DE102009020115B4 - Error correction method for a fiber optic sensor and measuring device - Google Patents

Abstract

An error correction method for a fiber optic sensor, comprising (a) providing backscatter measurement data of an optical polymer fiber; (b) determining a measured scattering profile from the backscatter measurement data; (c) providing a model of the optical polymer fiber, wherein the model is adapted to describe a time course of the backscattering behavior of the optical polymer fiber; (d) determining the time course of the backscatter behavior using the model; and (e) compensating the determined time course of the backscattering behavior.

Description

The present invention relates to an error correction method for a fiber optic sensor, in particular for a sensor with a polymer optical fiber. Furthermore, a measuring device is provided which carries out such an error correction method.

Fiber optic sensors are increasingly playing a role in monitoring extended structures such as buildings, dams or dykes. The fiber optic sensors can be used, inter alia, for strain measurement, since the scattering properties of the fiber differ for different strain states. Typically, backscatter measurements are used for such studies. These backscatter measuring methods allow to detect a local distribution of the measured quantity along the fiber. Among the backscatter measuring methods, the optical backscatter measurement in the time domain, also known as OTDR (Optical Time Domain Reflectometry), is the most widely used measuring technique. For example, in the OTDR method, a short pulse of light is radiated into the fiber and the backscattered light is recorded as a function of time. The transit time of the light in the fiber is 2 × d / c × n, where d is the simple distance along the fiber to the scattering point, c is the speed of light, and n is the effective refractive index in the fiber. The quantity c / n thus corresponds to the effective velocity v of the light in the fiber. In this way, a relationship between the time course and the local course of the signal can be established and the location of a scattering center in the fiber can be located. In addition to time domain analysis (OTDR), those skilled in the art are also familiar with correlation range analysis and frequency domain analysis as backscatter measurement methods.

Typically, glass fibers are used as optical fiber sensors, fiber optic strain sensors made of glass fiber already being known in the art. However, glass fibers are not suitable for measuring larger strains, for example over 1%. In contrast to the glass fiber sensors, sensors based on optical polymer fibers, in short POF (English: Polymer Optical Fiber), are also suitable for strain measurement for strains of more than 45%.

However, the amount of backscattering in POF sensors from about 1% strain no longer depends solely on the elongation but also on the time course of the elongation, d. H. For example, the time of stretching in relation to the time of measurement but also how fast the strain was performed.

Due to the time dependence of the backscatter the evaluation of the measuring signals is made more difficult. The error caused by the temporal decrease in scattering can make a reliable quantitative evaluation of the scattering profile in a strain profile for strains above about 2-5% virtually impossible.

In view of the above, an error correction method according to claim 1 and a measuring apparatus according to claim 16 are provided. Other aspects, details, advantages and features of the present invention will become apparent from the dependent claims, the description and the accompanying drawings.

According to one embodiment, an error correction method for a fiber optic sensor is provided. First, backscatter measurement data are provided, in particular backscatter measurement data obtained from a polymer optical fiber. From the backscatter measurement data, a measured scattering profile is determined. Furthermore, a model of the optical polymer fiber is provided. The model is adapted to describe a time course of the backscattering behavior of the optical polymer fiber. Based on this model, the time course of the backscatter behavior is determined. The time response of the backscatter behavior thus determined by the model is now used to compensate for the time behavior of the sensor fiber. In this way, a corrected strain profile can be obtained.

The described error correction method makes it possible to compensate for temporal changes in the backscattering behavior of the sensor fiber. In this way, the evaluation of the backscatter measurement data can be improved. In particular, the error correction method allows a reliable quantitative evaluation of backscatter measurement data from POF sensors. These are typically backscatter measurements obtained by time domain optical backscatter, optical backscatter in the correlation domain, or optical backscatter in the frequency domain. Furthermore, a corrected elongation profile of the optical polymer fiber can be determined, wherein due to the error correction method, even larger strains, in particular strains in the range of 3% to 50% of the fiber length, can be determined quantitatively.

According to one embodiment, the model used for error correction has a non-linear time response.

