range of application areas such as environmental, safety and security monitoring. As regards the links, fiber-optic sensors can be interconnected by wireless, copper wire or optical fiber [1]. Among them, the last one is the chosen technology.

The crucial feature of optical fiber is its dual functionality, it is not only a sensing structure in view of the measurand induced changes of the light properties that propagates in the fiber, but also a communication channel, meaning there is no need for an additional telemetry path, contrary to what happens in all other sensing technologies [2]. Other key advantages are: almost total immunity to external electromagnetic interference, multi-path reflections from natural and man-made objects and interruptions by bad weather; on top of that, it exhibits vast bandwidths and low transmission losses, enabling very wide geographical coverage; intrusive data interception is more difficult when using dielectric optical waveguide transmission media than free-space radio propagation or metallic wires; where necessary, fiber interconnects allow the sensors to be inserted within the structures being monitored, they can operate without electrical powering of local batteries outside the terminal nodes and they reduce the risk of sparking in combustible environments [3].

Consequently, fiber-optical sensor networks have emerged as a powerful tool for condition assessment of the system under consideration [2]. They have found a promising niche in the field of Structural Health Monitoring (SHM) which refers to the use of in situ, continuous or regular measurement and analyses of key structural and environmental parameters under operating conditions, for the purpose of warning impending abnormal states or accidents at an early stage to avoid causalities as well as giving maintenance and rehabilitation advice [4]. Fiber-optic sensor networks provide sensing solutions for almost all kind of applications and environments: from large scale structures, including bridges and other civil constructions to large natural environments [5].

Despite their marked advantages, optical fiber networks for sensors pose three challenges. The first is the need to increase the number of sensors that can be multiplexed on a single network while ensuring good signal quality. The most fundamental motivation for multiplexing fiber optic sensors is the cost, due to the fact that if the optoelectronic unit, which is the most expensive device in the network, is shared among a high number of sensing points the cost per sensing element decreases [6]. Wavelength division multiplexing is one of the best methods of multiplexing as it uses optical power efficiently, and also, it can be easily combined with other multiplexing methods, allowing a large number of sensors in a single fiber line [7]. The second demand is to ensure service continuity in the event of point failure(s) on the network. Resilience or self-healing is the ability of continue operating despite one or more points of failure on the network, which will be the key issue for practical FBG sensor systems [8]. The continued operation of the sensor network after accidental or malicious damage is of increasing importance when the structure being monitored is of high value (oil pipelines, power transmission lines, etc.); human safety is at risk (bridges, dams, chemical storage sites, nuclear plants, etc.) or perimeter security is a concern (airports, banks, etc.) [4]. The last one is to enable the possibility of remote sensing. The first two challenges are common to all optical fiber networks, while the last one, it is more specific and it is thoroughly discussed below.

Remote sensing using optical fiber systems has received an increasing attention in recent years due to the fact that it has proven to be a useful tool for monitoring a wide range of parameters in many fields. In general, the pivotal idea behind remote sensing concept is the continuous monitoring of structures from a central station located tens or hundreds of kilometers away from the field through the critical location of sensors, which send information to the central station, without the necessity of electrical power feeds in the remote locations. This remote capability allows immediate damage detection so that necessary actions can be quickly taken. Furthermore, this strategy removes the logistical inconvenience of electrical power feeds to remote locations [9,10].

More powerful remote sensing systems can find important applications in structural monitoring of large infrastructure components, such as oil or gas pipelines, ultralong bridges and tunnels, river banks and offshore platforms [9,11]. There are other promising applications of remote sensing to be highlighted. Firstly, tsunami detection and warning before their arrival to the coast, which is intended to mitigate as far as possible the disasters [12,13]; secondly, geodynamical monitoring such as surveillance of volcanic and tectonic areas which is used to predict the possible evolution towards critical stages or to detect landslides [14]; and finally, railway applications like train speed measurement, derailment, wheel defects and rail crack detection, to name but a few. Methods currently in use suffer from complexity and slow response times [15]. The methods of tsunamis detection gives an example of how of cumbersome the present techniques are: a pressure sensor, which needs electrical supply, is located in the bottom of the sea, this sensor sends acoustic signals to the buoy in the surface and it transmits the data to the satellite which launches it to the surveillance centers. Optical fiber systems, nevertheless, are secure and faster, offering very high accuracy as well as the possibility of real time measurement. These potential practical applications are the justification of this growing interest.

