Next Article in Journal
Hindrances to the Utilisation of the Metaverse for Net-Zero Buildings in South Africa
Previous Article in Journal
Building Energy Simulation of Eco-Efficient Composite Bricks Using a Novel Method to Convert Ceramic Bricks into Building Blocks: A Case Study in Iran
 
 
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Design of Sensorized Rail Pads for Real-Time Monitoring and Predictive Maintenance of Railway Infrastructure

by
Amparo Guillén
1,
Oswaldo Guerrero-Bustamante
1,2,
Guillermo R. Iglesias
3,
Fernando Moreno-Navarro
1 and
Miguel Sol-Sánchez
1,*
1
Laboratory of Construction Engineering, University of Granada, C/Severo Ochoa s/n, 18071 Granada, Spain
2
Department of Civil & Environmental Engineering, Universidad de la Costa, Barranquilla 08001, Colombia
3
Department of Applied Physics, Faculty of Science, University of Granada, 18071 Granada, Spain
*
Author to whom correspondence should be addressed.
Infrastructures 2025, 10(2), 45; https://doi.org/10.3390/infrastructures10020045
Submission received: 20 January 2025 / Revised: 11 February 2025 / Accepted: 13 February 2025 / Published: 19 February 2025

Abstract

Embedding sensors in rail pads allows for direct monitoring of train–track interaction, which is essential for preventive maintenance and sustainable management of railway infrastructure. Nonetheless, given the critical role that rail pads play in enhancing railway track performance and durability, it is crucial to define the optimal configuration of the sensorized pads (InterActive Pads) that ensures both mechanical reliability and functional accuracy. Also, before its widespread application, it is mandatory to provide calibration and modelling to allow for preventive maintenance, improving sustainable management. Thus, this research optimizes the design of rail pads with embedded piezoelectric sensors while validating its performance and developing calibration models to enable the implementation of preventive measures for railroad tracks. Laboratory tests identified the optimal sensor position at the rail pad extremity, featuring a half-embedded design with a gap beneath to ensure mechanical resistance and durability. Large-scale testing further facilitated the development of a calibration model that enhances diagnostic accuracy and supports proactive and sustainable maintenance strategies. The findings demonstrate a strong correlation between sensor signals and train-induced forces, allowing predictions of long-term track performance. This predictive capability enables more effective maintenance, reducing costs and improving safety. By providing a sustainable solution for railway management, this research lays the groundwork for future implementation on real tracks, offering a robust framework for proactive, data-driven maintenance strategies.

1. Introduction

From the 19th century to the present day, railway lines have been playing a pivotal role in both the economy and society, aiming to provide a reliable and efficient means of transportation. As such, they have gathered significant attention regarding the need for effective maintenance and cost management while considering that train traffic is hindered as little as possible and optimizing the flexibility of conservation [1,2,3,4]. Worldwide, extensive research endeavours have been seeking alternative methods for enhancing the efficiency and cost effectiveness of infrastructure conservation.
High railway maintenance costs have driven the search for more efficient maintenance strategies [1,5]. The maintenance procedure is crucial, as preventive actions help detect irregularities in track geometry—such as changes in nominal track gauge, longitudinal or transverse levelling, and track alignment. By addressing these issues early, the risk of derailments and delays in the transportation of people or goods could be significantly reduced. Cost savings of up to 12% when comparing the corrective and preventive maintenance [6] lead to numerous studies aiming to develop prevention strategies or models to identify the most efficient timing for pathway intervention [7,8,9]. Soleimanmeigouni I. et al. (2020) [10,11] studied the impact of track intervention at different inspection intervals, considering the action limit (AC), intervention limit (IL) and immediate action limit (IAL) as defined by Standard EN-1348-5 (2018) [12], which specifies limit values for intervention. Their research focused on the probability of isolated defects, such as longitudinal and transverse levelling, and predicted how long the track could be maintained in a certain condition. This underscores the importance of timing interventions based on the railway superstructure deterioration, emphasizing the value of preventive maintenance [13,14].
One of the key components of railway maintenance involves assessing the condition of infrastructure to guide decision making, often using track recording cars or other conventional methods for condition monitoring [15,16]. In recent years, technologies such as Structural Health Monitoring (SHM) have gained attention as tools for data acquisition to inform maintenance strategies [17]. SHM integrates advanced sensors to measure the health and integrity of railway structures, providing real-time data that can be analyzed to detect early signs of deterioration. While SHM is not a maintenance strategy itself, it supports decision making by enabling more efficient, proactive maintenance approaches. Given the strategic position of rail pads in controlling train–track interaction [18], several studies have explored embedding sensors within them [19,20,21,22]. However, due to the critical role of these pads in railway track performance, it is essential to define the optimal approach for sensor integration to ensure long-term reliability and accuracy in data measurement [22,23]. This careful integration process is crucial to ensure the reliability of acquired data, which underscores the complexity of sensor embedding. To achieve this, calibration of sensors is necessary both in laboratory settings and in the field. Additionally, the development of laboratory models that can be extrapolated to real-world conditions is essential for building predictive models. These models may include machine learning approaches [23,24], data-driven methods such as Artificial Neural Networks (ANN) [25,26] or field measurements [27].
In this context, the present paper aims to optimize, validate and prove the sensor integration in rail pads in order to achieve sustainable management for the railway infrastructure. This is crucial as the previous stage to a wider application and integration in railway tracks for reliable real-time monitoring system. In particular, this research focuses on the use of piezoelectric sensors embedded into rail pads since it has been previously seen as a suitable solution from different points of view [22,23,28,29], demonstrating its sensitivity to record key track factors like variations and discrepancies in the wheel-track contact as well as creating preventive maintenance models. Also, this type of device presents a linear behaviour with the changes in loads from such contact [30,31], providing a cost-effective smart material with the capacity to monitor the train–track contact, which could be of use in key applications such as detecting irregular contact due to flat wheels damaging the track, weighting vehicles in freight transportation and evaluating the evolution of track behaviour in transition zones or problematic areas. These are key indicators that are identified on time and could enrich the preventive model creation.
However, previous research has focused on proving the viability of using embedded piezoelectric sensors into the pad [22,23], but limiting to its superficial integration, which restricts its long-term performance and reliability. Therefore, the present research goes to a further step to optimize the design and validation of such smart pads by assessing the impact of diverse designing variables related to the process of embedding piezoelectric devices/sensors on the reliability sensitivity, and durability of smart rail pads. Also, once the optimal design is achieved, the efficient and necessary information from the vehicle–track interaction permits the application of preventive-calibration models. The objective is to cultivate an instrument that not only improves design and production processes but also contributes to the advancement of real-time load monitoring within rail operations. With this purpose, the research was divided into three main steps: (1) the design of the pad considering the optimal process of embedding sensors to provide a sensitive device; (2) the mechanical validation of the pad in order to prove the reliability and durability of the smart track component and (3) the development of a calibration model that with the obtained data could prevent the railway track maintenance activities.

2. Methodology

2.1. Materials

To carry out this research, several types of rail pads were manufactured at the laboratory in order to evaluate different ways to embed sensors, aiming to define the optimal design of smart pads for further validation, calibration and predictive modelling. This section explains the raw material and procedure for manufacturing the rail pads with different designs to include the sensors, the characteristics of the piezoelectric sensors proven in this research and the track components used to test the smart pad under simulated track conditions.

