Temperature measurement and damage detection in concrete beams exposed to fire using PPP-BOTDA based fiber optic sensors

A telecommunication-grade, single-mode optical cable with a polymer sheath and aramid yarn was used as a distributed temperature sensor. As illustrated in figure 1, the optical fiber had a tight polymer buffer (diameter: 900 μm), a polymer outer coating (outer diameter: 242 μm), a polymer inner coating (outer diameter: 190 μm), a silica (glass) core (diameter: 8.2 μm), and a silica (glass) cladding (outer diameter: 125 μm). Light waves propagate along the optical fiber with total internal reflection occurring at the interface of the core and the cladding [16]. The polymer buffer and coatings protected the glass from mechanical impact and from abrasion and environmental exposure. The buffer and coatings are burned off at approximately 300 °C to 400 °C. However, the glass core and cladding can sustain temperatures above 1000 °C. The optical fiber could slide in the sheath with negligible friction, so the distributed sensors measure Brillouin frequency shifts due to temperature changes. Once calibrated, the sensors can be used to evaluate the temperature changes from the measured Brillouin frequency shifts.

Zoom In Zoom Out Reset image size

PPP-BOTDA measures stimulated Brillouin backscattering light at any point along the length of an optical fiber using a pulsed pump wave and a counter-propagating continuous probe wave [21]. When the frequency difference between the pump and probe waves matches the frequency associated with the vibration of a crystal lattice, the probe wave is amplified and the frequency difference is referred to as the Brillouin frequency. The Brillouin frequency depends on the optical properties of the fiber, which change with the strain and temperature applied on the optical fiber. The arrival time of backscattered light is related to the location in the optical fiber where strain and temperature are applied. For silica-based single mode optical fibers, the Brillouin frequency ν_B typically ranges from 9 GHz to 13 GHz for light wavelengths of 1.3 μm to 1.6 μm and can be calculated by [21]:

$$\begin{eqnarray}&&{\nu }_{{\rm{B}}}=\displaystyle \frac{2{\nu }_{0}}{C}n{V}_{{\rm{a}}}\end{eqnarray} \tag{ 1 }$$

where ν₀ denotes the frequency of an incipient light wave, n denotes the refractive index of the optical fiber, V_a denotes the speed of acoustic wave in the fiber, and C (=3.0 × 10⁸ m s⁻¹) denotes the speed of light in a vacuum. The speed of the acoustic wave is given by [22]:

$$\begin{eqnarray}&&{V}_{{\rm{a}}}=\sqrt{\displaystyle \frac{(1-\mu )E}{(1+\mu )(\mathrm{1-2}\mu )\rho }}\end{eqnarray} \tag{ 2 }$$

where μ, E, and ρ respectively denote the fiber's Poisson's ratio, Young's modulus, and density.

The refractive index and density are related to both temperature and strain. The Poisson's ratio and Young's modulus are related to temperature only. Therefore, the Brillouin frequency varies with strain and temperature changes in the fiber. For a relatively small change of strain Δε and temperature ΔT with respect to their calibration values, the Brillouin frequency shift Δv_B can be linearly related to the strain and temperature changes [21]:

$$\begin{eqnarray}&&{\rm{\Delta }}{\nu }_{{\rm{B}}}=\,{C}_{{\rm{\varepsilon }}}{\rm{\Delta }}\varepsilon +{C}_{{\rm{T}}}{\rm{\Delta }}T\end{eqnarray} \tag{ 3 }$$

where C_ε and C_T denote the strain and temperature sensitivity coefficients, respectively. For an optical fiber that is free of strain change, the Brillouin frequency shift depends upon temperature change only. For large changes in temperature, the relationship is nonlinear [30]. When both strain and temperature change, an additional fiber that is free of strain can be incorporated for temperature compensation.

The sensitivity of Brillouin frequency to temperature was experimentally calibrated. The optical fiber was passed through an electric tube furnace where the temperature was monitored using a calibrated thermocouple. The furnace temperature was monotonically increased from 22 °C to 800 °C. Spatially-distributed Brillouin gain spectra were measured along the length of the test optical fiber using a Neubrescope^® data acquisition system (Model NBX7020). The five curves in figure 2(a) represent the one-point Brillouin gain spectra measured at five temperatures [30]. The peak of each curve represents the corresponding Brillouin frequency at a given temperature. The Brillouin frequency was then determined using a Lorentz curve-fitting algorithm and plotted against temperature, as shown in figure 2(b). The Brillouin frequency increased with temperature at a decreasing rate as the temperature increased from room temperature to 800 °C. The relationship between the Brillouin frequency and temperature was fitted using a parabolic equation with the coefficient of determination (R²) equal to 1.000.

