DESIGN OF A NEW FULL SCALE TEST-RIG FOR THE CALIBRATION OF INSTRUMENTED WHEELSETS

DESIGN OF A NEW FULL SCALE TEST-RIG FOR THE CALIBRATION OF INSTRUMENTED WHEELSETS

DESIGN OF A NEW FULL SCALE TEST-RIG FOR THE CALIBRATION OF INSTRUMENTED WHEELSETS

Challenge E: Bringing the territories closer together at higher speeds DESIGN OF A NEW FULL SCALE TEST-RIG FOR THE CALIBRATION OF INSTRUMENTED WHEELSETS Giorgio Diana 1 , Ferruccio Resta 1 , Francesco Braghin 1 *, Egidio Di Gialleonardo 1 , Marco Bocciolone 1 , Pietro Crosio 1 1 Dipartimento di Meccanica, Politecnico di Milano, Via La Masa 1, 20156 Milano, Italy *Corresponding author, email: francesco.braghin@polimi.it ABSTRACT Instrumented wheelsets have been used for years to assess running safety of railway vehicles. International Standards (EN 14363 [1]) prescribe a complete testing procedure for the approval of a new vehicle.

As far as running safety is concerned, the sum of the guiding forces in straight track (ΣY) and the derailment coefficient (Y/Q) in curve must be measured. Therefore, measuring wheel-rail contact forces is mandatory when the so-called “normal method” must be applied. In order to use wheelsets as a measuring device, a preliminary calibration procedure is required. Usually instrumented wheelsets are based on strain gauge bridges placed either on the axle or on the wheel web (or both). During the calibration phase forces of known magnitude are applied to the wheelset and the corresponding strain values are measured.

In this way it is possible to establish a correlation between the output of strain gauge bridges and the applied forces. The calibration procedure appears crucial in order to assess the accuracy of the measuring system. Rolling test-rig are often used to perform the calibration of the wheelset [2]. In this way, forces at wheel-rail contact interface are usually not directly measured but are determined by imposing the equilibrium of the system, thus reducing significantly the accuracy of the system. In order to improve the accuracy of the device, a new full scale test-rig has been designed at the Mechanical Engineering Department of Politecnico di Milano.

A bogie assembly made up of a bogie frame and 2/3/4 wheelsets, depending on the vehicle architecture, or a single wheelset in case of a non-bogie vehicle is mounted on the test-rig. This allows to correctly reproduce the real working conditions of the instrumented wheelset that are of fundamental importance: in fact, measured strains depend both of the contact forces and on reaction forces; hence, an error on the boundary conditions determines a re-distribution of reaction forces and thus a measuring error.

Instrumented rail elements are positioned under each wheel of the wheelset under calibration. Each rail element is equipped with 7 load cells in order to measure wheel rail contact force components in each direction (vertical, lateral and longitudinal). Vertical forces are applied to the bogie through the secondary suspensions by means of a load beam, whereas lateral and longitudinal forces are applied directly to the rail elements. 1. INTRODUCTION Test rigs for the calibration of instrumented wheelsets are of two kinds: rolling (such as the one shown in Figure 1) or non-rolling (see Figure 2).

Challenge E: Bringing the territories closer together at higher speeds DESIGN OF A NEW FULL SCALE TEST-RIG FOR THE CALIBRATION OF INSTRUMENTED WHEELSETS Giorgio Diana 1 , Ferruccio Resta 1 , Francesco Braghin 1 *, Egidio Di Gialleonardo 1 , Marco Bocciolone 1 , Pietro Crosio 1 1 Dipartimento di Meccanica, Politecnico di Milano, Via La Masa 1, 20156 Milano, Italy *Corresponding author, email: francesco.braghin@polimi.it ABSTRACT Instrumented wheelsets have been used for years to assess running safety of railway vehicles. International Standards (EN 14363 [1]) prescribe a complete testing procedure for the approval of a new vehicle.

As far as running safety is concerned, the sum of the guiding forces in straight track (ΣY) and the derailment coefficient (Y/Q) in curve must be measured. Therefore, measuring wheel-rail contact forces is mandatory when the so-called “normal method” must be applied. In order to use wheelsets as a measuring device, a preliminary calibration procedure is required. Usually instrumented wheelsets are based on strain gauge bridges placed either on the axle or on the wheel web (or both). During the calibration phase forces of known magnitude are applied to the wheelset and the corresponding strain values are measured.

In this way it is possible to establish a correlation between the output of strain gauge bridges and the applied forces. The calibration procedure appears crucial in order to assess the accuracy of the measuring system. Rolling test-rig are often used to perform the calibration of the wheelset [2]. In this way, forces at wheel-rail contact interface are usually not directly measured but are determined by imposing the equilibrium of the system, thus reducing significantly the accuracy of the system. In order to improve the accuracy of the device, a new full scale test-rig has been designed at the Mechanical Engineering Department of Politecnico di Milano.