2.1.1. Rail Pads

For the manufacturing of the pads, one type of granular polymeric material was employed that was melted to produce pads with different designs. The type of material used in this study was constant in order to assess the influence of varying the pad design and the process of embedding the sensor, without dependence on the material properties. In particular, the granular polymer was obtained by cutting waste waterproofing layers composed mainly of low-density polyethylene (LDPE) but also containing other polymers such as polypropylene (PP). The application of this material takes advantage of using a recycled component, which is of great abundance in the local area of this research. In addition, this material was selected as the main research material considering that previous research [23,32] had shown its feasibility for use in pad fabrication, while it serves the purpose of the article, which is more focused on optimizing the design of the sensor embedding, rather than developing the pad material.
The pads were geometrically designed with dimensions of 140 mm × 180 mm × 8 mm since these are common dimensions for the use of pads into fastener systems like those consisting of clip [33]. The pads were manufactured using a two-part mould as can be observed in Figure 1, this process being divided into two steps:
(1)
Obtaining a plain pad with the indicated dimensions (140 mm × 180 mm × 8 mm). For this purpose, 200 g of granulated polymer material (maximum particle size of 4 mm) was poured into a mould, heated at 190 °C for two hours and compacted by applying 18 kN on a metal plate that was placed on top of the molten material.
(2)
Obtaining a pad with surface geometry. The previous plain pad was heated to 140 °C, and a metallic plate, at 200 °C, with the specific geometrical surface was placed on the plain pad. This allowed for printing the geometry on the pad by applying a force of up to 30 kN.
Figure 1. Rutting and surface geometry process of the rail pad.
Figure 1. Rutting and surface geometry process of the rail pad.
Infrastructures 10 00045 g001
As a result of this process, the value of static stiffness (according to the Standard EN 13146-9) [33] of the pads, used as a reference for assessing the optimization of including the sensors, was around 150 kN/mm, which could be qualified as soft. This indicates the suitability of using this type of pads for studying the process of embedding the piezoelectric devices.

2.1.2. Sensors

The choice of sensor type for this research was the piezoelectric devices, which have highlighted its significant potential for detecting changes in strain and stresses from the rail–sleeper contact, as demonstrated by prior studies [22,23]. Also, the piezoelectric sensor was selected because of the low-cost material and was easy to implement.
The type of piezoelectric sensor consisted of a metallic circular base with a 35 mm diameter of a metallic base and 24 mm of the quartz sheet. The total thickness of these devices was around 0.35 mm, which makes them appropriate to be embedded in different track elements, as it was employed in this study and due to their low variability in the pad shape. To measure the electrical changes due to variations in infrastructure/traffic characteristics, they were connected to a data logger with the capacity to directly measure the voltage generated by the piezoelectrics.
The functionality of this sensor consisted of measuring the level of stress on the pad as a voltage unit when it experiences vibrations or deformations. This voltage was conducted by a cable to the controller of the testing machine to record the voltage signals as performed with other instruments used in laboratory tests.

2.1.3. Components of Test Configuration

To test the influence of the designing factors of the InterActive pad, laboratory tests were carried out simulating the composition of the superstructure of ballasted railway tracks. For this purpose, a 250 mm long UIC 54 type rail section and a 357 mm long prestressed concrete sleeper section were used, as can be observed in Figure 2a. Also, to simulate the track bed under the sleeper, a section of rubber mat with 10 mm of thickness was placed under the sleeper, presenting a dynamic stiffness at 10 Hz (Cdyn10Hz) close to 0.10 N/m3 according to DIN 45673-5. This configuration was used to assess the sensitivity of the pad with the embedded sensors since it was seen in previous studies [23] to be effective for this purpose.
Figure 2a also shows the cables connecting the piezoelectric included into the pad, with the controller unit of the equipment to record and analyze the voltage values generated when applying loads over the rail to simulate the distribution of loads that take place in ballasted railway tracks (Figure 2b). This configuration was used because in previous studies [23] it was seen as suitable to measure and record the signals from the piezoelectric sensors.
Figure 3 shows the full-scale test box that was employed in order to simulate the real train passing simulations. The test box was 100 cm long and 100 cm wide, compounded by (i) a ballast layer of 30 cm composed of ophitic aggregates with appropriate properties for this application according to EN 13450 [34]; (ii) a section of concrete sleeper (85 cm of length) commonly used in Spanish railway lines; (iii) a fastening system, of type VM, commonly used for concrete sleepers, including the rail pad with the sensor; and (iv) a rail section type UIC 54, 25 cm long. This type of test has been validated in previous studies [23], demonstrating the capability of the sensorized devices to detect various track effects in a controlled section designed to closely replicate the real-world environment in which they would be applied.

2.2. Testing Plan and Methods

Figure 4 shows that the testing plan was divided into 3 main stages: (i) the optimization of the design of pads including sensors, aiming to define key parameters like optimal position of the sensor into the pad and procedure of embedding; (ii) validation of the mechanical behaviour and durability of the InterActive rail pad and (iii) the development of calibration models to enable the implementation of preventive measures for railroad tracks.

2.2.1. Optimization of InterActive Pad Design

In order to define an optimal design of the rail pads incorporating a sensor to obtain reliable and versatile results to achieve sustainable management, it was necessary to carry out four steps, detailed in Figure 5. These steps were defined in order to achieve the optimal position and incorporation of the sensor inside the rail pad without disrupting its physical composition and integrity. Therefore, these four study steps were carried out:
  • Influence of horizontal and vertical position
It was required to define the position of the sensor on the rail pad; in order to analyze the influence of the horizontal position, three possible locations on the rail surface were studied (Figure 5—Step 1): (a) centre, tested in previous studies [22,23], being used as a reference in this article; (b) at the end of the pad under the rail seat area (this was selected because previous studies [20] have seen that this area of the pad is the least damaged during fatigue processes); and (c) outside the rail seat area, to see what happened when the sensor had no contact with neither the rail nor the sleeper, but some vibrations are transmitted through the pad centre to the end.
To further assess the impact of the vertical position of the sensor, it was investigated involving two distinct positions into the pad thickness (Figure 5—Step 2). Firstly, the performance of the sensor was examined when placed into the pad side with direct contact with the rail, referred to as “side up”. Secondly, the sensor behaviour was studied when positioned beneath the rail pad side, in contact with the sleeper, referred to as “side down”. The objective was to compare both variables and determine the pad side where it was more appropriate the embedding of the sensor, while aiming to find the position where the sensor would be less damaged.
2.
Influence of level of embedding sensor
After determining the vertical and horizontal positions of the sensor, the influence of the level of embedding the sensor of material was analyzed. There were three positions studied (Figure 5—Step 3): (a) a sensor on surface without being embedded, the load was applied to the sensor in an unembedded hole to be used as a reference, as this configuration had been used in previous studies [22,23], and in this article it was used to see the impact of embedding and covering the sensor with polymer; (b) an embedded sensor with half of the depth filled in; and (c) a sensor completely embedded with material. These last two positions depend on the amount of material located above the sensor, which was carried out by designing a metal piece for compacting the plastic over the sensor, using a process that avoids damaging the piezoelectric.
3.
Influence of incorporation of the sensor
With the aim of providing versatility and capacity to remove the sensor without the need to lift the track, this part of the study focused on assessing the viability of designing a removable sensorial piece to be extracted from the rest of the pad (Figure 5—Step 4). This could allow for the repairing of the sensor without the need to replace the whole pad or using different technologies and types of sensors depending on the objectives of the supervision, which would provide a versatile system with the capacity to adapt to the maintenance requirements of each moment.
For this purpose, the impact of designing a removable sensing device that fits into a cavity generated in the main pad was evaluated (referred to in this paper as “full-contact removable sensor”). Also, as it could require some space between both elements for the replacement of the sensing device, the impact of using a piece with a lower size than the cavity was assessed, and therefore, obtaining less contact between the sensor and the rest of the pad (named as “less-contact removable sensor”). On the other hand, to evaluate the effect of these designs, the results were compared to those obtained for a reference with an embedded sensor [23]. This allowed us to assess the differences in load transmission to the sensor and how it influenced the recorded signal.
Figure 5. Scheme of steps for the design of the sensor fitting into the rail pad.
Figure 5. Scheme of steps for the design of the sensor fitting into the rail pad.
Infrastructures 10 00045 g005
To assess the influence of these parameters, a series of tests were carried out to simulate different traffic loading conditions on the system including the rail pad with the embedded sensors in the different locations. In this paper, this test is referred to as “Dynamic loading simulating different traffic conditions”, adapted to the dynamic stiffness loading test according to Standard EN 13146-9 [33].

