1. Introduction
As a type of the most severe geological disasters, landslides can cause economic losses for billions of dollars around the world annually. Roads, bridges, oil-gas pipelines, and other infrastructures in mountain regions, as well as the people’s lives and property, are seriously threatened by landslides [
1].
Establishing the landslide monitoring and forecasting system (LMFS) is a feasible way to degrade the damage. Based on the manual methods, the early LMFSs [
2,
3] are built by observing the variation of the displacement of ground surface, the groundwater level, the plants and the other parameters in the slope area. With the development of science and technology, some mechanical instruments such as theodolites, inclinometers, and level gauges are used in the LMFSs [
4,
5]. However, due to the limited accuracy of the traditional measuring instruments, the early LMFSs can hardly satisfy the increasing engineering requirements.
Recently, by using the modern measuring instruments and techniques, several significant achievements in slope engineering fields have been obtained. Based on GPS and geodetic techniques, Puglisi et al. [
6] developed a remote system of ground deformation monitoring, which can real-timely measure the slope displacement and transmit signals wirelessly. Zhang et al. [
7] established a real-time remote system to monitor the landslides around the highway in mountain regions by utilizing the general packet radio service (GPRS) net of China Telecom. Aimed at the slopes of mountain highways, a remotely controlled system was built by Wu et al. [
8] to monitor and forecast the disasters by using GPRS, the trigger displacement meter, the grid pluviometer, and the other advanced techniques. Interferometry synthetic aperture radar (InSAR) was used by Perski et al. [
9] to measure the terrain deformation near the Wieliczka Salt Mine in Poland. Jia et al. [
10] proposed a static and dynamic factors-coupled forecasting model of regional rainfall-induced landslides, which quantitatively considered both the static and dynamic factors including the geological and geographical factors.
Although the modern methods and techniques are truly helpful to accurately measure the change of the parameters including the slope deformation, precipitation, soil moisture content and even the seepage pressure, the LMFSs still can hardly forecast the landslide efficiently as expected. Then, the geologists denoted that the variation of the above parameters is a necessary, but not sufficient, condition to the occurrence of the landslide.
As is well known, according to the Newton’s First Law, force is the source of the change of motion state. The landslide, as a kind of “motion”, is closely related to the change of the “force” inside the slope. Therefore, the sliding force in the potential landslide area should be regarded as the core parameter to be effectively measured. Accordingly, the force sensors are the key components in an effective LMFS.
Each type of sensor has its own advantages and limitations during the application. Due to the different properties of the key force-sensing element, the common force sensors can be divided into static and dynamic types.
In the aspect of static measurement, for instance, the differential resistor sensor [
11] and the resistance strain-gauge sensor [
12] are normally used into general industrial projects and laboratory applications. Vibrating wire sensors [
13,
14] are utilized in high stress fields for quasi-static tests due to the properties of high compressive strength. Wang et al. [
15] designed an improved type of vibrating wire sensor to measure the anchorage stress in underground engineering.
In the aspect of dynamic measurement, piezoelectric sensors are frequently used. By using piezoelectric acoustic emission sensors, Agioutantis et al. [
16] monitored the failure process of the Nestos marble in three points bending tests, and investigated the potential for accurate prediction of rock damage based on the measuring results. Karayannis et al. [
17] used the embedded cement-based piezoelectric sensors to ensure the safety of the concrete structures by measuring the dynamic force. By using the state-space method, Yan et al. [
18] studied the time-dependent behaviour of a simply-supported functionally graded beam bonded with piezoelectric sensors and actuators. Yang et al. [
19,
20] used the dynamic sensors for damage identification in the structural health monitoring field. Gu et al. [
21], Chalioris et al. [
22] and Voutetaki et al. [
23] used piezoelectric transducers as smart aggregates to evaluate and monitor structural health of reinforced concrete, since the piezoelectric transducers have the advantages of multiple monitoring functions such as dynamic seismic response detection, structural health monitoring and white noise response detection.