According to one embodiment, the model is designed as a mechanical model. For example, the mechanical model indicates at least one spring and at least one attenuator. In this case, the at least one spring and the at least one attenuator can be arranged in parallel or in series. Of course, the model may comprise a plurality of springs and / or attenuators, which are each arranged in parallel and / or in series. According to a development, the at least one attenuator has a non-linear behavior. In particular, the voltage on a shock absorber, depending on the first time derivative of the deflection on an attenuator, have a polynomial behavior. According to further embodiments, the model may also have other members that exhibit behavior other than springs or attenuators.

By models, as described above, the time-back scattering behavior of the sensor fiber can be predicted sufficiently well. In particular, mechanical models of the sensor fiber are well suited for describing the backscattering behavior. The mechanical models can be displayed relatively easily and can therefore be solved quickly and stably numerically.

Typically, the model is represented by a differential equation system that is solved to obtain the time course of the backscattering behavior. For some model classes, as described below, the differential equation system can be easily solved by the Euler method. In this way, the error correction with relatively little computational effort and in particular can be performed very quickly.

According to a further exemplary embodiment, a model scattering profile is determined on the basis of the model. The model scattering profile is then compared with the measured scattering profile, after which at least one parameter of the model is adapted in such a way that an error between the measured scattering profile and the adapted model scattering profile becomes as small as possible.

In this way deviations between the model and the real behavior of the fiber can be reduced. As a result, the model of the sensor fiber can be improved based on actually measured values, and then provides more accurate values for the time response of the backscatter. In particular, the method can be carried out iteratively in such a way that several measurements of backscatter measurement data take place in succession. For each of these measurements, the model scattering profile is then adapted as described, wherein in the adaptation of each of the adapted model scattering profile of the previous measurement is assumed. In this way, the model improves with each measurement and the error correction gets better and better.

According to another embodiment, a model strain profile is determined from the model, the model strain profile is compared to the measured strain profile, and at least one parameter of the model is adjusted such that an error between the measured strain profile and the fitted model strain profile becomes as small as possible , In this variant, therefore, the expansion profile is optimized directly, without the intermediate step must be gone through a correction of the measured scattering profile.

According to another embodiment, there is provided a measuring device comprising a polymer optical fiber, a light source adapted to introduce light pulses into the optical polymer fiber, a detector adapted to receive backscattered light from the optical polymer fiber as backscattered measurement data, and an evaluation unit which is arranged to execute on the backscatter measurement data the error correction method described above. Typically, the optical polymer fiber comprises polymethyl methacrylate, a fluoropolymer or a perfluoropolymer as an ingredient. Furthermore, the optical polymer fiber may be formed as gradient index fiber or step index fiber.

In particular, some of the optical polymer fibers can be subjected to elongation of up to 45% without damage.

Reference to the accompanying drawings, embodiments of the present invention will now be explained. Showing:

1 a schematic representation of a measuring device according to an embodiment;

2 a schematic representation of a measuring principle;

3 a spectral distribution of the relative increase in intensity of the scattered light for different strains;

4 the time decrease of the increased spread with time for an elongation of 20%;

5 a flowchart of an error correction method according to an embodiment;

6 a model as used in an error correction method according to an embodiment;

7 an iterative method for evaluation in several successive measurements.

The 1 shows a schematic representation of a measuring device 100 according to an embodiment. The measuring device 100 includes light sources 110 . 112 configured to emit light pulses into an optical fiber sensor 120 initiate. In the present embodiment, the measuring device 100 a first light source 110 on, the light pulses in a front end 122 the sensor fiber 120 initiates, as well as a second light source 112 , the light pulses in a rear end 124 the sensor fiber 120 initiates. It should be noted that measurements at the front and at the rear end of the sensor fiber 120 be performed consecutively. Therefore, it is also possible that the front end and the rear end of the sensor fiber are connected to the same light source and the same detector. This embodiment is in 1 not shown, but readily familiar to those skilled in the art. In principle, however, the measurements can also be carried out if the light is only fed in at one end of the sensor fiber.