2.2.2. Prove and Validation of Smart Pad

To validate the viability of the smart pad including the sensor, a three-stage assessment for the most appropriate designs and procedures for including the sensor was conducted, according to the previous results. In detail, the following evaluations were carried out:
(1)
Mechanical characterization: This stage aimed to examine the potential influence of the sensor design on the alteration of the pad stiffness and mechanical characteristics.
(2)
Durability testing: This phase focused on assessing the long-term viability and durability of the proposed solution, referring to climate and fatigue resistance.
(3)
Functionality assessment: The third stage involved evaluating the solution capacity to measure rail–track interactions when simulating the passage of trains with different levels of load per axle.
At every stage of proof and validation, specific assessments were conducted. The mechanical characterization phase involved static and dynamic stiffness tests performed in accordance with Standard EN 13146-9 [33]. Subsequently, durability testing assessed sensor resilience through fatigue test under repeated loads, thereby scrutinizing the pad durability with the piezoelectric sensor. Finally, to ensure the pad functionality, a simulated train passage test was carried out.

2.2.3. Calibration Modelling Development

Once the optimal rail pad design with integrated sensors was finalized and the complete monitoring system mechanically validated, a two-step calibration model for the InterActive pads was developed. This calibration model was divided into two steps:
(1)
Calibration: The first step involved calibrating the sensors embedded in the rail pads. This calibration was performed by evaluating the InterActive pads performance when applying loads similar to those of the train axles. In this study, this test was referred as “Train Simulation Test”. This test enabled the evaluation of the pad incorporating sensors, assessing their ability to detect various loads at different frequencies. The goal was to correlate the sensor measured signal with the varying load characteristics simulated during train passages. This would enable us to determine the railway track condition based on the recorded signal. In this way, the device would provide more reliable data for effective and sustainable management.
(2)
Modelling: In the second and final step, to develop the calibration model, large-scale tests were conducted, referred to in this article as “Full-scale testing box test under constant cyclic load”. These tests involved applying a large number of cycles of both, constant load cycling and train load simulation passing. The primary goal was to obtain substructure settlement values comparable to those observed in railway tracks. The tests evaluated the tendency of the section to settle based on the applied load level. By analyzing the received signal and using the calibration that correlates load and signal, the settlement caused by each load could be predicted, allowing for the determination of when maintenance would be required.

2.3. Testing Method

The dynamic loading test simulating different traffic conditions (carried out to evaluate the influence of the designing factors) was developed from adapting the tests recorded in the standard EN 13146-9 [33], which include the assessment of the rail pads under dynamic loads. The purpose of this test consisted of assessing the ability of the sensor to detect load changes in the rail–track interaction, which is the objective of the final design of the smart pad. This test included 5 dynamic load steps consisting of applying 1000 cycles at 5 Hz, with stresses of 200, 400, 800, 1600 and 2400 kPa, which were chosen because they represent the range of loads expected for this material during its application on tracks. In addition, to determine the susceptibility to changes in load frequency, 3 frequencies were studied for the 1600 kPa load [35].
The stiffness test, carried out for the evaluation of the impact of the sensor on the mechanical performance of the pad, was developed in consonance to the European standard 13146-9 [34]. For this purpose, two types of tests were carried out to measure the static and dynamic stiffness of the optimal designs for the smart pad. The static tests consisted of applying three loading cycles between 18 kN and 85 kN, at a rate of (15 kN/min) for the first and second loading cycle, and at 5 kN/min in the third cycle. From this process, the static stiffness was calculated as the slope of the force-displacement curve of the third loading cycle, between 18 kN (representing the force applied by the clip of the fastening) and 68 kN (expressed in the standard as the 80% of the maximum load and representing unfavourable loading conditions expected in tracks).
In the case of the dynamic stiffness test, this consisted of applying 1000 cycles of a cyclic load ranging from 18 kN to 68 kN (80% of the 85 kN maximum force). During this dynamic testing, plate displacements were measured at varying frequencies: 1 Hz, 5 Hz and 10 Hz. This thorough evaluation allowed us to gain valuable insights into the pad response to both static and dynamic loading conditions across different frequency ranges.
The fatigue test carried out when assessing the durability of the smart pad consisted of simulating repeated loads reproducing the stress transmitted by train curved sections (considered the most unfavourable testing condition) on the smart pad. The test was adapted from Standard EN 13146-4 [36] to focus on determining the durability of the sensor. It consisted of applying 3 × 10 6 cycles at 5 Hz with an amplitude between 5 kN and 75 kN, under a load angle of 33° from the vertical direction over the head of a rail section mounted on the system pad-fastening sleeper. The durability assessment was conducted for the sensor, following the criteria outlined in Standard EN 13146-4 [36]. The criterion adopted required that the amplitude of the sensor signal, as measured after the fatigue test, should not exceed a variation between +20% and −20% of the average derived from the amplitudes measured at the beginning of the test.
The test for simulation of the train passage was designed in consideration of a procedure previously applied to smart rail pads [22,23]. This test focused on a loading sequence consisting of a series of loads with stresses between 0 and 2000 kPa (the same level of loading expected on rail pads for a singular section under the central sleeper), with the capacity to produce different states of deformation in each system, to be detected by the piezoelectric sensor. This could be appropriate for monitoring variations in track deflections under various traffic conditions over a constant section, or changes in sleeper movements under the same load level because of different configurations/states of the system, like those associated with the “hanging sleeper” phenomenon (increasing the sleeper oscillation accelerations) or abrupt stiffness changes, among others.
For the large-scale tests conducted on the testing box, a series of application cycles simulating train passages (as described earlier) was performed. This included a total of 31,250 simulated train passages, representing a train travelling at 100 km/h with 100% of the estimated load. These simulations equated to 500,000 applied load cycles, allowing long-term analysis of the InterActive Pad sensor behaviour. The tests evaluated whether the device could consistently measure load signals during vehicle–track interactions without a decrease in signal amplitude and maintain accurate measurements under varying load applications. To calibrate these sensors, tests were conducted by running each train at least once with varying loads—specifically, at 100 km/h with loads of 50%, 75% and 100% of their maximum capacity.

