Introduction

The removal of formwork during concrete construction should be done only when concrete has gained sufficient strength to withstand its self-weight. IS 456:20001 suggests ensuring that a concrete member should attain a minimum of 65–90% of its 28th -day strength before removal of the formwork. The minimum attained strength depends on the type of member. The rate of hardening of concrete or the development of concrete strength depends on the ambient temperature and moisture condition that as a result affects the formwork removal time. The early removal of formwork results in excessive deformations, cracking, reinforcement slippage, creep, or even failure/collapse of the structure. It increases the chances of failure especially while progressive construction, putting human life, time, and material at risk. The collapse of an under-construction 26-story Skyline Plaza building in Virginia is an example of such failure due to premature removal of formwork2. In the current construction practices, codal provisions are used for deciding the appropriate formwork removal time for ordinary Portland cement with Supplementary Cementing Materials (SCMs) and rapid hardening cement. Hitherto, the standard cubes/ cylinders prepared from the same mix are tested under compression to estimate the strength of concrete before formwork removal. IS 516:19593 suggests curing the cubes in the moist air of relative humidity 90% and at a temperature of 27 ± 2 °C for the first 24 h and then be submerged in water maintained at a temperature of 27 ± 2 °C. These curing conditions are not always possible to be maintained in the field while curing. Apart from this, the use of new types of cement and the addition of additives has made it complex to estimate the correct ‘in-situ’ strength gain of concrete at a specific time of curing. This prolongs the removal of formwork to be on a safer side and results in delays in the construction. The current demand for increased construction speed puts tremendous pressure on the early removal of formwork. Hence, it is important to determine the in-situ strength of concrete. Monitoring the early strength gain in concrete can be very useful in determining the time for formwork removal and applying further loading4. Generally, non-destructive techniques are preferred over destructive techniques, as these do not induce any damage to the structure. Ultrasonic pulse velocity (UPV) and Schmidt hammer (SH) tests are widely used to determine the strength gain in concrete5. The results from these tests are used for developing statistical correlations for estimating the in-situ strength6,7. Since these techniques were developed to indicate the quality of concrete, they can only estimate the in-situ strength in a probabilistic approach. Also, they are time-consuming and challenging to perform in inaccessible locations. The conventional concrete mixes are also supplemented with additives like nano-particle8, and synthetic fibers9 to achieve ultra-high strength, and sometimes with incinerated agricultural waste10, to provide sustainable waste management solutions. Thus, there is a need for a better monitoring technique that can indicate the strength gain of these new types of concrete and while considering all the above-discussed issues. The estimation of concrete strength based upon the electro-mechanical impedance technique using the embedded Resin Jacketed Piezo (RJP) sensors is a new technique that is a possible solution for all the above-mentioned problems. The RJP sensors are Lead Zirconate Titanate (PZT) based on thick circular piezo patches coated with epoxy resin, which can sustain the rough conditions while casting.

The piezo-based sensor configurations have been successfully demonstrated in providing condition monitoring solutions to various engineering problems. The new applications are not limited to conventional civil engineering issues like corrosion, damage detection, fatigue monitoring, etc. The Electro-Mechanical Impedance (EMI) technique is a proven method to detect structural changes in the vicinity of the sensors11,12,13,14. The transducers used for the EMI technique are patches made of piezoelectric materials like PZT, which can act as both sensors and actuators simultaneously and eventually as mechatronic impedance transducers due to their direct and converse effect capabilities. The EMI technique can be used for long term monitoring effectively by using self-diagnostic measures, which can compensate for the sensor degradation over a period of time15. In the EMI technique, a PZT patch is instrumented on a structure using high-strength adhesive and subjected to an alternating voltage excitation from an impedance analyzer/ LCR meter, sweeping through a particular frequency range, generally a subset of 30–400 kHz. At a given frequency, the patch actuates the structure and the structural response is, in turn, sensed and measured in terms of the electro-mechanical admittance of the patch, consisting of real and imaginary components, the conductance, and the susceptance, respectively. In this manner, frequency plots termed conductance (G) and susceptance (B) signatures, are generated across the specified frequency range. A plot of ‘G’ over a sufficiently wide band of frequency serves as a diagnostic signature of the structure and is called the conductance signature. Bhalla and Soh11 advocated using both G and B to extract components of impedance (Z) for improved structural identification diagnosis.