In typical embodiments, the fiber sensor is implemented as optical polymer fiber (POF). In this case, the POF sensor as part of polymethylmethacrylate (PMMA), a fluoropolymer or a perfluoropolymer have. Furthermore, the optical polymer fiber may be embodied as gradient index fiber or as step index fiber. According to one embodiment, a sensor fiber of PMMA with a core diameter of 1 mm and a Stufenindexpofil is used. Such fibers are available, for example, from Mitsubishi, Toray and Asahi. The fibers are also available with different core diameters for the same material. The fibers can be tens of percent stretched without damage (> 45%) and have a with the strain steadily increasing scattering. Furthermore, these fibers exhibit attenuation in the range of 160 dB / km at the common wavelength of 650 nm. In addition, the input light pulses due to the step index profile along the fiber greatly. A gradient index fiber of PMMA with 1 mm core diameter is available from the company optimedia. This fiber has increased scattering only in the elongation range between 1.5% and 6%. In addition, this fiber is significantly more fragile than the above-mentioned step index fibers. All of these PMMA fibers are typically light sources 110 . 112 used, which work in the visible wavelength range between 400 nm and 700 nm.

According to other embodiments, for the fiber sensor 120 Gradient index fibers manufactured by Asahi Glas and Chromis, which consist of the perfluoropolymer Cytop ^® . Attenuation and pulse expansion are comparatively low for these fibers and are in the range of 30 dB / km. However, an increase in scattering can only be observed in the expansion range between 1.5% and 3%. These fibers can also be used in the visible range, but are typically used with light sources operating in the wavelength range from 850 nm to 1300 nm.

Polymer fibers can be made, for example, from thermoplastics. Among the thermoplastics, for example, optical fibers made of polycarbonate or polystyrene can be produced in addition to PMMA. Furthermore, optical fibers can be made from cyclic polyolefins. Other fluoropolymers suitable for making optical fibers include hexafluoroisopropyl 2-fluoroacrylate (HFIP 2-FA), polytetrafluoroethylene (PTFE), tetrafluoroethylene-hexafluoropropylene (PFE), and tetrafluoroethylene-perfluoroalkylvinyl ether (PFA). Furthermore, optical fibers can also be made with deuterated polymer based on PMMA. In addition to these thermoplastics, optical fibers can also be made from elastomers such as polysiloxanes.

Furthermore, the measuring device comprises 100 detectors 130 . 132 that are set up, from the fiber sensor 120 backscattered light as backscatter measurement data. In this case, an optical backscatter measurement in the time domain, an optical backscatter measurement in the correlation domain, or an optical backscatter measurement in the frequency domain can be performed. By way of example but not limited to the general public, an optical time domain backscatter measurement (OTDR) is described here. These are the detectors 130 . 132 set up to register the scattered light time-resolved. According to one embodiment, both at the front and at the rear end of the sensor fiber 120 optical couplers 140 . 142 provided, which decouple the scattered light from the fiber and to the respective detector 130 . 132 conduct.

Finally, the measuring device comprises 100 another evaluation unit 150 , The evaluation unit 150 receives the from the detectors 130 . 132 and is adapted to perform on the backscatter measurement data an error correction method described below according to an embodiment of the present invention. Typically, the evaluation unit comprises 150 a computer on which a computer program executing the error correction method described below is executed. The evaluation unit 150 is typically further set up, from the backscatter measurement data or from the corrected one obtained by means of the error correction method Backscatter data an expansion profile of the sensor fiber 120 to obtain. In particular, the expansion profile becomes a stretched part 125 the sensor fiber 120 receive.

In 2 is the underlying measuring principle shown schematically. First, the sensor fiber is at the front and at the rear end 120 by means of the detectors 130 . 132 respective measured signals recorded time-resolved. In the measurement signals, the corresponding fiber sections are then identified on the basis of reference reflections (eg Fresnel reflections at the beginning or end of the fiber or at plug connections). It should be noted here that the signals, viewed from the two reference points, are temporally inverted relative to one another since the scattering events were recorded at the opposite ends of the sensor fiber. This is followed by a sharpening of the signals in order to compensate for the pulse broadening caused by the sensor fiber. For example, such a sharpening can be done with the aid of a blurring matrix, wherein the pseudo inverse of the blur matrix is used for sharpening. Typically, the pseudoinverse of the blur matrix is determined once for a particular meter and then stored. The sharpening of the signal is then achieved by simply multiplying the signal vector by the pseudoinverse of the blur matrix. Subsequently, the signal obtained at the rear end of the sensor fiber is time-inverted in order to obtain correct position values when superimposing the two measuring signals. By superimposing the two measurement signals thus obtained, the attenuation profile, that is to say the attenuation at a respective location of the sensor fiber, and the scattering profile, that is to say the intensity of the backscatter at a respective location of the sensor fiber, can be separated from one another. Typically, the intensity of the backscatter at a particular location of the sensor fiber is indicated by the intensity of the scattered light, typically normalized to the light intensity present, at the location.