It is worth noting that He et al. [
24,
25,
26], Tao et al. [
27] and Yang et al. [
28] developed a real-time remote LMFS based on the vibrating wire sensor shown in
Figure 1 due to its high strength property, which has been applied in some slope engineering projects and obtained remarkable achievements by long-termly measuring the changing tension of the monitoring anchor cable. Thus, it proved that the LMFS [
24,
25,
26,
27,
28] based on the force sensors has made a great breakthrough in the landslide monitoring field. However, for the reason that the working principle of the LMFS [
24,
25,
26,
27,
28] depends on the long-term evolutionary trend analysis of the sliding force since the only vibrating wire sensor in the existing sensing system has the weak ability for dynamic measurement, the monitoring and forecasting efficiency of the LMFS [
24,
25,
26,
27,
28] is constrained and the transient disturbing signals can hardly be captured. Unfortunately, the escaping transient disturbance has become progressively more critical to the development of landslides.
On condition that the sliding force is considered as the generating mechanism, the landslides can be divided into natural and disturbance-induced types, whose evolutionary processes of sliding force are shown in
Figure 2.
The natural landslide is a kind of quasi-static evolutionary behaviour. Strong weather variations, including heavy storms and blizzards, would affect the internal structure and external loading conditions in the potential landslide area. Thus, the rise of the sliding force develops together with the decline of the sliding resistance force in the slope. Once the sliding force exceeds the resistance, the landslide will occur theoretically.
The disturbance-induced landslide has a kind of relatively rapid evolutionary process. Under the natural conditions, the sliding force and the resistance are in equilibrium. The excavation unloading effect induced by human engineering activities, as well as the change of natural conditions, would make the equilibrium state vulnerable. Once an accident occurs, the consequent disturbance will abruptly break the equilibrium and cause landslides.
Currently, disturbance-induced landslides have become more frequent with the increasing human engineering activities. The randomness and non-controllability of disturbance-landslides further increase the monitoring and forecasting difficulties. Thus, the requirement of increasing the dynamic monitoring ability of the existing LMFS is more urgent.
In view of the situation that the disturbance-induced landslides occur more frequently and the lack of dynamic monitoring sensors in the existing LMFSs, as well as considering the conventional dynamic monitoring sensors can hardly adapt the high stress condition in slope engineering, a high-performance piezoelectric force sensor is presented in this paper.
This paper is organized as follows. Firstly, the piezoelectric force sensor is designed, in which two technical indexes based on the loading conditions in the practical slope engineering are prearranged and two key techniques are employed. Secondly, the advisable-dimensional prototype of the sensor is assembled, which can be theoretically validated to satisfy the two presented indexes. Thirdly, the calibration experiments are employed via the independently invented static and transient loading mechanism and the results show that the sensor has fine linearity and stability. Fourthly, the low-frequency correction method is proposed and experimental verified to improve the low-frequency measuring reliability of the sensor. Finally, the conclusions summarize the paper and state that the piezoelectric sensor can complement the existing LMFS for dynamic disturbance monitoring with its excellent static and dynamic properties.
3. Assembly of the Sensor
Figure 10 shows the prototype and the main components of the piezoelectric sensor. The connection pattern of the sensors in the sensing system is depicted in
Figure 11. As can be seen in
Figure 10a,b, the protruding blocks on the front and back of the piezoelectric sensor are utilized to embed into the vibrating wire sensor and the cable lockset respectively, as is shown in
Figure 11. Three force-sensing elements are inside the piezoelectric sensor. Each element shown in
Figure 10c comprises two PZT-5 piezoelectric patches and two copper conducting slices with the same cross section. As
Figure 10c shows, the positive electrodes of the two piezoelectric patches are both connected by the copper conducting slice (ii), on which a lead wire is attached. The connection method shown in
Figure 10c is widely used in the practical application due to the advantages such as improving the signal-to-noise ratio and eliminating the extra insulation between the positive electrode and the basic structure. All the output voltage signals of the three force-sensing elements are converged at the outlet of the piezoelectric sensor via the lead wires placed into the ring-like groove in the base. Before being collected, the final output voltage signals are regulated by the embedded capacitive circuit shown in
Figure 9.