The development of new smart sensors has replaced many arduous practices in many aspects of the engineering world. Many techniques have been used in the literature for strength gain monitoring including the UPV, Schmidt hammer, and surface-bonded piezo sensors. Researchers have developed PZT patches of different sizes and shapes to optimize the utility of piezo based sensors in various applications. The sensitivity of these sensors towards surrounding stiffness and strength modulus makes them effective in monitoring concrete hydration/ strength. In 2009, Song et al.16, presented the methodology to prepare smart aggregates (SA) using PZT patches to be embedded in concrete for hydration monitoring and damage detection. They proposed an active sensing system involving two SA’s acting as an actuator and sensor. They also reviewed the health monitoring of concrete structures using Piezo based sensors. In a recent review, Lim et al.17 compiled almost all the studies related to the hydration monitoring/ strength gain monitoring of concrete using thin PZT patches in different configurations in a review paper. They have addressed almost all the important issues related to the monitoring of concrete hydration using PZT based sensor EMI technique. They have also emphasized the studies including field applications to identify the influencing factors. Recently Lu et al.18, have assessed the strength gain and dynamic modulus of the cementitious materials using a prism-based configuration. They proposed a miniature prism bonded PZT patch as non-destructive testing to monitor the strength gain of concrete instead of using destructive tests. Table 1 displays the experimental details and findings of hydration monitoring using PZT patches by various researchers. Since the last decade, the feasibility of monitoring the strength gain using piezo sensors via electromechanical impedance (EMI) technique has already been established, however, a calibration relation to estimating the actual in-situ concrete strength is not provided by any of these studies. Also, these feasibility studies are based on a limited (max up to 42) number of samples and statistical parameters that largely depend upon the frequency range considered for the study. The selection of a particular frequency range depending upon the researcher’s experience could end up in dubious results when applied in the field. Apart from the experimental studies, Lu et al.19 developed a new EMI based theoretical model for monitoring the strength gain in cementitious materials. Some researchers have also explored the embedded PZT configuration for concrete strength monitoring, however, their suggested embedded configuration for the PZT patches is either complex to fabricate or they have used a thin square PZT patch which does not produce a sharp first resonant peak in the EMI signature20,21,22,23.

The present study aims at investigating the effectiveness of the embedded sensors in monitoring the gain in the concrete strength based on the shift in the first resonant peak of conductance signature in a more ‘conclusive’ and ‘consistent’ manner. Even though many researchers have proposed and developed several piezo-based configurations (summarized in Table 1) for hydration monitoring, however, they are either complex or not easy to implement in real-life on-site field conditions. There are several concerns with these listed piezo-based monitoring techniques, such as the position of the PZT patch, sensor protection, temperature, and cement hydration effects, and epoxy cover thickness. The current study attempts to address the above shortfalls related to the existing piezo based configurations by developing a unique solution based on a basic piezo-based configuration as described by Negi et al.24. This technique is more practical, easy to fabricate and execute in the site conditions. Another researchers group25,26 also used a similar piezo configuration and successfully monitored concrete hydration by studying its natural frequencies and the respective conductance values. However, this was again a feasibility study that only established the fundamental concept. Therefore, the present study extends the concept of estimating the concrete compressive strength in real-life field conditions using embedded RJP sensors by utilizing mathematical models developed in this study based on a sufficiently large number of tests. A total of 180 samples in the laboratory and three sets of readings in an under-construction bridge were tested in the present study that adds to its novelty and makes it a thorough rigorous investigation that duly considers the on-site field conditions including the size factor, temperature, and humidity. A correlation between the changes of the EM parameters of the embedded piezo patch and the changes in the hydration or strength gain of the concrete has been established in this study. Further, the effect of other environmental parameters like temperature and humidity on the developed correlations has also been to be studied in the present research.

Though numerous researchers have presented feasibility studies of employing piezo-based sensors to monitor concrete strength increase, these studies have yet to be field validated. A field testing of any proposed methodology brings several issues to the notice, which aids in its further development. The end goal of any research is to implement it on the ground for fully extorting its benefit to the end users. This can be quantified by designating particular Technology Readiness Level (TRL) as defined by NASA to discuss the maturity of developing technology. It is emphasized here that another very important feature of this study which adds to its novelty is that it aims to enhance the technology readiness level (TRL) of concrete hydration monitoring technique from TRL 3 (establishing proof of concept) to TRL 9 (proving the system in operational environment). The proposed technique is recommended to be used in addition to the existing practices at the construction site for effective concrete strength estimation which can help in better decision making during crucial stages of construction e.g. removal of formwork in the casting yards to increase the construction rate. Also, the embedded sensors can be crucial in determining the in-situ strength of the structure at a later stage. It is recommended to use in-situ strength for designing retrofitting measures for an existing structure27.