A criterion of classification optical fiber sensors (OFS) is according to the spatial distribution of the measurand, so, they can fit into two classes: distributed OFS, for example, those based on Brillouin scattering or Raman scattering; and discrete sensors such as fibre Bragg gratings (FBG) or Fabry-Perot.

Distributed optical fiber sensors are attracting more and more interest thanks to their wide range of potential industrial applications in strategic sectors such as energy, security, defense and transportation, among others. In particular, fiber sensors based on Brillouin optical time-domain analysis (BOTDA) exploit the dependence of the Brillouin frequency shift parameter on strain and temperature [2,16–18]. These distributed Brillouin sensors are able to measure with highly accurate over long single mode fibers exceeding several tens of kilometers. This fact allows monitoring civil structures to which optical fibers can be attached with highly accurate distributed measurements over long single-mode fiber. Nevertheless, the measurement range of these systems has a trade-off between the spatial resolution and the measurement range. For this reason, current research in BOTDA sensors has two different scopes: firstly, high-resolution sensors with centimeters spatial resolution, but for relatively short-distances fibres and, secondly, long-range BOTDA sensors, which are able to perform measurements in tens of kilometre fibres with metre resolutions. In addition, if these schemes were for remote sensing, this measurement range should be divided by a two factor, because of the fact that stimulated Brillouin scattering only occurs when incident pump light is contradirectional to the signal. For this reason, some recent published systems, whose length is 100 or 120 km [19–22], would reach 50 km or 60 km in a remote sensor system. This effective range limits their use in certain applications in which the distance to monitor is larger. We consider that a detailed discussion of this promising field goes beyond the scope of this review.

Regarding Fabry-Perot sensors [23–25], they have been developed for a variety of applications. In fact, they can provide high sensitivity, dynamic range, and response speed for measurement of temperature, strain, pressure, displacement and magnetic field. However, their development has not been as quickly as some had expected a decade ago because weak interferometric signal, costly signal processing and difficulty in wavelength division multiplexing are two major intrinsic drawbacks, limiting their applications considerably. Advances in technology and the availability of commercial Fabry-Perot sensor products will make them cost competitive in an increasing number of fields. They have some areas where rapid growth is expected during next decade: optical networks at large, and in particular, biomedical applications, smart military and commercial structures, industrial equipment monitoring, and in the oil and gas industry [26]. Consequently, few studies have been published on fiber optic networks for remote sensing including Fabry-Perot sensors, one of them reaches 50 km [27,28].

Among the wide variety of available sensors, both optical and non-optical, Fiber Bragg Gratings (FBGs) are the strongest candidates for this kind of systems due to the numerous advantages they offer. They present roughness in hostile environments, good linearity, simple demodulation concepts, electromagnetic immunity, compactness, embedding capability, commercial availability, small size and low cost. On top of that, one of the major advantages can be attributed to their wavelength-encoded information, thus the information remains immune to power fluctuations along the optical path. Another attractive benefit is their high multiplexing capability. These inherent characteristics make them attractive for applications in harsh environments and smart structures [29,30]. But, their cross-sensitivity effect, Bragg wavelength shift is simultaneously sensitive to both strain and temperature, supposes a great handicap in real engineering applications. Thereby, some methods are needed to discriminate both measurements [31,32].

Two additional important issues must be taken into consideration when fiber-optic remote sensing systems are designed. Firstly, the interrogation system, and secondly, the most suitable amplification method must be chosen to compensate for the losses undergone by the light.