Six screws and three set screws are used for the assembly of the piezoelectric sensor, which are shown in
Figure 10e. As can be seen, plate (i), plate (ii), and the base are fastened by the six screws with uniform angles. The three set screws are bolted into plate (i). The three moving slices, each of which is thinner than plate (ii), are butted on the top of the force-sensing elements by the heads of the set screws, respectively. The height of the single force-sensing element is designed to be slightly higher than the depth of corresponding cylindrical groove in the base and the height of the movable slice is lower than that of the plate (ii) when plate (ii) is placed on the top surface of the base, as can be seen in
Figure 10d.
Thus, most of the external pressure applied on plate (i) will be transferred onto plate (ii), while only a small part onto the movable slices as well as the force-sensing elements. Due to the high compressive strength of the basic steel-made structure, including the plate (i), plate (ii), the base and the other components, the sensor can bear the high pressure exceeding the compressive strength of piezoelectric ceramic and reach what the Index 1 requires.
It is worth mentioning that all the screws including the six screws and the three set screws should be fully tightened and all the connection gaps among the components of the sensor, as well as the exposed grooves in the relative components should be sealed up to isolate from the outer moist. Besides, as the PZT-5 piezoelectric ceramic is brittle and easily damaged by concentrated force, the contact of the PZT and the adjacent components, such as the conducting slice (i) and (ii), should be plane-to-plane type.
With the connection between the outlet of the piezoelectric sensor and the DAE via a data wire, the response voltage signals to the changing external forces can be obtained. The DAE used in this paper is the LMS Spectrum Testing System developed by Belgium LMS Co.
The parameters of the piezoelectric sensor and its main components are presented in
Table 3, based on which, it can be theoretically verified whether the sensor can meet the Indexes 1 and 2 by using the two techniques of SSPDM and CCVDM in
Section 2.2.
Figure 12 presents the equivalent spring model of the piezoelectric sensor based on
Figure 10, where
K1 in Part B denotes the total stiffness of the three force-sensing elements in parallel. Each element contains two piezoelectric patches, whose total stiffness is
K11, and two copper conducting slices, whose total stiffness is
K12. Then
K1,
K11 and
K12 can be obtained as
where
E11 and
A11 are the elastic modulus and the electrode surface area of the PZT-5 piezoelectric patch, respectively, which can be acquired in
Table 1 as 117 MPa and 314 mm
2. Similarly,
E12 and
A12 are those of the copper conducting slices, which are 100 MPa and 314 mm
2.
l11 and
l12 mean the total height of the two piezoelectric patches and the two copper electrode slices, respectively, which are 6 mm and 10 mm. Thus,
K1 can be obtained as 6.12 × 10
9 N/m.
Meanwhile,
K2, the total stiffness of the steel main body in Part B can also be calculated as 2.1 × 10
11 N/m when substituting the corresponding parameters in
Table 3 into Equation (3).
As a result, when the piezoelectric sensor is bearing external ultimate static load of 1500 kN and transient force amplitude of 500 kN, the component forces applied on the force-sensing elements,
F1, can be respectively calculated as 43.23 kN and 14.41 kN by using Equation (4). The former is less than the total ultimate compressive capacity of the three piezoelectric force-sensing elements, which can be obtained by the product between the parameters of compressive strength and cross sectional area shown in
Table 1. The latter can be used to calculate the response voltage via Equation (2) and the theoretical solution is 1025 V, which is more than 100 times greater than the regular signals since the maximal measuring range of the LMS Spectrum Testing System is within 10 V. Therefore, the advisable
C2-to-
C1 ratio in the capacitive circuit shown in
Figure 9 can be selected as 1/200, so that the final output voltage for the ultimate transient load is 5.01 V and the theoretical sensitivity coefficient of the sensor is 0.01 V/kN.