Table 1 Literature details on hydration monitoring of concrete using PZT patches.
Full size table

Concrete strength estimation using RJP sensors

The process of fabrication of RJP sensors in laboratory conditions is discussed below. The strength monitoring of concrete using the embedded RJP sensors is initiated from the time the sensors were embedded inside the concrete just after casting and continued till 28 days after casting. The correlation models have been developed based on a sufficiently large number of samples to estimate the concrete strength using RJP sensors via the EMI technique. The developed models have been extended to the field conditions as well.

Fabrication of RJP sensors

In the present study, thick circular (10 mm dia, 1 mm thick) PZT patches, are protected with a single thin layer of two-part epoxy (Araldite standard epoxy adhesive) to protect the piezo and soldered terminals against the adverse effects of water, and environmental conditions during the construction stages are named as an RJP sensor24. The PZT patches used in this study were confirmed to the properties of PZT-5 A, with a d33 constant of 390 × 10− 12 m/V. This configuration provides a sharp first resonant frequency peak, and the shifting of this peak is measured to estimate the change in concrete strength. A typical view of an RJP sensor is shown in Fig. 1 with the entire experimental setup. A total of 120 RJP sensors were fabricated identically in laboratory conditions. The EMI signatures of the RJP sensors were acquired during three stages of their preparation, namely, (a) raw PZT stage when the PZT was not soldered, (b) soldered PZT patch, and (c) RJP (unbonded) when a protective epoxy layer was added on the soldered PZT. The RJPs were allowed to rest for 24 h, for the curing of the epoxy. The gravity ensures the smooth flow of epoxy over the PZT patch forming a uniform protective layer of 100 μm. The variation of peak conductance and peak frequency in these three stages of RJP sensor fabrication for 120 PZT patches is shown in Fig. 2. It is observed that the peak conductance value of a free PZT patch at the first resonance frequency decreases when its terminals were soldered and it reduces further after the addition of the protective epoxy layer. However, there is a negligible shift in the peak frequency values after soldering the terminals but, the value reduces on adding the epoxy layer as it increases the mass of the system. These results were used for selecting the RJP sensors with similar peak conductance and peak frequency values to ensure consistency among the functioning of the sensors while experimentation. Standard deviation (SD) for each state is calculated for both, the peak conductance and peak frequency. The lower standard deviation in the final state (SD = 0.003) for the peak conductance values proves its better consistency as compared to the peak frequency values. Such high similarity in the peak frequency and simultaneous conductance values aids in understanding the correlation between the shifts and the features to be monitored, which in this case is concrete strength.

Fig. 1
figure 1

Experimental setup for acquiring EMI signature of the RJP embedded inside the concrete cubes immediately after casting.

Full size image
Fig. 2
figure 2

Variation of (a) peak conductance and (b) peak frequency of PZT patch in the raw, soldered, and unbonded RJP condition.

Full size image

Hydration monitoring using RJP sensors

The characterization of the raw material used in the study including cement, coarse aggregates, fine aggregates, and admixture for preparing the concrete mix was done using standard tests. Five cubes (each) conforming to M40 grade with and without accelerator were cast. Mix proportions of 1:2.83:1.66 and 1:2.77:1.71 (Cement: Coarse Aggregates: Fine Aggregates) were used for concrete mixes with and without the accelerator, respectively. The Water Cement ratio used was 0.4 for both the set of mixes. The polycarboxylate ether (PCE) based super-plasticizer and accelerator content produced from41 was kept constant as 1% and 2% by weight of cement, respectively. During casting, five RJP sensors (RJP-1 to RJP-5) were embedded in the five cubes of M40 without an accelerator and five RJP sensors (RJP-6 to RJP-10) each were embedded in the five cubes of M40 with the accelerator at the center of the cube. The cubes were covered with plastic wraps to avoid any water loss.