3. Analysis of Results

3.1. Parametric Design Study

Figure 6 analyzes the sensor position on the rail pad when analyzing the possibility to incorporate the device at the end of the pad under the rail seat area or outside the rail seat area (Influence of the horizontal position) and whether, in the chosen position, it would be side up or side down (Influence of the vertical position). The results represent the percentual variation between a reference (embedded centred sensor in the rail pad) compared to the different horizontal and vertical positions. The graph shows the signal variation when applying three different stress levels: 800, 1600 and 2400 kPa, representing positive values when increasing the signal recorded at varying such parameters, and negative values when decreasing the sensitivity of the sensor.
The results indicate that positioning the sensor at the end of the pad under the rail seat area yields the highest positive signal variation—up to 414% compared to a centred sensor—likely due to lateral pad deformation transmitting strain to the sensor. This positioning not only enhances sensitivity to changes in wheel–rail interaction but also could offer durability benefits, as the end of the pad is less prone to damage during service [37]. Negative percentage variability is observed when the sensor is positioned outside the rail area. This indicates that such a placement is not advantageous. To ensure reliable results, it is preferable to position the sensor in contact with both the rail and the sleeper. Additionally, placing the sensor on the bottom face of the pad (pad–sleeper contact) showed similar signal variation to the upper face (pad–rail contact). Overall, the findings highlight the advantages of end placement for creating versatile, durable and adaptable smart rail pad technologies.
Moreover, Figure 7 displays the percentual signal variation for the same three stress levels (800, 1600 and 2400 kPa) employing as a reference the completely embedded centred sensor. In this case, the level of embedding the sensor (without embedding or embedded with recess) was analyzed and as well as the possibility of the sensor to be removed from its place (fitted into the rail pad or with a partial space between the removable device and the rail pad hole). The embedded with recess sensor displayed the highest signal variation percentages (26, 61 and 121%); this could be associated with the fact that this inferior cavity could allow some bending of the pad at that point, facilitating the vibration of the sensor and therefore, its sensitivity.
The analysis revealed that the removable sensor fully integrated into the pad (maintaining complete contact with the surrounding surface) exhibited up to 300% greater signal variation compared to the embedded sensor, demonstrating its effectiveness for train–track interaction monitoring. This design offers practical advantages, including the ability to remove and replace the sensor when needed or adapt it to evolving monitoring requirements or future technological advancements. However, the partially integrated removable sensor, while still capable of detecting load changes, produced lower signal levels, potentially reducing reliability. Thus, ensuring full contact between the sensor and the pad is essential for accurately measuring strain and stress distribution from the centre to the end of the pad.

3.2. Mechanical Validation

3.2.1. Static and Dynamic Stiffness

Figure 8 displays the static and dynamic stiffness of the two types of pads; the Reference pad (the pad including an embedded sensor in the centre, as studied in previous research) [13,14]; and the InterActive pad (the removable one was considered for this study, since it gave reasonable results while providing other advantages as it is removable). In addition, it is the most unfavourable case due to the composition requiring various elements, so it is the one employed as the most critical one in the stage of validation. As can be observed, the static stiffness values were 183 kN/mm for the Reference pad and 219 kN/mm for the InterActive pad, which could be qualified as medium and stiff. On the other hand, the dynamic stiffness values were 3016 kN/mm for the Reference pad and 3629 kN/mm for the InterActive pad.
The InterActive pad shows an increase of 16% compared to the reference, which may be linked to the inclusion of the sensor. However, it still qualifies as a medium-stiffness pad, similar to the reference. The dynamic stiffness also increases by 17%, but the dynamic/static ratio remains unchanged, confirming that the sensor’s inclusion does not significantly alter the pad’s mechanical properties. Therefore, this validates that the sensor, even in a removable system that modifies the rail pads geometry, does not substantially affect the mechanical properties of the assembly. Although, these stiffness values are mostly associated with the type of polymer used to manufacture the pads and the geometric design, so they will depend mostly on these parameters. Also, in this study, a recycled polymer was used as an example, but the solution could apply to any type of pad.

3.2.2. Fatigue Test

To evaluate the InterActive rail pad durability, the percentual variation of the piezoelectric sensor signal in three loading steps, before and after the fatigue test, was measured. As can be observed in Figure 9, in the case of the centre sensor pad (Reference), the variant was 181% for the 30 kN load, 231% for the 60 kN load and 454% for the 90 kN load. This suggests that increasing the load amplifies fatigue variation, indicating that as the material undergoes fatigue loads, it becomes more vulnerable to higher stresses. Nonetheless, the one that showed the lowest percentual variations before and after the fatigue test, taking into consideration the piezoelectric sensor signal, was the InterActive pad. For the pad in study, the percentages were 5% and 54%, and in the last load step, the signal was increased so the percentual variation was 24%. This could be due to the clearance between the sensor and the pad, the sensor being the least damaged component of the rail pad.
Based on the test results, the InterActive pad demonstrated less variability compared to the Reference pad in the tests conducted before and after fatigue. This indicates the accurate capacity of the InterActive pad to keep detecting changes in load level even after an elevated number of load cycles applied. Ensuring more consistent performance for monitoring in both the short and long term makes it highly valuable for assessing the railway structure and anticipating its behaviour, enabling the creation of predictive models by the same.
Secondly, to study the InterActive pad in depth, Figure 10 shows the values of the sensor signal amplitude during the fatigue process. As mentioned previously, a criterion was implemented, as different load cycles were applied, the sensor continued measuring loads for values between +20% and −20% of the average of the amplitudes. The results indicate that the sensor conserved its ability to detect variations in load levels following the fatigue test. Consequently, it could be asserted that the sensor demonstrated commendable durability by maintaining its capacity to detect load changes and delivering consistent performance after the fatigue test.

3.3. Calibration Modelling Development

3.3.1. Calibration: Train Simulation Test

Figure 11 displays the results from different train simulation load tests, in this case for 50, 75 and 100% of load capacity, at 100 km/h. The graph indicates a clear relationship between the applied load and the recorded signal magnitude. The 100% train simulation generates the largest dynamic response, followed by the 75% and 50% simulations, which demonstrate progressively smoother and less variable behaviour. This pattern highlights the increasing impact of higher loads on the system response.
Furthermore, Figure 12a,b illustrate the linear response of the train simulation load tests for each load percentage (50%, 75% and 100%), as well as the dispersion of the results for three measurements analyzed at each load level. Firstly, on Figure 12a, it could be seen that the combination of each train load displayed a reliability of about 98%. This confirms the results observed in Figure 11, where each of the train axles was easily detected. A 76% variation is observed in the load applied, ranging from the lowest axle value of 13.26 kN to the highest value of 56.25 kN. Similarly, the signal emitted by the InterActive pad shows a 91% variability between its lowest and highest detected voltage values. These high variations indicate a steeper slope and, consequently, greater sensor sensitivity, suggesting an improved ability to detect tension changes in the fastening system where it is applied.
Additionally, on Figure 12b, when analyzing the dispersion results, the 50% train simulation load results displayed dispersion values of about 1.5, which were elevated for sensor calibration and, conversely, with the load increasing the dispersions decreased. This indicated a sensor better performance for higher load values, a significant dispersion decrease was observed with trains at 75% load, as their values drop below the 0.30 dispersion, being the lowest ones for the 100% train load simulation dispersion results, with dispersion values of 0.09 which could be more indicated for sensor calibration results. This could be due to the fact that, as the load increases, the sensor produces stronger signals, leading to reduced variability and a lower margin of error, further enhancing result reliability.
Thus, the InterActive pad would be expected to provide more reliable results at higher loads, which is particularly beneficial for railroads where train loads are substantial. Therefore, tests were conducted with trains loaded to 100%, representing the most demanding conditions for the track. This approach provides more reliable results for both model calibration and field application.