4. Calibration Experiments
Conducting the calibration experiments and acquiring the sensitivity are absolutely necessary for the sensors before their practical engineering applications. The purpose of the calibration for the sensors [
11,
12,
13,
14,
15,
16,
17,
18,
19,
20,
21,
22,
23,
24,
25,
26,
27,
28], as well as the sensor in this paper, is similar, which is to obtain the relation between the input physical parameter, such as force, and the output parameter, such as voltage. However, the requirements of the calibration experiments are different, which depend on the practical application environments.
For the sensor in this paper, the calibration experiments should as closely as possible simulate the actual static and dynamic loading conditions in slope engineering. Thus, the static and dynamic characteristics of the sensor can also be verified. However, considering the severe stress conditions in practical slope engineering, conventional testing machines can hardly match the high requirements. For instance, the drop hammer impact tester can provide only dynamic load (without static preload); the fatigue tester can provide both static preload and dynamic load, but the loading amplitudes are limited. Therefore, this section presents an independently invented static and dynamic loading mechanism, which can provide step-load with different amplitudes on the basis of the preload. Thus, it can meet the high static and dynamic loading requirements.
For the perspective of frequency-domain analysis, the step-load is a typical wide-frequency range exciting signal. The steeply rising stage and its peak value of the step signal are mainly composed of the high-frequency components, while the platform stage is low-frequency components [
30,
31]. Then, using the step-load with different amplitudes as the input excitation to calibrate the sensor can mostly eliminate the nonlinear response errors caused by the low-frequency piezoelectric charge leakage. Thus, it is advisable to employ the wide-frequency step-load for the calibration of the sensor. However, the common defect of low-frequency measuring errors for the piezoelectric sensor cannot be avoided in the practical engineering application, since the dynamic disturbing signals are various, containing the low-frequency disturbance certainly. The low-frequency correction method will be specially introduced in
Section 5.
4.1. Experimental Setup
As can be seen in
Figure 13 and
Figure 14, the whole loading mechanism comprises a preloading system and a step-loading system, which can respectively provide the high enough static preload and the transient load with certain amplitudes. The preloading system is located upon the platform, which contains the components of the Hydraulic Press, beam (i), support (i), the lugs and the pins, while the step-loading system under the platform is composed of the quick-release hook, beam (ii), support (ii), the matching lugs and the pins, and the other components. By the pinned connections, both the upper and lower loading systems can be regarded as levy mechanisms.
Both the two loading systems provide load on the sensing system through the anchor cables, where the sensing system and its connecting pattern are shown in
Figure 11. The anchor cables are made up of six steel strands. As a kind of hollow structure, each of the sensing components including the piezoelectric sensor, the vibrating wire sensor and the cable locksets is crossed in the middle by the anchor cables. Cable lockset (i) is clamped at the top of the anchor cables and cable lockset (ii) is at the bottom. The two locksets are also fixedly connected with and inside lug (ii) and lug (iv), respectively. Thus, the whole sensing system can be compacted by lockset (ii) under the platform when applying load as follows.
At the first step for performing the static preload,
A, the free end at the right of the beam (i), can be slowly pulled up by the Hydraulic Press, which can provide both uniaxial tension and compression force, as is shown in
Figure 14. Then,
B is consequently raised since
O at the left hinged end of beam (i) can be considered as the hinged support of the levy mechanism. As a result, the lockset (i) stretches upward the anchor cables. Thus, the sensing system including the piezoelectric sensor and the vibrating wire sensor is tightly compressed by lockset (ii). By monitoring the reading of the vibrating wire sensor, the static preload can be controlled to reach the design values.