As far as interrogation systems are concerned, they must allow obtaining the Bragg wavelength shift when a physical measurand acts on the grating. Consequently the determination of the Bragg wavelength allows quantifying the state of a particular parameter. The simplest method is using conventional spectrometers which are widely used in laboratories. However, high precision optical spectrum analyzers are unsuitable for real sensor systems, not only because of their high cost and large-sized but also because their slow scanning speed limits dynamic sensing [33]. Interrogation techniques must provide high sensitivity to Bragg wavelengths shifts, large measurement ranges, immunity to optical power fluctuations, low environmental sensitivity, extension to sensor multiplexing, simplicity and low cost [34]. Different methods have been reported for measuring the wavelength-encoded temperature or pressure changes of FBG. But probably, the sensor interrogation system with the greatest commercial success is based on a wavelength swept laser, which scans the reflection spectra of the Bragg gratins using the power detected by a photodiode. However, this type of systems is inherently limited in its reach due to Rayleigh scattering noise [10]. A matter of fact, the choice of the interrogation method depends upon several factors like type and range to be measured, accuracy and sensitivity required, number of sensors being interrogated and cost of the instrumentation. In conclusion, there is no a perfect interrogation method suitable for all the applications, but there is enough flexibility in order to reach an optimum interrogation method for any application [5,29,31,33,34]. By way of example, in the case of the measurement frequency is of prime importance, FBG interrogation methods can be classified as in reference [35]. Some cases of bridge monitoring where high frequency is needed to capture strain variations due to passage of speeding trucks can be found in the literature. To give an example, reference [36] is able to reach a data acquisition up to values of 500 Hz, thus, the assessment of the structural behaviour under current traffic conditions can be performed. In special, they show the results obtained during the passage of vehicles on normal traffic conditions, trucks circulation at an average speed of 50 km/h, logged with an acquisition rate of 200 Hz. As a matter of fact, recently some preliminary studies on the use of FBG sensors for monitoring railway footprints of high-speed trains, passing at speeds between 200 and 300 km/h, in the Madrid–Barcelona (Spain) high-speed line, are being developed with a FBG sensors interrogation system which allows the simultaneous monitoring of fours sensors at 8,000 samples/s [37].

The methods proposed and reported in the scientific literature, with the essential motivation of extending the distance while maintaining a good signal to noise ratio, usually include optical amplification to compensate the losses [38]. The first systems were based on broadband light sources in which case the distance was limited to a maximum of 25 km, mainly due to Rayleigh Scattering [39]. In order to surpass this limit, FBG sensors systems based on a fiber linear laser scheme are a promising alternative due to the fact that they have demonstrated several advantages, such as high resolution or high signal-to-noise ratio (SNR) against noise environments in practical applications. These systems usually include Raman amplification [40], or Raman amplification merged with other kinds of amplification: Brillouin, Erbium doped fiber or both [41–43]. To the best of our knowledge the longest distance covered by a FBG sensor system for a single FBG reported to date reached 253 km, the displacement sensor system based on a fiber loop mirror and a long period grating inside of the loop mirror is interrogated by a commercial optical time-domain reflectometer (OTDR) [44]. Following these approaches this research field is being extensively investigated at present.

Distributed Raman amplification (DRA) is the most utilized optical amplification in remote sensing networks. The systems profit from the main benefits of the Raman amplification. Firstly, it is a process inherent to germane-silicate fibers. This is a great advantage because, as no special fiber is needed to obtain Raman amplification, the transmission fiber acts as an amplifier itself. Thus, it makes the amplifier configuration very simple. Secondly, gain may be achieved at any signal wavelength simply choosing an appropriate pump wavelength. As a consequence of this second advantage, the amplification wavelength range may be extended and the ripples in the spectral gain may be reduced by using multiple pumps at different wavelengths and adjusting carefully their powers. These properties are extremely useful for the development of broadband multichannel systems, in which a large number of signals are multiplexed. Finally, as DRA name suggests, it has the capability to provide distributed amplification [45–47].