After the casting, the first set of initial signatures was acquired hourly using the impedance analyzer, E4980A42 in the frequency range of 20 kHz to 450 kHz. The entire acquisition setup is shown in Fig. 1. After 24 h, the cubes were transferred to a controlled humidity chamber at 90% Relative Humidity (RH) and 25 °C temperature, instead of submerging them in the water tank (Fig. 1). This type of curing ensured conditions similar to a field where complete submergence of an RC member is not always possible. The standard deviation (SD) shown in Fig. 2 in the peak conductance and peak resonance values of the RJPs indicates the uniformity in the resonance properties of the prepared RJP sensors. The conductance signature of a typical RJP sensor, embedded inside the cube from Day 0 (immediately after casting) to Day 28 of casting is shown in Fig. 3. The first resonant peak shifted downward and rightward with an increase in the number of days after casting. A prominent reduction in the overall conductance signature was observed during the initial hours of hydration (Day 0). This happened due to the presence of a considerable amount of water during the initial hours34. Later, the reduction in the peak frequency value becomes gradual after a day of casting. This is attributed to the combined effect of hydration (occurring during the initial four hours) and increase in the Young’s Modulus (thus strength) of the concrete with time. This also increases the peak frequency value with the increasing age of concrete.

Both of the concrete batches (with Acc and without Acc), showed an initial dip for the first two hours of monitoring, in the peak frequency value (Fig. 4). In the first three hours of hydration, the corresponding increase in mechanical impedance (resistance to motion) of the surrounding medium to the motion of the PZT patch is related to a decrease in amplitude and a slight decrease in the frequency of the first resonant mode42. Later, the peak resonant frequency started increasing until Day 4 and then becomes constant. Also, in all the cases, it was observed that the peak conductance value reduced significantly during the first day of hydration in comparison to the latter days (Fig. 5). The inset of Fig. 5 showed a further gradual reduction in the peak conductance values until Day 4 followed by negligible change till Day 28. The variation in resonant frequency and peak conductance value with hydration time up to Day 4, indicates a strong correlation with strength gain in concrete.

Fig. 3
figure 3

Conductance of RJP in unbounded and bonded condition from Day 0 (immediately after casting) to Day 28 of casting.

Full size image
Fig. 4
figure 4

Peak frequency for different RJP sensors embedded in concrete cubes from Day 0 to Day 28 of casting (a) without accelerator and (b) with accelerator.

Full size image
Fig. 5
figure 5

Peak conductance for different RJP sensors embedded in concrete cubes from Day 0 to Day 28 of casting (a) without accelerator and (b) with accelerator.

Full size image

Early strength detection using RJP sensors

Considering the previous observations of negligible changes in both peak frequency and peak conductance values after Day 4, it was decided to restrict the monitoring till the 7th day after casting in further experimentation. To generate correlation models between RJP sensor parameters and the early concrete strength, different concrete grades with and without accelerators were cast. Ten different concrete mixes as listed in Table 2, were considered to achieve sufficiently scattered concrete strength values for a better correlation. The mix design procedure was based on IS: 10,26243 while keeping in view the durability requirements as per IRC: 11244. The superplasticizer and accelerator content was kept constant in all the mixes as 1% and 2% by weight of cement, respectively. In total 18 cubes were cast for each concrete mix. For each concrete mix, only three cubes were instrumented with the RJP sensors. The non-instrumented cubes were compressed in a set of three in a Universal Testing Machine (UTM), UH 2000 kNIR at a loading rate of approximately 140 kg/sq cm/min3 to determine the compressive strength on the 18 h, 1 day, 2 days, 3days and, 4 days after casting, whereas the EMI signatures were acquired simultaneously from the instrumented cubes. On the 7th day, the three instrumented cubes were compressed after recording the signatures of the embedded RJP sensors.

Table 2 Proportion of concrete mixes.
Full size table

The three EMI parameters, RJP resonant peak widening, the peak conductance shift, and the peak frequency shift were calculated from the acquired signatures for all the mixes at different days after casting. Figure 6 shows the average variation in RJP resonant peak widening of the first resonance peak derived from the EMI signatures of RJP sensors. The RJP resonant peak widening (\(\xi\)) was derived based on the following half-power bandwidth formula46.

$$\:=\frac{{f}_{2}-{f}_{1}}{2{f}_{\text{n}}}$$
(1)

where fn is the peak frequency value in the conductance signature of the RJP sensor, \(\:{f}_{1}\) and \(\:{f}_{2}\) are the frequencies corresponding to an amplitude, \(\:Peak\:conductance\:value\:at\:{f}_{\text{n}}/\sqrt{2}\). From Fig. 6, it is observed that for all the concrete mixes (with and without accelerator); the peak widening initially rises rapidly until the 18th hour and then achieves a downward slope till Day 4 and lastly becomes constant. The higher peak widening till Day 1 is due to the presence of considerable free water in the cubes, which gets consumed in the hydration reactions/evaporated later with the increasing age of the cubes.