The second step is to provide transient step-load on the basis of the first step, which can be divided into two substeps. Firstly,
A1, the free end at the left of beam (ii), can be quasi-statically pulled up by the pulley mechanism shown in
Figure 15b. Thus, the total compression force applied on the sensing system is further increased. That is, an additional part of preload is applied on the sensing system by the lower loading system in this substep. By monitoring the reading of the vibrating wire sensor, this additional preload can be quantitatively controlled. Secondly, using the quick-release hook, the additional preload applied on the sensing system in the first substep can be unloaded transiently. Thus, the step-load with the specific amplitude is achieved via the lower loading system.
Meanwhile, the response voltage of the piezoelectric sensor can be recorded by the LMS Spectrum Testing System during the action of the step-load. With the output response voltage Δ
U, and the input specific amplitude of step-load Δ
F, the sensitivity coefficient of the piezoelectric sensor
α can be obtained, that is,
In particular, the maximum tension that the Hydraulic Press can provide is 600 kN and the ultimate bearing capacity of the quick-released hook is 5 t (almost 50 kN) so that the pulley mechanism shown in
Figure 15b can provide tension of 100kN. As is shown in
Figure 13, the length of
OA is designed as 2 times more than
OB and
O1A1 is 5 times more than
O1B1. Based on the leverage principle, the upper and lower loading system shown in
Figure 13 and
Figure 14 can apply enough high static load of 1800 kN and transient load with amplitude of 600 kN , which can satisfy the requirements expressed in Indexes 1 and 2, respectively.
It is also worth mentioning that the strength and stiffness of all the components in the loading mechanism have been theoretically verified and qualified under the ultimate static and dynamic conditions.
4.2. Experimental Scheme
Table 4 presents the scheme of the calibration experiments, which can be divided into four sets according to the different static preload levels, namely, 300 kN, 600 kN, 1000 kN and 1500 kN. Each set can be further classified into several subsets based on the amplitude of the step-load. Take subset 1-1 as an example. The first step is to apply the static preload of 300 kN on the sensing system, while at the second, the transient step-load with the amplitude of 50 kN will be applied.
4.3. Experimental Results
The peak response voltage of every subset is also reported in
Table 4. Considering the limited space and the similarity of the response signals to the step-load, only the subsets 1-2, 2-2, 3-3 and 4-1 in every set are given, as is shown in
Figure 16.
By means of the linear fit, the fitting lines of the experimental results can be obtained in
Figure 17. As can be seen, the piezoelectric sensor has a satisfactory wide range of linearity, since the goodness of fit of the each fitting straight line is more than 96%. The sensitivity coefficient of the sensor can be acquired as 0.0081 V/kN by calculating the average slope of the fitting straight lines.
In addition, all of the fitting straight lines are almost coincident with each other and their slopes are quite similar. It indicates the piezoelectric sensor has a stable sensitivity under different preload levels.
Particularly, the subset 4-1 shown in
Table 4 can be used to check the static ultimate compressive bearing capacity of the piezoelectric sensor, since the peak static load has reached the static index of 1500 kN. As is shown in
Figure 16d, the response result of subset 4-1 have the consistent property with the other sets and follows the linear law shown in
Figure 17. It illustrates that the sensor maintains the good behavior under the ultimate preload condition. Therefore, the piezoelectric sensor has satisfied the static-load requirement of Index 1.
Besides, the amplitude of the transient load in subset 3-3 is 500 kN, which has reached maximum of the dynamic measuring index, the response result in
Figure 17 shows that it also follows the linear law with the other subsets. That is, the measuring range of the piezoelectric sensor is as wide as 500 kN at least, which is enough to satisfy Index 2.
Therefore, both the prearranged Indexes 1 and 2 for the piezoelectric sensor has been well satisfied experimentally.
It is noteworthy that the experimental sensitivity coefficient of the piezoelectric sensor is less than 20% with the theoretical result, which can be acceptable when considering the error source of the limited piezoelectric property and machining precision. Moreover, the coherence of the theoretical and experimental results can also verify the helpfulness and effectiveness of techniques 1 and 2 in
Section 2.2.