DRA is able to overcome the losses at every point along the transmission fiber, or in other words, it utilizes the transmission fiber in the network as the Raman gain medium to obtain amplification. Moreover, its use also encloses two crucial assets: improving the signal to noise ratio and reducing the nonlinear penalty due to the fact that the overall excursion that the signal experienced is reduced. The signal is prevented from decaying and increasing as much as it would make if the amplification were discrete. This improved noise performance may be used in different ways: extending the reach between repeaters, improving the transmission capacity, or the most interesting for remote sensing, expanding the total reach of the transmission system. There is, even, a forth possibility, the margin released can be used for decreasing the signal power injected in order to postpone the onset of nonlinearities or reduce their effects [48].

Although distributed Raman amplification is widely included in the remote sensor network due to its attractive advantages, it also entails some limitations which hamper the reach [45]. Two of the most severe challenges to overcome are, firstly, Rayleigh reflections and, secondly, nonlinear effects that may be enhanced in distributed Raman amplifiers since the path average signal power is higher along the transmission fiber when compared with unpumped fiber. In addition, when pump and signal are copropagating the cross-coupling between the pump and the signal also must be considered [45]. Among them, the multipath interference (MPI), primarily attributed to double-Rayleigh scattering (DRB) along the length of the transmission fiber, is the most important obstacle when the reach of the transmission system wants to be expanded [49,50]. It occurs in an optical fiber due to small inhomogeneities or microscopic variations in the refractive index [51]. An intuitive if not rigorous explanation of double Rayleigh scattering is as follows: light is scattered randomly in all directions and a small portion of the signal reflects back due Rayleigh backscattering, this reflection is amplified by the Raman gain and reflected once again due to Rayleigh backscattering and it can be recoupled into the forward direction. Therefore, the double Rayleigh scattering is amplified twice and it scales as the length since the gain is accumulated over tens or hundreds of kilometers. In summary, DRB produces a large number of low power replicas of the signal with random delays and phases that propagate along with the signal. The problem associated with this extra noise is due to its inherent characteristic of occupying the same spectral region as the signal and consequently is not easily distinguish it from the real signal. So far electrical measurements have been preferred for characterization of DRB [52]. The use of multiple gain stages optically isolated one from another is a well-known technique of reducing the DRS, but usually when distributed amplification is used is not possible to include [53]. Another way to limit the DRS is to combine the Raman amplifier with and EDFA to form a hybrid amplifier [51].

Table 1 summarizes the state of the art of remote sensing systems for optical fiber sensors in chronological order taking into account the most representative characteristics of the systems. When the first fiber-optic sensor networks were designed with some kilometers [54,55], they were considered as remote sensor systems.

However, nowadays fiber-optic sensor systems can cover hundreds of kilometers. Because of this, the summary table only contemplates remote sensor systems with, at least, some tens of kilometers.

This section is devoted to explain more carefully the most representative remote fiber-optic sensor systems for fiber optic sensors presented in Table 1, discussing their schemes, pros and cons. They are going to be evaluated not in chronological order as in the list, but taking into account the system length.

In the course of the previous section, the key parameters in the design of long-distance sensor systems for remote sensing have been discussed. In this section, we want to gather briefly the most important points.

The first consideration is about amplification. When the transmission channel is composed by a span of hundreds of kilometers, if the modulated signal by the sensors wants to be detected, the accumulated losses must be compensated. The most widespread way is the addition of distributed Raman amplification [39,40,57–59,61,65,66,69] due to its advantages. Other possibility is the used of hybrid amplifiers such as the combination of Raman with Brillouin amplification [68] or Raman with EDF amplification [11,42]. Whatever the utilized amplification, its inherent noises and other awkward effects must be counterbalanced. Only a few systems have achieved long distance without amplification [44,62].