The peak conductance frequency shift corresponds to the percentage difference between the peak conductance frequency values at different ages of concrete normalized with respect to the values measured just after casting (Zeroth hour). Hence, as the peak conductance shift (%) is relative, there will be no value for Day 0. From Fig. 7, it is observed that the peak conductance shift is slightly increasing with the age of the cubes. The peak conductance shifts are rather closely spaced for all the concrete mixes with slightly higher values for the concrete mixes with the accelerator. The lower values of peak conductance shift for M35 are attributed to inconsistency caused due to human error while casting this concrete mix. Figure 8 shows that the peak frequency shift values also exhibit an increasing trend with the increasing age of concrete. In contrast to the peak conductance shift, the peak frequency shift is sparsely spaced for all the concrete mixes. Also, the latter’s values are slightly higher for the concrete mix with the accelerator. Due to the accelerator, the concrete attained a rapid strength up to Day3-4. This phenomenon is also reflected in the higher changes in the peak frequency shift in concrete mixes with accelerators. However, the ultimate shift is similar for both cases as the accelerator does not improve the final strength gain of the concrete.

Fig. 6
figure 6

RJP resonant peak widening for different concrete mixes with the age of concrete (a) No Accelerator Case and (b) With Accelerator Case.

Full size image
Fig. 7
figure 7

Peak conductance shift for different concrete mixes with the age of concrete (a) No Accelerator Case and (b) With Accelerator Case.

Full size image
Fig. 8
figure 8

Peak frequency shift for different concrete mixes with the age of concrete (a) No Accelerator Case and (b) With Accelerator Case.

Full size image

To find the compressive strength of a Mix after specific days of curing, three non-instrumented cubes, cast from the same batch were tested under the UTM. The average strength achieved by both with and without accelerator batches is shown in Fig. 9 with their designated days after casting. The strength of the concrete mixes with the accelerator was slightly higher as compared to those without the accelerator, as the accelerator increases the rate of strength gain at the initial stages but does not enhance the 28-day strength. The variation of the average concrete compressive strength along with the peak conductance shift and peak frequency shift for different concrete mixes is shown in Fig. 10. It can be observed, the slope of plots for the concrete mixes with accelerator in both the cases of peak frequency shift and peak conductance shift, is slightly higher than in the case of with out accelerator. Also, the peak conductance shift (between 80 and 95%) is higher for both accelerator and with out accelerator case than peak frequency shift (between 20 and 40%), indicating it a better strength indicator.

Fig. 9
figure 9

Strength of concrete cubes (a) without accelerator and (b) with accelerator for different concrete mixes along with the age of concrete.

Full size image

Development of RJP based mathematical models

This section presents the development of the mathematical models between RJP parameters and achieved concrete compressive strength based upon 180 concrete cubes. The average values of the concrete compressive strength, peak conductance shift, peak frequency shift derived from the conductance signature of the RJP sensors embedded in the ten different concrete mixes are plotted in Fig. 11. Three correlation models namely, Model-A, Model-B, Model-C have been found between the compressive strength and corresponding peak conductance shift, peak frequency shift, and their combined effect respectively.

$$\:\text{M}\text{o}\text{d}\text{e}\text{l}-\text{A}\:{S}_{\text{t}}=57.64{G}_{\text{S}}^{5.298}$$
(2)
$$\:\text{M}\text{o}\text{d}\text{e}\text{l}-\text{B}{S}_{\text{t}}=139.6\:{F}_{\text{S}}^{1.176}$$
(3)
$$\:\text{M}\text{o}\text{d}\text{e}\text{l}-\text{C}{S}_{\text{t}}=70{F}_{\text{S}}^{0.018}{G}_{\text{S}}^{7.265}$$
(4)

where, ‘St’ denotes the compressive strength (N/mm²) and ‘Gs’ and ‘Fs’ denote the peak conductance shift (infraction) and the peak frequency shift (infraction) derived from the conductance signature of the embedded RJP sensor, respectively. For example, an 80% percentage shift in peak frequency value will be written as Fs = 0.8 in the models for calculating the estimated concrete strength. The highest value of the coefficient of determination (R2)29 for Model-C proved it to be a better model than the other two for concrete strength determination using RJP sensors.

Fig. 10
figure 10

Strength Vs peak conductance shift for different concrete mixes (a) without accelerator and (b) with accelerator and strength vs. peak frequency shift for different concrete mixes (c) without accelerator and (d) with accelerator.

Full size image
Fig. 11
figure 11

Proposed model for estimating concrete strength based on (a) peak conductance shift, (b) peak frequency shift, and (c) combined effect of peak conductance and peak frequency shift in a 3-D plot.