In 3 shows a spectral distribution of the relative increase in intensity of the scattered light for different strains of the sensor fiber in the range of 1% to 40% elongation. The plot is logarithmically chosen to show the typical λ ^-4 dependence that results from Rayleigh scattering in the sensor fiber. This in 3 shown scattering behavior of the sensor fiber for different strains now allows to establish a relationship between the scattering profile and the elongation profile of the fiber. In other words, over the in 3 As shown, the scattering intensity at a location of the sensor fiber can be translated into a strain at that location. The relative increase in scattering in the stretched fiber compared to the unstretched fiber can thus be assigned an elongation value. In this way, the sensor fiber can be used as a strain sensor.

However, the amount of backscatter in POF sensors from about 1% strain no longer depends solely on the strain itself but is also time-dependent. In particular, the POF sensors show a decrease in the increased scattering with time, ie a relaxation behavior. This is in 4 shown in which the increase in the spread is plotted against time. The 4 clearly shows the decrease in the increased due to the elongation scattering with time. In the example of 4 was set an elongation of 20%, in principle, this decay of the scattering but can be practically observed for all strains above about 1%. In general, the stronger the impressed strain, the stronger the relative decrease of the measured quantity with time. Furthermore, the elongation can also depend on the time course of the expansion, ie, for example, the time of stretching in relation to the time of measurement but also how fast the stretching was carried out.

Due to the time dependence of the backscatter the evaluation of the measuring signals is made more difficult. In other words, after a certain amount of time, the strain appears smaller than it really is. Due to the uncorrected measurement data, one would thus underestimate the strain in the sensor fiber systematically, whereby the error is time-dependent. The error caused by the temporal decrease in scattering thus makes a reliable quantitative evaluation of the scattering profile in a strain profile for strains above approximately 2-5% impossible without corresponding error correction.

In 5 Fig. 10 is a flowchart of an error correction process 500 shown according to an embodiment. In a first step 510 Backscatter measurements of an optical polymer fiber are provided. Subsequently, a measured scattering profile from the backscatter measurement data is determined (step 520 ). It has already been described above how a scattering profile can be determined from the backscatter measurement data, so that a further explanation of this is omitted here. In one step 530 a model of the optical polymer fiber is provided. The model is adapted to describe the time course of the backscattering behavior of the sensor fiber, in particular an optical polymer fiber. Typically, the model has an inhomogeneous stress distribution. Such an inhomogeneous stress distribution is typically also found in the sensor fibers. Furthermore, the model typically exhibits a non-linear time behavior, as is also found, for example, in the sensor fibers (cf. 4 ). Suitable models are described in more detail below. Based on the model will now be in step 540 the time course of the Backscatter behavior determined. On the basis of the time course of the backscattering behavior determined in the model, this time behavior is now compensated in the measured scattering profile.

The described error correction method makes it possible to compensate for temporal changes in the backscattering behavior of the sensor fiber. In this way, the evaluation of the backscatter measurement data can be improved. In particular, the error correction method allows a reliable quantitative evaluation of backscatter measurement data from POF sensors. Furthermore, a strain profile of the optical polymer fiber can be determined on the basis of the corrected backscatter profile, wherein due to the error correction method even larger strains, in particular strains in the range of 3% to 50% of the fiber length, can be reliably determined quantitatively.

In 6 is a model 600 as used in an error correction method according to an embodiment. Here, a mechanical model is used, ie a model consisting of coupled mechanical elements. The model according to 6 has two mutually parallel branches. The first branch comprises a first spring S ₁ and disposed with the first spring S ₁ in series first damping element _d1. The second branch comprises a second spring S ₂ and disposed with the second spring _S2 in series with second damping element D. ₂ The deflection of the first and the second spring is time-dependent and given by z ₁ or z ₂ . The two attenuators D ₁ , D ₂ have a non-linear behavior. The model 600 can be described by a differential equation system, which is numerically solvable, for example by means of the Euler method.