3.3.2. Modelling: Full-Scale Testing Box Tests Under a Constant Cyclic Load

In order to create a model of prediction of track settlement, Figure 13 displays the results of settlement of a ballasted section reproduced by a full-scale testing box. This allowed for predicting the track behaviour under different load levels to be detected by the InterActive Pad, and therefore, correlating the signal measured in track with the prediction in track settlement.
As observed in Figure 13, diverse load cycling tests were carried out at a 5 Hz frequency, applying a total of 500,000 cycles representing train load axles for a 40, 50 and 60 kN of amplitude load. These values correspond to axle loads of 16 t (assuming 50% on the central sleeper and 25% on the adjacent sleepers—aligning with findings from [33]), 20 t and 24 t. Hence, three tendency settlement slopes could be identified, Slope I for the first 150,000 load cycles, Slope II between 150,000 and 250,000 and Slope III in the last 250,000 cycles of the test.
Furthermore, it could be observed that as the slope number increases, the settlement difference for each load cycle step also increases. It could be observed that, for all three load levels, the slope is steeper in the first stage due to a greater initial recompacting of the ballast layer, as supported by various authors and prior studies [38,39]. As the number of load cycles increases, the slope decreases, indicating that the layer becomes more consolidated. Furthermore, when comparing different load levels, the slope increases with higher loads, with this trend being more pronounced in the early stages. Notably, the slope shows a more significant increase when the load shifts from 50 kN to 60 kN (around a 40% increase in settlement).
Continuing the analysis and development of the calibration model, Figure 14 presents the correlation between the train loads to be monitored by the sensor signal output (primary y axis), as seen previously in Figure 12, as well as the correlation with the trend to settlement (secondary y axis) of the track depending on train load and stage, as seen in Figure 13. So, combining the results of sensor calibration, depending on level of load, with the results from the testing box predicting track settlement, the calibration model to the correlated sensor signal with track behaviour could be defined.
In detail, Figure 14 includes the calibration equation for the sensor ( y = 0.4582 x 5.8728 ); Y being the voltage displayed and X the load applied on the track) and, moreover, it represented each of the previously analyzed slopes (Slope I, Slope II and Slope III) for track settlement. These slopes illustrate the behaviour of the railway substructure settlement over the application of various loading cycles and, consequently, over time, with each slope tendency line accompanied by its respective equation.
The slopes were determined by calculating the values corresponding to each cycle interval described earlier. The linear equations for each slope were y I = 6.04 × 10 7 x 1.80 × 10 6 x ; y I I = 4.20 × 10 7 x 3.16 × 10 6 ; and y I I I = 9.88 × 10 8 x 4.71 × 10 7 ; Y being the settlement tendency and X the load applied, with Slope I producing the highest seating tendency results, followed by Slope II with lower values and Slope III with the least. This trend aligns with previous analyses, showing that as the number of cycles increases, the settlement stabilizes, resulting in reduced deformation values. Additionally, it remains evident that as the applied load increases, the tendency for settlement also increases, with the 60 kN load case exhibiting a settlement of 3.824 × 10 5 for Slope I, 2.992 × 10 5 for Slope II and 6.248 × 10 6 mm/cycle for Slope III, for example.
Calibration Model Implementation and Validation
To validate the previously proposed calibration model, it was implemented based on the results obtained after subjecting the InterActive Pad to a total of 500,000 train load cycles with varying axle loads. Figure 15 demonstrates a schematic procedure of the model application on the railway track field. Starting with the signal voltage values obtained from the InterActive Pads when a load is applied to the railroad, and incorporating the calibration model, it becomes possible to determine the number of applied cycles and the section settlement at that moment. This information could then be used to inform decisions on implementing appropriate maintenance strategies.
Thus, as seen in Figure 14, with the signal recorded (y), the load applied by the train wheel (x) on the rail could be obtained and transmitted to the InterActive Pad:
W h e e l   L o a d   k N x = 2.04 y + 14.55
And, therefore, with the wheel load obtained and the load cycle knowledge previously processed, following the application of 500,000 cycles on the testing box, it aligns with the Slope III section indicating that
T e n d e n c y   t o   s e t t l e   m m c y c l e   y = 9.88 × 10 8 x 4.71 × 10 7 ;
where “x” is the wheel load applied, previously calculated, and y is the tendency to settle.
In this case, with the passage of 31,250 train axles at a speed of 100 km/h and using the parameters derived from the previously discussed equations, Table 1 was obtained as an example of an application to predict the settlement at three different points (cycles 150,000; 250,000 and 500,000, as examples), supposing that the sensor recorded a signal of 20 V. The structure’s settlement for each cycle could be determined assuming the same load is applied throughout all cycles. For example, at 150,000 cycles, the settlement could be approximately 3.16 × 10 5 ; mm, at 250,000 cycles, 2.01 × 10 5 mm and at 500,000 cycles, 5.00 × 10 6 mm. In this case, a settlement of 12.27 mm was observed, corresponding to the scenario where the train is loaded with 55.35 kN and subjected to approximately 500,000 load cycles. This aligns with the trends shown in Figure 13, illustrating settlements for different loads within the 50 kN to 60 kN range.
Additionally, Figure 16 compares the settlement predicted by the implemented model, when the train loads are included in the equation, with the measured settlement in the testing box after 500,000 train simulation load cycles at full capacity. These 500,000 cycles reflect field conditions where cumulative loads are recorded over time, in contrast to the previous numerical case, which represents a single point load application.
Results reflect that the final settlements differed by only 6%, with the model predicting 6.30 mm and the testing box measuring 6.73 mm. This close agreement validates the calibration model’s potential for real-world application on railway tracks.
Furthermore, the slopes of both settlement curves exhibited parallel behaviour, with major differences in Slope II of 14% at 150,000 cycles and 10% at 250,000 cycles. As observed earlier, the initial slope is the steepest, reflecting the most unstable behaviour up to 150,000 cycles. Beyond this point, the slope decreases as the number of cycles increases, indicating a stabilization of the ballast mass. Despite these variations, the model provides an accurate estimation of track substructure settlement, a critical parameter for assessing structural health.
This laboratory validation demonstrates that the model could be effectively extrapolated to field conditions, incorporating the application of both constant loads and train axles with varying loads or specific irregularities in the track–vehicle interaction. These factors could also be analyzed to predict whether they might result in uneven settlements yielding results that closely align with the actual behaviour of the railway structure and its settlement [40].