Full size image

Validation of RJP based on mathematical models

The above-developed models were validated in the laboratory condition for two sets of cubes (Set-1 and Set-2) with different concrete grades (18 cubes for each set). Mix proportions for the concrete grade M50 (without accelerator) and M60 (with accelerator) corresponding to Set-1 and Set-2, respectively, considered for validation are listed in Table 2. Compressive strength after 18 h, 1, 2, 3, 4, and 7 days of casting was determined for three cubes (each) of Set-1 and Set-2. The correlation between the measured and the estimated compressive strength using Model-A to Model-C is shown in Fig. 12. It can be seen from the figure that the Models predicted the concrete strength closely when the compressive strength of concrete was more than 20 N/mm2. The free water present in the initial stage affects the conductance signature of the RJP sensors47. The lamb waves generated by the vibrating piezo patches interact with the liquid-solid boundaries resulting in a fluctuating conductance signature away from the peak resonant frequency (refer Day 0, 0th hour to Day 0, 4th hour signature in Fig. 3). Also, the high-frequency lamb waves generated by the vibrating sensor get damped due to the presence of water and reduce the area of sensing of the sensor47. As soon as the free water reduces significantly after Day 1, the confidence level in the predicted strength improves. Hence, after the 1st day of casting, the accuracy in the predicted values of the models was improved for both sets. It is thus observed for both the sets, that all the three models are effectively predicting the compressive strength when it is higher than 20 N/mm2.

Figure 13 shows the plot of error percentage in the strength estimation based upon the proposed RJP sensor-based models along with the increasing age of concrete. It is observed for both the sets, that all the three RJP based developed models are very well predicting the compressive strength of concrete with an error less than 10% after Day 2 of casting. However, these models proved unsuitable for predicting the compressive strength for Day 1 in both sets, as the error is unusually high. This is due to the reason already stated above that the EMI readings of the RJP sensors get prominently affected due to the presence of free water around them. The error values for Set-1 (M50) in comparison to Set-2 (M60) are higher. Overall, based upon the error percentage shown in Fig. 13, Model-C proves to be the best among all the three RJP based developed models for compressive strength prediction after Day-2.

Fig. 12
figure 12

Correlation between the measured and predicted strength using piezo-based models for the (a) Set-1 and (b) Set-2.

Full size image

Ambient temperature and humidity corrections for RJP based mathematical models

The models proposed above have been developed at a constant temperature of 25 °C and relative humidity of 90%. However, in actual site conditions, these conditions vary a lot. Hence, there are several factors, including size factor, temperature, and humidity which may affect the efficacy of the proposed RJP sensors-based model developed above. In this section, the effect of temperature and humidity is discussed to formulate corrections for the previously developed models. The size factor will be studied in future research as an extension of this study.

Fig. 13
figure 13

Error in the prediction of concrete strength using RJP based models for (a) Set-1 and (b) Set-2.

Full size image

The EMI signatures of the embedded RJP sensors were captured while exposing three concrete cubes (M40, without accelerator) to a range of temperature and humidity in the humidity chamber shown in Fig. 1. The temperature was varied from 10 °C to 60 °C and the RH was varied from 40 to 90%. The peak conductance and peak frequency values were recorded from the conductance signature of the RJP sensors and their variation with increasing temperature and humidity is shown in Fig. 14. The conductance signatures showed an upward and leftward shift with an increase in temperature. Thus, the peak conductance value of the RJP sensor increases, and the peak frequency value reduces as shown in Fig. 14. The chamber humidity showed a negligible effect on both the peak conductance and peak frequency values. Table 3 provides the corrections to be incorporated in the peak conductance shift and peak frequency shift for different temperatures and relative humidity. The average temperature and humidity for 7 days at the site were found to be 25 °C and 90% (due to regular curing), respectively using a handheld probe. Therefore, in the present case, the correction values for the peak conductance shift and peak frequency shift corresponding to the temperature and humidity condition of 25 °C and 90%, respectively are 0.0 as observed from Table 3. The peak conductance and peak frequency shift for the in-between temperature and humidity values as mentioned in Table 3 shall be calculated by bilinear interpolation. These correction values shall be added/subtracted (as per the sign) in the peak conductance shift and peak frequency shift fraction to incorporate the correction for temperature and humidity in the models [Eq. (3) to (5)] proposed in the previous section. The higher correction values for the peak conductance shift indicate its higher sensitivity towards the change in temperature and humidity in comparison to the peak frequency shift.