Zoom In Zoom Out Reset image size

Each concrete beam was tested in the compartment fire setup shown in figure 3. The setup, which was located under a 6 m × 6 m exhaust hood, consisted of a burner rack, an enclosure and water-cooled supports for the beam specimen. The rack supported four independent natural gas diffusion burners made of sheet metal, each with dimensions of 300 mm × 300 mm × 140 mm (length × width × height). The two middle burners were fueled with natural gas from their bottom through the burner cavity and a 20 mm thick ceramic fiber blanket for gas distribution. The burners were manually regulated using a needle valve on the gas line, and the energy content of the supplied gas was measured with an expanded uncertainty of less than 2.4% [31]. An enclosure constructed of steel square tubes, cold-formed steel C-channels, and gypsum boards lined with refractory fiber board formed the fire test space above the burner rack. The enclosure was open at the bottom and the two end faces in the longitudinal direction of the beam, creating the compartment flame dynamics depicted in figure 3. The heated area created by the enclosure had dimensions of 380 mm × 400 mm × 1830 mm (height × width × length), in which the entire beam was engulfed in flame.

Zoom In Zoom Out Reset image size

The test beams were simply supported at a clear span of 0.5 m on two supports constructed from 1-1/2'' Schedule 40 pipe (outer diameter: 48 mm), which were, in turn, supported on concrete blocks. The only mechanical load on the beam was its self-weight. The supporting pipes were water-cooled so that the exiting water temperature did not exceed 50 °C, which limited undesired thermally-induced movement of the supports.

Four concrete beams were tested, designated as Beam 1, Beam 2, Beam 3, and Beam 4 in figure 4. Each beam was 152 mm in depth, 152 mm in width, and 610 mm in length. Normal weight concrete with nominal 28-day compressive strength of 42 MPa was used (concrete casting and curing are detailed in section 3.3). A deformed steel reinforcing bar with 25.4 mm diameter was placed approximately in the center of each beam. The geometry of the beams was not intended to be representative of construction practice, but rather to provide simple specimen geometry to test the optical fibers. Each test beam was instrumented with four, glass sheathed, K-type, bare-bead thermocouples, and one distributed fiber optic sensor. The four thermocouples (TC1, TC2, TC3, and TC4) were embedded within the beam at the quarter span and the mid-span as depicted in figure 4. TC1 was at the quarter span at the 1/2 depth of the beam and near the rebar; TC2, TC3, and TC4 were at mid-span at the 1/4, 1/2, and 3/4 depth of the beam, respectively. Additional thermocouples were deployed on the inner wall of the compartment and throughout the test setup to characterize the test environment and monitor safety-relevant temperatures. The optical fiber (in figure 1) was loosely passed through a protective polymer sheath (diameter: 1 mm) and could freely slide within the sheath. The sheath had a thermal conductivity of 3.0 W m⁻¹ · K⁻¹, which was slightly higher than the thermal conductivity of the concrete, which is typically less than 2.0 W m⁻¹ · K⁻¹ [32]. The sheath was in direct contact with concrete and isolated the optical fiber from the effect of strain in concrete. Thus, the distributed sensor was subjected to temperature change only. It was deployed within the concrete following the path illustrated in figure 4. Capital letters A to G designate key locations on the optical fiber. For example, A and G corresponded to the entrance and exit points of the optical fiber with respect to concrete. The optical fiber was firmly attached on a fishing wire every 150 mm using superglue; and the fishing wire was tightly fixed at both ends to the walls of casting mold to keep the fiber straight during casting of the concrete.