The model described above is capable of quantitatively correctly reflecting the decrease in scattering intensity over time. In this case, the voltages at the individual springs F _n ¹ = z _n ¹ · S ^{1 are} converted into a scatter factor by a function g (F _n ¹ ). The index i stands for the different branches in the model and the index n for the measurement or the time of the measurement. In the simplest case, the function g corresponds to a weighted sum of the components of F _n ¹ , so that the conversion takes place via a simple multiplication with these constants. The sum of the individual scattering factors then forms the scattering factor, which can also be found in the scattering profile. The voltage F ¹ (t) on a spring is time-dependent and given by the product of spring constant S ¹ and deflection z ¹ F ¹ (t) = z ¹ (t) · S ¹ .

As a result, the scattering factor calculated from F _n ¹ is also time-dependent, as a result of which, for example, the in 4 Describes relaxation behavior described. As already explained, the attenuators, as well as the springs, can be linear (F ¹ (t) d / dt (z (t) -z ¹ (t)) * D ¹ ) or have a non-linear time response (z ¹ (t) * S ¹ ) ^m = d / dt (z (t) -z ¹ (t)) * D ¹ .

Here, d / dt (z-z ¹ ) is the change of the attenuation of the attenuator with time and m is an exponent. The variable z is equivalent to the strain ε at the respective location of the fiber. In this respect, therefore, the temporal change of z is equivalent to the temporal change of the strain ε. In the simplest case, it is assumed that the change in the elongation is uniformly distributed over the time interval between the measurements dt. According to other embodiments, however, models with different temporal expansion curves are also used.

In principle, other than polynomial functional characteristics (m = 1, 2, 3...) Can also be used in the model. However, even the relatively simple model shown with two nonlinear attenuators arranged in parallel shows good results for m = 2. It should be noted, however, that in principle the number of springs and attenuators in the model is arbitrary. Likewise, the topology of the network, which describes the couplings of the mechanical members together, is basically arbitrary. However, it makes sense to quickly and easily determine the timing, a simple network topology with as few nodes as possible, d. H. mechanical links, to use.

The model first gives a series of ordinary differential equations for F ¹ (t). Furthermore, the constants S ¹ and D ^{1 are} selected identically at each location of the fiber. However, there is for each location along the fiber in each case an expansion z or ε and own displacements z ^{1 of} the respective springs. By solving the differential equation system for each location along the fiber, the F ¹ can now be determined. The time dependence is taken into account so that the F ¹ (t) in each case for measurement times n = 1, 2, 3, ... as F _n ^{1 are} determined. For this purpose, the starting values z _n , z _n ^{1 of} the model are used from previous measurements, as well as an estimated Δz and the time interval Δt _n between the current and the preceding measurement. The scatter factors for the respective location of the fiber can then be newly determined from the F _n ¹ . These can now be used for comparison with the measured dispersion factors. The determined time dependence of the local scattering factors can now be used to compensate for the time dependence in the measured scattering profile. In other words, the time-dependent decrease of the scattering intensity can now be taken into account.

In 7 an iterative method for evaluation in several consecutive strain measurements is shown. In this case, first the deflections z _n-1 and z _n-1 ^{1 of} the preceding measurement and the time interval Δt _n for this measurement are used as input variables of the model. Furthermore, a local change of the strain Δε _{n is} estimated and used as an input quantity in order to calculate the resulting scattering profile therefrom. The advised value Δε _n for the change in the elongation is then optimized until the calculated model scattering profile agrees as well as possible with the measured scattering profile. In addition to the strain values for each location point along the sensor fiber, the states for all z _n ¹ are also stored for the calculation of the strain profile from the next measurement, since these again serve as the starting point for the next calculation. In other words, the model contains a separate state for each location along the fiber with the local state variables z _n or ε _n and z _n ¹ , which sufficiently describe the entire history of the fiber and are adapted with new measurement data after each calculation.