Table 3 Correction for temperature and humidity in the peak conductance and peak frequency shift.
Full size table
Fig. 14
figure 14

Effect of temperature and humidity on the (a) peak conductance shift and (b) peak frequency shift.

Full size image

Extention of RJP based models to real-life field condition

The RJP based models were extended to the real-life field condition attempting the validation of the results. Three RJP sensors (S1, S2, and S3) were embedded in the pier (width = 500 mm) of an under-construction bridge in Faridabad, India as shown in Fig. 15. While casting, 18 separate cubes were cast from the same concrete for determining the compressive strength of the concrete. A set of three cubes (out of 18 cubes in total) were tested for concrete compressive strength in the UTM after 18 h, 1, 2, 3, 5, and 7 days of casting. The EMI signature of the three RJP sensors (embedded in three different cubes) was acquired just after the concreting, using the impedance analyzer, (Agilent LCR meter E4980A)34 in the frequency range of 20 kHz to 450 kHz. The LCR meter was used on battery power using an uninterruptible power supply. This also ensured very less noise due to surrounding electrical and magnetic interference, while acquiring signatures. The EMI signature of the RJP sensors S1, S2, and S3 are shown in Fig. 16. During the concreting at the site, intermediate connection issues were leading to overload values in the impedance analyzer readings which got resolved. At an overload condition, the LCR meter doesn’t show any value which is beyond its measuring range. Therefore, the EMI signature of the RJP sensors just after embedding is not included in the plots. The overall behavior of the EMI signature for the RJP sensors S1 to S3 is similar to that observed in Fig. 3 plotted for the RJP sensors embedded in the cubes in the laboratory. It indicated a downward and rightward shift in the first resonant peak of the signature with an increase in the age of concrete depicting an increase in the Young’s Modulus, strength, and stiffness of the concrete with time.

Fig. 15
figure 15

Experimental setup at the site, pier of an under-constructed bridge at Faridabad, India.

Full size image
Fig. 16
figure 16

Conductance signature of RJP sensors (a) S1 (b) S2 and (c) S3.

Full size image

The shift is very prominent during the initial hours of the hydration (Day 0) and afterwords becomes gradual. From the EMI signatures for the initial 4 h, it can be noticed that the EMI signature of sensor S2 corresponds more closely to that observed in the laboratory when compared with the sensors S1 and S3. This can be attributed to the location of the sensors, as the latter was embedded close to the exposed surface leading to loss of water due to evaporation in contrast to the sensor S2 which was located in the mid of the pier where the water content during the initial 4 h is considerably greater. The peak conductance shift and the peak frequency shift were determined as mentioned above and plotted against the age of the concrete in Figs. 17 and 18, respectively. The average values of the peak conductance shift displayed in Fig. 17 increase rapidly till the 2nd day after casting following a gradual increase subsequently. In contrast, the average values of the peak frequency shift plotted in Fig. 18 increase progressively through the 3rd day after casting following a negligible increase subsequently. At a certain age of the concrete, the values of the peak frequency shift for three RJPs are more scarcely spaced as compared to the peak conductance shift values exhibiting higher repeatability in the latter. It is also noticed that the values of both the peak conductance shift and peak frequency shift for the RJP sensor S1 and S3 are near to each other, unlike RJP sensor S2. This is again related to the position of the three embedded RJPs as stated previously. From Fig. 15, it can be seen that the RJP S1 and S3 were near the exposed surface of the pier consequently more affected by the ambient conditions as compared to S2. Figure 19 depicts the variation in average concrete compressive strength, along with peak conductance shift and peak frequency shift, for different concrete mixes. Figure 19(a) shows that the peak conductance shift plots have a greater slope than the peak frequency shift plots, indicating that they are more sensitive to changes in concrete compressive strength. This is also consistent with the laboratory findings.

Fig. 17
figure 17

Peak conductance shift with respect to the age of concrete for different RJP sensors embedded at the site.

Full size image
Fig. 18
figure 18

Peak frequency shift with respect to the age of concrete for different RJP sensors embedded at the site.

Full size image
Fig. 19
figure 19

(a) Strength Vs peak conductance shift and (b) Strength Vs peak frequency shift for different RJP sensors embedded at the site.