Zoom In Zoom Out Reset image size

Data from the fuel delivery system and thermocouples were measured continuously using a National Instruments data acquisition system (NI PXIe-1082). Thermocouple data were recorded using 24-Bit Thermocouple Input Modules (NI PXIe-4353). Data were sampled at 90 Hz with average values and standard deviations recorded in the output file at 1 Hz. The manufacturer-specified standard limit of error for the thermocouples is 2.2 °C or 0.75% (whichever value is greater). This error represents a standard uncertainty of the thermocouple itself and does not account for possible additional sources of uncertainty, such as the data acquisition systems and calibration drifts during use. The Neubrescope^® data acquisition system (model: NBX7020) was used to take Brillouin gain spectra from the distributed fiber optic sensors [33]. Using a pulse length of 0.2 ns, 2 cm spatial resolution measurements are obtainable, and the accuracies for strain and temperature measurements were reported to be 15 με and 0.75 °C, respectively, at an average count of 2¹⁴ [33]. In this study, the measurement distance and the spatial resolution were set at 50 m and 2 cm, respectively, meaning the Brillouin frequency shifts of two points spaced at no less than 2 cm could be distinguished over a 50 m fiber length. The scanning frequency ranged from 10.82 GHz to 11.50 GHz, which approximately corresponded to a target temperature range of 20 °C to 800 °C [28]. The acquisition time varied from 15 s to 25 s, depending on the scanning frequency range.

Type III Portland cement, Missouri river sands, and small aggregates (maximum grain size of 5 mm) at 640 kg m⁻³, 800 kg m⁻³, and 400 kg m⁻³, respectively, were used. The water-to-cement ratio was 0.45 by mass. To improve the flowability of self-consolidated concrete, a polycarboxylic acid high-range water reducer was used at a dosage of 1% by volume of the water content. The initial slump flow was measured to be between 280 mm and 290 mm, ensuring that no vibration was required and optical fibers in the specimens would not be disturbed during casting. After concrete casting, the beams were trowel-finished and covered with wet burlap pieces and a plastic sheet for 1 day. The beams were then demolded and air-cured for 36 days prior to testing. In the first 32 days, the beams were cured at a temperature of 25 ± 3 °C and a relative humidity of 40 ± 4%. In the last four days before testing, the beams were exposed to a less controllable environment (transport and indoor space) where the temperature and humidity were about 30 ± 5 °C and 30 ± 10%, respectively. The beams were weighed every three days. The mass loss over time was measured and plotted in figure 5. The thermal conductivity and the specific heat of the concrete were measured to be 1.8 W m⁻¹ · K⁻¹ and 1.7 kJ kg⁻¹ · K⁻¹, respectively.

Zoom In Zoom Out Reset image size

The four concrete beams were exposed to fire at increasing intensities under the control of burner's heat release rate (HRR). Figure 6 shows the test protocols of the four beams. The HRR was held constant at each target level. The durations corresponding to each sustained HRR value for the four beams are listed in table 1.

Zoom In Zoom Out Reset image size

A relatively longer duration was adopted at the first target level to more slowly heat the test beams, reduce the temperature gradient, and drive off some of the residual moisture near the surface. The intent was to delay spalling of the concrete, which tends to occur with a relatively high (>10% by weight) moisture content [34]. Beam 1 and Beam 2 were tested with the same protocol. They were pre-heated at a HRR of 25 kW for 45 min, and, then, heated at each elevated HRR for 10 min. Beam 3 was heated at 25 kW for 30 min, which was 15 min less than that of Beam 1 and Beam 2, in order to shorten the test duration. For Beam 1 to Beam 3, fire was extinguished when excessive spalling occurred at a HRR of 160 kW for various time durations. Beam 4 was first heated for 30 min at 15 kW, which was 10 kW less than that of the other three beams. The fire was extinguished for 3 min between 40 kW and 80 kW to allow visual observation of Beam 4. Then, fire was re-ignited, and the HRR was increased from 160 kW to 200 kW until excessive spalling occurred.

As the temperature in Beam 4 increased, the concrete cracked and spalled as shown in figure 7. Surface cracks were observed when the air temperature within the compartment reached about 300 °C. Cracking continued as the temperature increased and at 450 °C spalling occurred at the corners and edges of beam. Sudden fracture through the entire specimen happened shortly after the air temperature reached 800 °C to 1000 °C. Similar damage progressions were observed for each of the other beams. The mechanisms causing cracking and spalling of concrete at elevated temperatures are complicated and involve a series of physicochemical reactions and stresses induced by thermal gradients [35–38]. It is noted that fracture behavior is governed by the temperature of the beam, not the surrounding air. Moreover, the observed damage sequence was specific to our test member geometry, material properties, conditions at the time of testing (e.g. beam moisture content), and thermal loading protocols applied.