In the iterative method, therefore, a model scattering profile is determined on the basis of the model, the model scattering profile is compared with the measured scattering profile, and then the estimated elongation adjusted in such a way that the error between the measured scattering profile and the matched model scattering profile is as far as possible gets small. For example, this optimization can be done with the least squares method, but other optimization methods are also suitable. Typically, as described above, the method is carried out iteratively in such a way that several measurements of backscatter measurement data are made and for each measurement the model scattering profile is adapted. In the adaptation, it is assumed in each case of the adapted model scattering profile of the previous measurement.

The present invention has been explained with reference to exemplary embodiments. These embodiments should by no means be construed as limiting the present invention.

Claims (20)

Error correction method for a fiber optic sensor, comprising (a) providing backscatter measurement data of a polymer optical fiber; (b) determining a measured scattering profile from the backscatter measurement data; (c) providing a model of the optical polymer fiber, wherein the model is adapted to describe a time course of the backscattering behavior of the optical polymer fiber; (d) determining the time course of the backscatter behavior using the model; and (e) compensating for the determined time course of the backscattering behavior. The error correction method of claim 1 further comprising determining a measured strain profile of the optical polymer fiber from the backscatter profile. Error correction method according to one of the preceding claims, wherein the model has an inhomogeneous stress distribution. An error correction method according to any one of the preceding claims, wherein the model has non-linear timing. An error correction method according to any one of the preceding claims, wherein the model is a mechanical model. The error correction method of claim 5, wherein the mechanical model comprises at least one spring and at least one attenuator. Error correction method according to claim 6, wherein at least one spring and at least one attenuator are arranged in parallel. Error correction method according to claim 6 or 7, wherein at least one spring and at least one attenuator are arranged in series. An error correction method according to any one of claims 6 to 8, wherein the at least one attenuator has non-linear timing. The error correction method of claim 9, wherein the first temporal derivative of the at least one attenuator has a polynomial behavior. Error correction method according to one of the preceding claims, wherein the model is represented by a differential equation system and the differential equation system is solved in step (d). An error correction method according to any one of claims 2 and 3 to 11 as far as dependent on claim 2, wherein a model strain profile is determined from the model, the model strain profile is compared to the measured strain profile, and at least one parameter of the model is fitted in such a way that a mistake between the measured strain profile and the adapted model-strain profile is as small as possible. Error correction method according to one of the preceding claims, wherein a model scattering profile is determined based on the model, the model scattering profile is compared with the measured scattering profile, and at least one parameter of the model is adapted in such a way that an error between the measured scattering profile and the adapted model scattering profile is as small as possible. Error correction method according to claim 12 or 13, wherein the method is performed iteratively in such a way that a plurality of measurements of backscatter measurement data takes place and for each measurement the model expansion profile or the model scattering profile is adapted, wherein in the adaptation of each adapted model strain profile or the model scattering profile of the previous measurement is assumed. Error correction method according to one of the preceding claims, wherein the backscatter measurement data were obtained by an optical backscatter measurement in the time domain, an optical backscatter measurement in the correlation domain, or an optical backscatter measurement in the frequency domain. Measuring device comprising an optical polymer fiber, a light source adapted to introduce light pulses into the optical polymer fiber, a detector configured to receive backscattered light from the optical polymer fiber as backscatter measurement data, and an evaluation unit which is set up to execute an error correction method according to one of the preceding claims on the backscatter measurement data. A measuring device according to claim 16, wherein the optical polymer fiber comprises polymethyl methacrylate, a fluoropolymer or a perfluoropolymer as an ingredient. A measuring device according to claim 16 or 17, wherein said optical polymer fiber is a gradient index fiber or a step index fiber. A measuring device according to any one of claims 16 to 18, wherein the optical polymer fiber can be subjected to elongation of up to 45% without damage. Measuring device according to one of claims 16 to 19, wherein the optical polymer fiber is a thermoplastic, in particular polycarbonate, polystyrene, a cyclic polyolefin, hexafluoroisopropyl 2-fluoroacrylate (HFIP 2-FA), polytetrafluoroethylene (PTFE), tetrafluoroethylene-hexafluoropropylene (PFE), tetrafluoroethylene Perfluoroalkylvinyl-ether (PFA) or a deuterated polymer based on PMMA, or an elastomer, in particular a polysiloxane.

Images