Full size image

The in-situ concrete strength was predicted using average values of the three RJP sensors for the peak conductance shift and the peak frequency shift in the proposed models given by Eq. (3) to (5). The in-situ concrete strength predicted using the proposed models by the RJP sensors S1 to S3 with respect to the cube strength measured in the laboratory after 18 h, 1, 2, 3, 5, and 7 days of casting is shown in Fig. 20. It is observed that the in-situ concrete strength predicted using all three proposed models by the RJP sensors S1 and S3 are very close to each other owing to their similar location (near the exposed surface). The strength predicted by RJP sensor S2 is the giving the closest estimate with respect to cube strength when compared to RJP sensors S1 and S3. This is expected as the latter is exposed to the hotter environment as compared to the RJP sensor S2 leading to comparatively lower strength as compared to that predicted by the RJP sensors S1 and S3 as depicted by Model A and Model C.

Fig. 20
figure 20

Correlation between the cube strength and predicted strength using piezo-based models for the RJP sensors S1, S2 and S3.

Full size image

Learnings from field study

For the first time, the EMI method has been performed in the real-life field condition for the concrete strength prediction. Several learnings both from the technical outcomes as well as from the execution point of view are listed below:

  1. 1.

    This study clearly shows that there is a clear difference in the in-situ concrete strength achieved in the outer region of the pier in comparison to the inner region. However, in the field practices, the entire strength of the concrete member is considered to be analogous to the strength estimated from the lab cubes test. It is recommended to embed the sensors away from the exposed surfaces of the structure to avoid the effect of water loss due to evaporation. That is the reason for better strength prediction by the sensor (S2) which was embedded in the mid of the pier.

  2. 2.

    During the concreting at the site, intermediate connection issues were leading to overload values in the impedance analyser readings which got resolved later after two hours of casting. At an overload condition, the LCR meter doesn’t show any value which is beyond its measuring range. This was happening due to wet site conditions during the initial monitoring. Therefore, it is recommended that a robust impedance analyzer shall be used which can withstand the harsh site conditions including dust, wet environment, and high temperature. A hand battery-operated equipment shall be preferred.

  3. 3.

    During the initial few hours of monitoring, many a time electric shocks were experienced due to wet site conditions. Therefore, proper electric gloves are recommended from a safety point of view.

  4. 4.

    The corrections of temperature and humidity can be further improved by embedding sensor probes along with the RJP sensors. In the future, the RJP, temperature, and humidity probes can be integrated as a single unit, giving a further improved model for determining the actual in situ strength of concrete using the EMI approach.

Conclusions

The focus of this research was to use RJP sensors embedded in concrete to measure in-situ concrete strength. The RJP sensors were simple to fabricate and use during the installation process. The sensors withstood the rigorous casting conditions as they were thick and covered by a resin coating. The conductance signatures exhibited a downward and rightward shift with an increase in the age of the concrete. This is attributed to the combined effect of hydration (occurring during initial hours) and increase in the Modulus (thus strength) of the concrete with time. The same reason caused the increase in the peak frequency (except during the initial two hours) with an increase in the number of days after casting. For all the RJP sensors, the peak conductance values reduced significantly till Day 1, followed by gradual reduction till Day 4, and finally negligible change till Day 28. The peak frequency values were also varying correspondingly but with an increasing trend. This establishes the feasibility of monitoring concrete strength using RJP sensors.

For the early strength detection, ten different concrete mixes (18 cubes for each mix with three of them instrumented with RJP sensors) were cast. The resonant peak widening in the signatures was found to be increasing till the 18th hour after the casting; reduced for Day 1 and then became almost constant with the increasing number of days till the 7th day after casting. Based upon the average values of the compressive strength of 180 concrete cubes (ten concrete mixes) and corresponding peak conductance shift, peak frequency shift, and their combined effect, three correlation models were developed for concrete strength estimation which were successfully validated in the laboratory. After the 1st day of casting, improved accuracy (> 90%) in the prediction models was witnessed. Model C (with minimum error) outperformed the other two models for the estimation of concrete compressive strength. The corrections for the temperature and humidity have also been proposed. The EMI signatures were found more sensitive to the change in temperature as compared to the humidity.

Further, the developed RJP based models were extended for the field conditions by embedding the RJP sensors in the pier of an under-construction bridge. The concrete strength was estimated using three embedded RJP sensors based on the proposed mathematical models. Based upon the results of the three RJP sensors, it is recommended to embed the sensors away from the exposed surfaces of the structure to avoid the effect of water loss due to evaporation. This study clearly showed that there is a clear difference in the in-situ concrete strength achieved in the outer region of the pier in comparison to the inner region. However, in the field practices, the entire strength of the concrete member is considered to be analogous to the strength estimated from the lab cubes test, which is therefore inappropriate. This establishes the superiority of the proposed methodology over the traditional techniques and recommends its use for improved decision making at the site.