4. Discussion

This study aligns with existing research on the sensing of railway track components, particularly smart rail pads. However, unlike previous studies [19,23], it advances the field by optimizing both the design and integration of these sensors from a functional and mechanical perspective. The findings highlight the critical importance of considering design parameters when incorporating sensors into rail pads, the horizontal placement of the sensor being a key factor that significantly enhances sensitivity in detecting wheel–rail interactions. The placement of sensor at the pad extreme allows for monitoring irregularities in the contact and stresses induced in the fastening while also improving durability, corroborating previous research [37]. On the other hand, although the vertical position of the sensor has shown minimal influence on its performance, placing it beneath the pad (pad–sleeper contact) could enhance durability by shielding it from external climatic conditions.
Furthermore, exploring the removal of the sensor from the rail pad itself could lead to a more versatile and adaptable design, better suited to future technological advancements. Accordingly, this study employed piezoelectric sensors, building upon the work of Sol-Sánchez et al. [23]. These sensors are not only widely used but also among the most cost effective, directly influencing their durability when embedded in the rail pads. However, alternative sensor types could be explored, as suggested by other researchers [19,21,22], piezoelectric materials such as PZT (lead zirconate titanate), BT (barium titanate) or CdS (cadmium sulfide) could offer different performance characteristics. For example, PZT might provide higher sensitivity due to its greater piezoelectric constants, while BT and CdS could offer unique benefits in certain environments. Additionally, the optimization of the sensor electrical hardware could further improve the signal quality. For instance, using coaxial cables to reduce noise in signal transmission could enhance the accuracy of the measurements, which is a factor the authors plan to explore in future studies.
Moreover, in terms of durability, the sensorized pads exhibited appropriate fatigue resistance, consistent with the findings of previous studies [23]. This allows for both short- and long-term data collection, which should be further validated in real-world railway infrastructure applications. These results also have identified predictive models that associate the sensor’s received signal with the applied load level. Laboratory tests have further correlated this with settlement tendencies. This aligns with previous studies [7,40] aimed at developing predictive models for preventive maintenance. However, these models are limited to laboratory conditions and require field validation. Such validation would enable an assessment of both predictive capabilities and the sensors’ effectiveness in detecting train-wheel load characteristics, permitting the real-time measurement of bogie loads and weights, which could be particularly appropriate in freight railway lines, for example.

5. Conclusions

This article examines the feasibility and optimization of rail pad designs incorporating embedded piezoelectric sensors, specifically for real-time load monitoring in rail infrastructure for preventive maintenance operations. Focusing on key factors such as reliability, sensitivity and durability, the study seeks to enhance the production process of these smart monitoring components, contributing to more effective structural health monitoring in railway systems. Achieving this optimized design enabled the development of a calibration model for the sustainable management of railway infrastructure. Based on these findings, the following conclusions could be drawn:
-
Results showed that the most influential parameter on the sensor placement was related to its location along the pad under the rail foot, obtaining the highest sensitivity to monitor changes in traffic and track conditions when placing the piezoelectric sensor at the end of the rail pad. This is associated with the higher pad movements and displacements at the edge of the contact rail–sleeper, therefore, leading to a higher ability of the sensor to measure such interaction efforts.
-
Nonetheless, it must be considered that both the embedding sensor and the removable sensor led to a slight modification of rail stiffness. The pad with the hole for incorporating the removable sensor showed a 16% increase, maintaining the dynamic/static ratio implicating a stable behaviour of rail pads on its mechanical properties.
-
The laboratory tests into the fatigue process demonstrated that the most appropriate method for sensor inclusion was designed as a system composed of the main pad plus a removable sensor insert. This design not only demonstrated durability over the millions of fatigue cycles but also ultimately proved itself to continue to fulfil the role of monitoring track conditions.
-
The calibration results confirmed the ability to detect different axles under varying train loads. Additionally, the dispersion for trains loaded at 100% was approximately 0.09, demonstrating the pads’ high sensitivity and effective performance under heavy loads, fulfilling their intended functionality on the track.
-
Large-scale tests, designed to simulate conditions on the railroad track, demonstrated the behaviour of the system under the application of severe load cycles, up to 60,000 newton load magnitude. Even after applying this number of cycles, the sensor exhibited only a 0.78% decrease in reliability when measuring various trains.
-
With the obtained data, a calibration model was developed in which the passive slopes of the railway superstructure were defined and linked to the behaviour of the InterActive Pad over the same number of cycles. By analyzing the relationship between the signal obtained and the settlement tendency slope for the given cycle count, a clear correlation could be established between the sensor signal and the settlement slope; this allowed for an assessment of the current condition of the track, facilitating informed decisions regarding necessary maintenance actions.
-
Finally, when validating the calibration model, the final settlement comparative value of 6.30 and 6.73 mm could be obtained in the case of the train simulation cycles applied in the laboratory testing box. This highlights the model capability to accommodate trains with constant loads, axles carrying varying loads and irregularities in the vehicle–track interaction. This provides a predictive tool to be implemented on railway maintenance activities and managing an active and sustainable management program.
This article presents the optimal design of InterActive rail pads, including those equipped with a removable sensor, showcasing their capability to monitor railway track conditions under various scenarios. The paper underscores the essential role of rail pads in railway infrastructure, as they directly influence the rigidity and behaviour of the system under train loads. A comprehensive study was conducted to identify the ideal placement of piezoelectric sensors within the rail pads, ensuring accurate and reliable data for extensive research. With this optimal InterActive Pad design, a sustainable calibration model for management could be developed, enabling preventive maintenance actions that address the structural health of the railway system.
The objective was to implement intelligent technologies that seamlessly integrate with elements of the railway infrastructure, facilitating the collection of data for sustainable maintenance while minimizing monitoring costs. Moving forward, the next step would be to deploy these InterActive Pads in operational railway sectors to enable the large-scale extrapolation of these models.

Author Contributions

Conceptualization, M.S.-S.; methodology, M.S.-S.; software, G.R.I.; validation, A.G., formal analysis, A.G. and O.G.-B.; investigation, A.G. and O.G.-B.; resources, M.S.-S.; data curation, A.G.; writing—original draft preparation, A.G.; writing—review and editing, M.S.-S.; visualization, F.M.-N.; supervision, M.S.-S., O.G.-B. and G.R.I.; project administration, M.S.-S.; funding acquisition, M.S.-S. and F.M.-N. All authors have read and agreed to the published version of the manuscript.

Funding

The present study has been conducted within the framework of a project with the acronym “InterActive Pads” and titled “Proof of Smart Pads for Monitoring Vehicle–Track Interaction” (PDC2022-133966-I00), funded by the Ministry of Science, Innovation and University of Spain (MICIU/AEI/10.13039/501100011033) and by the European Union Next Generation EU/PRTR.

Data Availability Statement

The data that support the findings of this study are available from the author upon reasonable request.

Conflicts of Interest

The authors declare that there are no conflicts of interest.