Secondly, we have to consider the network topologies. The most common are star and bus. The first one usually uses postamplifiers while bus topology employs distributed amplification. In particular, bus architecture is one of the most widely used, mainly owing to its simple cabling requirements when compared with star networks, and its potential to increase the degree of integration of different sensors [6]. An improved version of the single bus is the double bus or ladder structure. It is composed by two different fibers: the first one is used to send the optical signal towards the sensors, and the second one collects the signal from the sensors. This enhanced topology improves the signal to noise ratio and offers the possibility of multiplexing hundreds of sensors [79-81].

The kind of sensors and its location in the network also must be contemplated. In reference [57,58], any kind of optical sensors can be multiplexed in the network which confers an extra degree of freedom in its design. Thus, it can be adapted to every application: distributing the sensors at critical points and even combining different kind of sensors. In most of the developed remote sensor schemes, however, the chosen sensors are FBGs because of its appealing features [41,43,68,69]. The FBGs disposal within the systems is usually in series or in parallel. The first one is a good option when the network is not based on a multiwavelength fiber laser [62,69] because, in this case, achieving power equalization for all the channels is not an easy task which restricts severely the number of multiplexed FBGs [39,59]. This serial topology allows utilizing the power efficiently but the performance of one sensor depends directly on the previous ones. Parallel arrangement, on the other hand, is more suitable for systems based on multiwavelength fiber lasers if channel equalization is considered [66]. Each behavior sensor is independent from the others, in consequence, it makes easier the development of self-healing networks, enables an easy repair of the sensors when needed or exchange the kind of sensor depending on the needs. On the other hand, the power requirements of parallel disposition are much higher.

One of the most restricting factors in the remote sensor systems is the accumulated and amplified Rayleigh scattering. Unlike other broadband noise like ASE, Rayleigh scattering is centered at the same wavelengths as the signal. For this reason, avoiding this noise or distinguishing the contribution of the noise and the signal are complex tasks. Till now, two different ways have been presented. The first one uses an heterodyne detection system [43,68] which relegates the Rayleigh noise to low frequencies and the signal is transferred to noise-free RF frequencies. The second one includes a high-speed pulse-driven swept light source, which is turned off immediately after wavelength sweeping finishes and so the Rayleigh noise is reduced [44,62].

Multiplexing or self-healing capability are other important issues to address as well. In real applications of remote sensor systems, at least, tens of sensors are needed to control the whole structure under survey because either they usually have huge dimensions or the parameters to be assessed are several. In consequence, apart from reducing the cost and complex of the networks, multiplexing capability helps to have an adequate evaluation of all key parameters. Moreover, resilient remote sensor systems must concentrate the attention. Till now, we dare to say that only one scheme has been proposed combining long distance and resilient [66]. Resilience is the ability of continue operating despite one or more points of failure on the network, which will be the key issue for practical FBG sensor systems. The continued operation of the sensor network after accidental or malicious damage is of increasing importance when the structure being monitored is of high value; human safety is at risk or perimeter security is a [4]. Much work must be dedicated to this point.

At all events, no combination of the aforementioned aspects is the answer for all the applications, but there is indeed enough flexibility in order to reach an optimum scheme for any application. To finish, we would like to point out that the foremost aspect to considerer in the design of real remote sensor systems is the economic conditions. Thus, simplicity and cost always must be in mind.

Multi point sensing fiber optic networks for remote sensor systems are a promising technology with a wide field of practical applications but with some challenges to face. This overview shows the great evolution of these systems in recent years. The most representative systems are explained in ascending order of the system length: beginning with the shortest 22 km and 50 km to finish with the longest till now of 250 km. Along the discussion, the foremost factors on such matter, as kind of amplification, network topology, pump efficiency, type of sensors, noise to overcome, multiplexing and self-healing capability, simplicity and cost are taking into consideration to evaluate each system.