Zoom In Zoom Out Reset image size

Concrete cracking and spalling affect the heat transfer behavior of the beam as illustrated in figure 8. Before cracks appear, the temperature of internal concrete increases mainly due to thermal conduction. Due to the low thermal conductivity of concrete, the temperature of the concrete away from the surface increases slowly. Once cracks appear, they can be filled with hot air. The internal concrete is then subjected to thermal radiation and convection along the crack, in addition to thermal conduction through the concrete cover. Therefore, the temperatures in the vicinity of cracks can increase more rapidly. Since the spatially-distributed sensors have greater chances to cross the cracks than discrete thermocouple beads, they are a potential method to monitor cracking within concrete.

Zoom In Zoom Out Reset image size

When the beams were subjected to fire, the temperature within the concrete gradually increased. At each HRR, the spatially-distributed Brillouin frequencies along the length of fiber optic sensor were measured and converted to temperatures using the calibration curve in figure 2(b). Figures 9(a)–(d) show the temperature time histories in Beam 1 to Beam 4, respectively. The measurement results from the thermocouples and distributed fiber optic sensors were compared.

Zoom In Zoom Out Reset image size

As expected, temperatures were typically highest at the bottom, mid-span of the beam (TC2) and lowest at the center of the beam (TC1 and TC3). At a HRR of 25 kW and 40 kW for Beam 1 to Beam 3 and at 15 kW to 40 kW for Beam 4, temperatures within the concrete increased almost linearly with time and the maximum discrepancy between the measurements from the thermocouples and fiber optic sensors was 9%. The discrepancy was likely due to the slightly different positions of the sensors in the specimen. However, at a HRR of 80 kW for Beam 1 to Beam 4, temperatures increased at higher rates over time and the discrepancies between the measurements were increased up to 54%. As discussed in the next section, these larger discrepancies are believed to be due to the influence of concrete cracking or spalling on the heat transfer. Indeed, the spikes in the thermocouple measurements immediately prior to fire extinction were due to exposure of the thermocouple beads to heated air after spalling.

Figures 10(a)–(d) present temperature distributions in the four beams, measured from the distributed fiber optic sensors. The horizontal axis represents the distance along the distributed fiber optic sensor, starting from the pump end of the Neubrescope. The vertical axis represents temperature. The location of the distributed fiber optic sensor within the test beam was marked using capital letters from A to G (refer to figure 4). The length between A and G was embedded within concrete. The other portions of the fiber were exposed to air, and, in particular, the part of the fiber length within the compartment was subjected to heated air. As indicated in figure 10, air temperatures up to 600 °C were measured by the distributed fiber optic sensors. However, as the distance from the compartment increased, the temperature dropped to room temperature. While the air temperatures measured by the fiber optic sensors generally agree with the compartment thermocouple measurements (comparison between figures 6 and 10), a direct comparison is not made since the sensors were not co-located inside the compartment.

Zoom In Zoom Out Reset image size

The optical fibers between points A and G measured the temperature distributions within the concrete beams. Non-uniform temperature distributions are demonstrated in figure 10. Overall, temperatures between B and C and between E and F are higher than those between C and E. This is because fiber sections B-C and E-F were closer to the exterior surfaces of the beam than C-E. Furthermore, temperatures between B and C are higher than those between E and F. This is because the bottom surface was closer to fire and had a higher temperature than the top surface, and B-C and E-F were respectively close to the bottom and top surfaces of the beams.

When the air temperature was below 400 °C corresponding to a HRR of below 40 kW, temperatures over B-C, C-D, D-E, and E-F sections were approximately constant in the four beams. However, when the air temperature exceeded 400 °C, peaks appeared as indicated by the dashed circles in figure 10. For instance, as shown in figure 10(a), the temperature at the middle of C-D was higher than that at C or D at a HRR of 80 kW. Additionally, the thermal gradients of C-D and D-E were approximately the same and symmetrical to D. This behavior is believed to be due to cracking in the concrete.

Concrete cracking and spalling may break the distributed fiber optic sensor in two mechanisms, as respectively illustrated in figures 11(a) and (b): opening (Mode I) and sliding (Mode II). Crack opening in concrete influenced the heat transfer behavior but had no effect on sensor function as evidenced from the test results since there was no bond between the optical fiber and its protective sheath. However, sliding can be detrimental to the distributed sensor by bending the optical fiber into an acute angle, which results in a significant signal loss and thus a reduction of signal-to-noise ratio. Fibers in a section of spalling concrete were sheared off as illustrated in figure 11(c). Once broken, the distributed sensor with PPP-BOTDA measurements was no longer functional. However, BOTDR measurements can still be taken from one end of the optical fiber to determine the location of the spalling [21].

Zoom In Zoom Out Reset image size