References

  1. Buurman, B.; Gkiotsalitis, K.; van Berkum, E.C. Railway Maintenance Reservation Scheduling Considering Detouring Delays and Maintenance Demand. J. Rail Transp. Plan. Manag. 2023, 25, 100359. [Google Scholar] [CrossRef]
  2. Macedo, R.; Benmansour, R.; Artiba, A.; Mladenović, N.; Urošević, D. Scheduling Preventive Railway Maintenance Activities with Resource Constraints. Electron. Notes Discret. Math. 2017, 58, 215–222. [Google Scholar] [CrossRef]
  3. Peng, F.; Kang, S.; Li, X.; Ouyang, Y.; Somani, K.; Acharya, D. A Heuristic Approach to the Railroad Track Maintenance Scheduling Problem. Comput.-Aided Civ. Infrastruct. Eng. 2011, 26, 129–145. [Google Scholar] [CrossRef]
  4. Odolinski, K.; Boysen, H.E. Railway Line Capacity Utilisation and Its Impact on Maintenance Costs. J. Rail Transp. Plan. Manag. 2019, 9, 22–33. [Google Scholar] [CrossRef]
  5. Jovanovic, S. ECOTRACK: A Track Maintenance Management System Directed towards Optimal Life Cycle Costs. In Proceedings of the Conference: Asset Management in Railway Infrastructure Conference—IQPC, London, UK, 13 November 2000. [Google Scholar]
  6. Lakušić, S. Road and Rail Infrastructure III. In Proceedings of the 3rd International Conference on Road and Rail Infrastructure, Split, Croatia, 28–30 April 2014. [Google Scholar]
  7. Lopes Gerum, P.C.; Altay, A.; Baykal-Gürsoy, M. Data-Driven Predictive Maintenance Scheduling Policies for Railways. Transp. Res. Part C Emerg. Technol. 2019, 107, 137–154. [Google Scholar] [CrossRef]
  8. Le-Nguyen, M.H.; Turgis, F.; Fayemi, P.E.; Bifet, A. Real-Time Learning for Real-Time Data: Online Machine Learning for Predictive Maintenance of Railway Systems. Transp. Res. Procedia 2023, 72, 171–178. [Google Scholar] [CrossRef]
  9. Giunta, M.; Bressi, S.; D’Angelo, G. Life Cycle Cost Assessment of Bitumen Stabilised Ballast: A Novel Maintenance Strategy for Railway Track-Bed. Constr. Build. Mater. 2018, 172, 751–759. [Google Scholar] [CrossRef]
  10. Soleimanmeigouni, I.; Ahmadi, A.; Khajehei, H.; Nissen, A. Investigation of the Effect of the Inspection Intervals on the Track Geometry Condition. Struct. Infrastruct. Eng. 2020, 16, 1138–1146. [Google Scholar] [CrossRef]
  11. Soleimanmeigouni, I.; Ahmadi, A.; Nissen, A.; Xiao, X. Prediction of Railway Track Geometry Defects: A Case Study. Struct. Infrastruct. Eng. 2020, 16, 987–1001. [Google Scholar] [CrossRef]
  12. UNE-EN 13848-5; Aplicaciones Ferroviarias. Vía. Calidad Geométric Ade La Vía. Parte 5: Niveles de Calidad Geométrica. Plena Vía y Aparatos de Vía. Asociación Española de Normalización: Madrid, Spain, 2018.
  13. Mohammadi, A.; Amador-Jimenez, L.; Nasiri, F. Reliable, Effective, and Sustainable Urban Railways: A Model for Optimal Planning and Asset Management. J. Constr. Eng. Manag. 2020, 146, 6. [Google Scholar] [CrossRef]
  14. Li, P.; Tang, Y.; Zheng, Z.; Wang, Z.; Zhuang, Y. Semiautomated Railway Line Information Modeling Based on Asset Management Data. J. Constr. Eng. Manag. 2024, 150, 10. [Google Scholar] [CrossRef]
  15. Sauni, M.; Luomala, H.; Kolisoja, P.; Nummi, T. Advancing Railway Asset Management Using Track Geometry Deterioration Modeling Visualization. J. Transp. Eng. A Syst. 2022, 148, 04021106. [Google Scholar] [CrossRef]
  16. Neuhold, J.; Vidovic, I.; Marschnig, S. Preparing Track Geometry Data for Automated Maintenance Planning. J. Transp. Eng. A Syst. 2020, 146, 5. [Google Scholar] [CrossRef]
  17. Roveri, N.; Carcaterra, A.; Sestieri, A. Real-Time Monitoring of Railway Infrastructures Using Fibre Bragg Grating Sensors. Mech. Syst. Signal Process 2015, 60–61, 14–28. [Google Scholar] [CrossRef]
  18. Shen, C.; Zhang, P.; Dollevoet, R.; Zoeteman, A.; Li, Z. Evaluating Railway Track Stiffness Using Axle Box Accelerations: A Digital Twin Approach. Mech. Syst. Signal Process 2023, 204, 110730. [Google Scholar] [CrossRef]
  19. van Schalkwyk, M.H.; Gräbe, P.J. Condition Monitoring of Train Wheels Using a Cost-Effective Smart Rail Pad. Eng. Res. Express 2022, 4, 035045. [Google Scholar] [CrossRef]
  20. Woschitz, H.; Woschitz, H. Development of A Rail-Strain-Pad Using FBG Sensors. In Proceedings of the 5th International Conference on Structural Health Monitoring of Intelligent Infrastructure (SHMII-5) 2011, Cancún, Mexico, 11–15 December 2011. [Google Scholar]
  21. Zhang, S.L.; Koh, C.G.; Kuang, K.S.C. Proposed Rail Pad Sensor for Wheel-Rail Contact Force Monitoring. Smart Mater. Struct. 2018, 27, 115041. [Google Scholar] [CrossRef]
  22. Sol-Sánchez, M.; Castillo-Mingorance, J.M.; Moreno-Navarro, F.; Rubio-Gámez, M.C. Smart Rail Pads for the Continuous Monitoring of Sensored Railway Tracks: Sensors Analysis. Autom. Constr. 2021, 132, 103950. [Google Scholar] [CrossRef]
  23. Sol-Sánchez, M.; Castillo-Mingorance, J.M.; Moreno-Navarro, F.; Mattinzioli, T.; Rubio-Gámez, M.C. Piezoelectric-Sensored Sustainable Pads for Smart Railway Traffic and Track State Monitoring: Full-Scale Laboratory Tests. Constr. Build. Mater. 2021, 301, 124324. [Google Scholar] [CrossRef]
  24. Sainz-Aja, J.A.; Ferreño, D.; Pombo, J.; Carrascal, I.; Casado, J.; Diego, S.; Castro, J.; Rivas, I. Machine Learning-Based Parametric Analysis of Railway Systems. In Proceedings of the Sixth International Conference on Railway Technology: Research, Development and Maintenance, Prague, Czech Republic, 1–5 September 2024; pp. 1–10. [Google Scholar]
  25. Popov, K.; De Bold, R.; Chai, H.-K.; Forde, M.C.; Ho, C.L.; Hyslip, J.P.; Kashani, H.F.; Long, P.; Hsu, S.S. Big-Data Driven Assessment of Railway Track and Maintenance Efficiency Using Artificial Neural Networks. Constr. Build. Mater. 2022, 349, 128786. [Google Scholar] [CrossRef]
  26. Bhadouria, A.S.; Braga, J.A.P.; Mishra, R.P.; Andrade, A.R. Data-Driven Approach for Condition-Based Maintenance of Freight Train Wheelsets Using Markov Decision Process. pp. 1–15. In Sixth International Conference on Railway Technology: Research, Development and Maintenance; Civil-Comp Conferences: Edinburgh, UK, 2024; Volume 7, Paper 7.14; ISSN 2753-3239. [Google Scholar] [CrossRef]
  27. Tsunashima, H.; Nakano, T. Track Management Method Based on In-Service Vehicle Vibration Measurements. In New Research on Railway Engineering and Transport; IntechOpen: London, UK, 2023; pp. 1–14. [Google Scholar] [CrossRef]
  28. Castillo-Mingorance, J.; Sol-Sánchez, M.; Moreno-Navarro, F.; Gámez, M. A Critical Review of Sensors for the Continuous Monitoring of Smart and Sustainable Railway Infrastructures. Sustainability 2020, 12, 9428. [Google Scholar] [CrossRef]
  29. Su, D.; Sun, C.; Wang, L. Design and Evaluation of a Piezoelectric-Electromagnetic Energy Harvester with a Lever Structure. Mech. Syst. Signal Process 2024, 220, 111642. [Google Scholar] [CrossRef]
  30. Prego, F.J.; Solla, M.; Núñez-Nieto, X.; Arias, P. Assessing the Applicability of Ground-Penetrating Radar to Quality Control in Tunneling Construction. J. Constr. Eng. Manag. 2016, 142, 5. [Google Scholar] [CrossRef]
  31. Castillo-Mingorance, J.M.; Sol-Sánchez, M.; Mattinzioli, T.; Moreno-Navarro, F.; Rubio-ámez, M.C. Development of Rail Pads from Recycled Polymers for Ballasted Railway Tracks. Constr. Build. Mater. 2022, 337, 127479. [Google Scholar] [CrossRef]
  32. ET 03.360.574.2; Suelas Bajo Traviesa Para Alta Velocidad. 1a Edición: May 2010; Administrador de Infraestructuras Ferroviarias: Madrid, Spain, 2010.
  33. UNE-EN 13146-9; Railway Applications. Track. Test Methods for Fastening Systems. Part 9: Determination of Stiffness. Asociación Española de Normalización: Madrid, Spain, 2020.
  34. UNE-EN 13450; Aggregated for Railway Ballast. Asociación Española de Normalización y Certificación: Madrid, Spain, 2003.
  35. Milne, D.; Masoudi, A.; Ferro, E.; Watson, G.; Le Pen, L. An Analysis of Railway Track Behaviour Based on Distributed Optical Fibre Acoustic Sensing. Mech. Syst. Signal Process 2020, 142, 106769. [Google Scholar] [CrossRef]
  36. UNE-EN 13146-4; Railway Applications. Track. Test Methods for Fastening Systems. Part 4: Effect of Repeated Loading. Asociación Española de Normalización: Madrid, Spain, 2020.
  37. Hu, P.; Wang, H.; Tian, G.; Liu, Y.; Li, X.; Spencer, B.F. Multifunctional Flexible Sensor Array-Based Damage Monitoring for Switch Rail Using Passive and Active Sensing. Smart Mater. Struct. 2020, 29, 095013. [Google Scholar] [CrossRef]
  38. Sol-Sánchez, M.; D’Angelo, G. Review of the Design and Maintenance Technologies Used to Decelerate the Deterioration of Ballasted Railway Tracks. Constr. Build. Mater. 2017, 157, 402–415. [Google Scholar] [CrossRef]
  39. Tutumluer, E.; Qian, Y.; Hashash, Y.Y.M.A.; Ghaboussi, J.; Davis, D.D. Field Validated Discrete Element Model for Railroad Ballast; Railtec Illinois: Urbana, IL, USA, 2011. [Google Scholar]
  40. Soleimanmeigouni, I. Predictive Models for Railway Track Geometry Degradation. Doctoral Dissertation, Luleå University of Technology, Luleå, Sweden, 2019. ISBN 9789177903109. [Google Scholar]
Figure 2. (a) Simulation of the composition of the super-structure; (b) scheme demonstrating the distribution of loads on a railroad track section.
Figure 2. (a) Simulation of the composition of the super-structure; (b) scheme demonstrating the distribution of loads on a railroad track section.
Infrastructures 10 00045 g002
Figure 3. Testing box description.
Figure 3. Testing box description.
Infrastructures 10 00045 g003
Figure 4. Testing plan.
Figure 4. Testing plan.
Infrastructures 10 00045 g004
Figure 6. Parametric design study to evaluate the positioning of the sensor inside the rail pad.
Figure 6. Parametric design study to evaluate the positioning of the sensor inside the rail pad.
Infrastructures 10 00045 g006
Figure 7. Parametric design study to evaluate the way of incorporating the sensor inside the rail pad.
Figure 7. Parametric design study to evaluate the way of incorporating the sensor inside the rail pad.
Infrastructures 10 00045 g007
Figure 8. Static and dynamic stiffness of the Reference pad and the InterActive pad.
Figure 8. Static and dynamic stiffness of the Reference pad and the InterActive pad.
Infrastructures 10 00045 g008
Figure 9. Percentual variation of measurements of the Reference and InterActive pad before and after the fatigue test.
Figure 9. Percentual variation of measurements of the Reference and InterActive pad before and after the fatigue test.
Infrastructures 10 00045 g009
Figure 10. Sensor amplitude during fatigue test.
Figure 10. Sensor amplitude during fatigue test.
Infrastructures 10 00045 g010
Figure 11. Train simulation test results for different train load percentages.
Figure 11. Train simulation test results for different train load percentages.
Infrastructures 10 00045 g011
Figure 12. Train simulation test results for different train load percentages: (a) calibration linear demonstration and (b) results dispersion.
Figure 12. Train simulation test results for different train load percentages: (a) calibration linear demonstration and (b) results dispersion.
Infrastructures 10 00045 g012
Figure 13. Box tests applying three load levels to evaluate full-scale ballast settlement.
Figure 13. Box tests applying three load levels to evaluate full-scale ballast settlement.
Infrastructures 10 00045 g013
Figure 14. Calibration of train simulation test results and settlement tendency both depend on the load application in the track.
Figure 14. Calibration of train simulation test results and settlement tendency both depend on the load application in the track.
Infrastructures 10 00045 g014
Figure 15. Proposed maintenance prediction model: from data processing to maintenance strategy formulation.
Figure 15. Proposed maintenance prediction model: from data processing to maintenance strategy formulation.
Infrastructures 10 00045 g015
Figure 16. Calibration model verification.
Figure 16. Calibration model verification.
Infrastructures 10 00045 g016
Table 1. Results of analysis of calibration model.
Table 1. Results of analysis of calibration model.
CyclesSlope (mm/Cycle)Signal Output (V)Load Applied (x)Tendency to Settle (mm/Cycle)Structure Settlement (mm)
At 150,000I2055.350.00003164.74
At 250,000II2055.350.00002019.77
At 500,000III2055.350.000005012.27
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Guillén, A.; Guerrero-Bustamante, O.; Iglesias, G.R.; Moreno-Navarro, F.; Sol-Sánchez, M. Design of Sensorized Rail Pads for Real-Time Monitoring and Predictive Maintenance of Railway Infrastructure. Infrastructures 2025, 10, 45. https://doi.org/10.3390/infrastructures10020045

AMA Style

Guillén A, Guerrero-Bustamante O, Iglesias GR, Moreno-Navarro F, Sol-Sánchez M. Design of Sensorized Rail Pads for Real-Time Monitoring and Predictive Maintenance of Railway Infrastructure. Infrastructures. 2025; 10(2):45. https://doi.org/10.3390/infrastructures10020045

Chicago/Turabian Style

Guillén, Amparo, Oswaldo Guerrero-Bustamante, Guillermo R. Iglesias, Fernando Moreno-Navarro, and Miguel Sol-Sánchez. 2025. "Design of Sensorized Rail Pads for Real-Time Monitoring and Predictive Maintenance of Railway Infrastructure" Infrastructures 10, no. 2: 45. https://doi.org/10.3390/infrastructures10020045

APA Style

Guillén, A., Guerrero-Bustamante, O., Iglesias, G. R., Moreno-Navarro, F., & Sol-Sánchez, M. (2025). Design of Sensorized Rail Pads for Real-Time Monitoring and Predictive Maintenance of Railway Infrastructure. Infrastructures, 10(2), 45. https://doi.org/10.3390/infrastructures10020045

Article Metrics

Back to TopTop