A bogie assembly made up of a bogie frame and 2/3/4 wheelsets, depending on the vehicle architecture, or a single wheelset in case of a non-bogie vehicle is mounted on the test-rig. This allows to correctly reproduce the real working conditions of the instrumented wheelset that are of fundamental importance: in fact, measured strains depend both of the contact forces and on reaction forces; hence, an error on the boundary conditions determines a re-distribution of reaction forces and thus a measuring error.

Instrumented rail elements are positioned under each wheel of the wheelset under calibration. Each rail element is equipped with 7 load cells in order to measure wheel rail contact force components in each direction (vertical, lateral and longitudinal). Vertical forces are applied to the bogie through the secondary suspensions by means of a load beam, whereas lateral and longitudinal forces are applied directly to the rail elements. 1. INTRODUCTION Test rigs for the calibration of instrumented wheelsets are of two kinds: rolling (such as the one shown in Figure 1) or non-rolling (see Figure 2).

Challenge E: Bringing the territories closer together at higher speeds DESIGN OF A NEW FULL SCALE TEST-RIG FOR THE CALIBRATION OF INSTRUMENTED WHEELSETS Giorgio Diana 1 , Ferruccio Resta 1 , Francesco Braghin 1 *, Egidio Di Gialleonardo 1 , Marco Bocciolone 1 , Pietro Crosio 1 1 Dipartimento di Meccanica, Politecnico di Milano, Via La Masa 1, 20156 Milano, Italy *Corresponding author, email: francesco.braghin@polimi.it ABSTRACT Instrumented wheelsets have been used for years to assess running safety of railway vehicles. International Standards (EN 14363 [1]) prescribe a complete testing procedure for the approval of a new vehicle.

As far as running safety is concerned, the sum of the guiding forces in straight track (ΣY) and the derailment coefficient (Y/Q) in curve must be measured. Therefore, measuring wheel-rail contact forces is mandatory when the so-called “normal method” must be applied. In order to use wheelsets as a measuring device, a preliminary calibration procedure is required. Usually instrumented wheelsets are based on strain gauge bridges placed either on the axle or on the wheel web (or both). During the calibration phase forces of known magnitude are applied to the wheelset and the corresponding strain values are measured.

In this way it is possible to establish a correlation between the output of strain gauge bridges and the applied forces. The calibration procedure appears crucial in order to assess the accuracy of the measuring system. Rolling test-rig are often used to perform the calibration of the wheelset [2]. In this way, forces at wheel-rail contact interface are usually not directly measured but are determined by imposing the equilibrium of the system, thus reducing significantly the accuracy of the system. In order to improve the accuracy of the device, a new full scale test-rig has been designed at the Mechanical Engineering Department of Politecnico di Milano.

A bogie assembly made up of a bogie frame and 2/3/4 wheelsets, depending on the vehicle architecture, or a single wheelset in case of a non-bogie vehicle is mounted on the test-rig. This allows to correctly reproduce the real working conditions of the instrumented wheelset that are of fundamental importance: in fact, measured strains depend both of the contact forces and on reaction forces; hence, an error on the boundary conditions determines a re-distribution of reaction forces and thus a measuring error.

Instrumented rail elements are positioned under each wheel of the wheelset under calibration. Each rail element is equipped with 7 load cells in order to measure wheel rail contact force components in each direction (vertical, lateral and longitudinal). Vertical forces are applied to the bogie through the secondary suspensions by means of a load beam, whereas lateral and longitudinal forces are applied directly to the rail elements. 1. INTRODUCTION Test rigs for the calibration of instrumented wheelsets are of two kinds: rolling (such as the one shown in Figure 1) or non-rolling (see Figure 2).

DESIGN OF A NEW FULL SCALE TEST-RIG FOR THE CALIBRATION OF INSTRUMENTED WHEELSETS

Challenge E: Bringing the territories closer together at higher speeds Figure 1. Example of rolling test rig (courtesy of Lucchini RS) Figure 2. Example of non-rolling test rig (courtesy of DB and HBM) Rolling test rigs are mainly made up a of two rail discs on which the wheelset to be calibrated is set. In order to apply known loads on the wheelset, hydraulic/electric load actuators equipped with load cells are used [3]. Typically, two vertical actuators (to simulate the load transfer), one lateral actuator (to reproduce the overall lateral force that is applied to the wheelset through the bogie) and two longitudinal actuators (to impose the desired angle of attack) are used.

Moreover, to take into account the relative wheel-rail position, displacement transducers (e.g. laser sensors) at contact level are used. The advantage of such test rigs is that the wheelset can be tested in working conditions that are similar to those encountered during line tests. Instead, the typical main drawback is due to the fact that no direct measure of wheel-rail contact forces is achievable [2]: wheel-rail contact forces are usually determined either through dynamic equilibria of the whole wheelset, knowing the forces applied by the actuators and the wheelset position, and/or through indirect measures, e.g.

the deformation of rail discs near the contact area. Moreover, the lateral forces on the two wheels, as well as the wheel-rail relative lateral position, cannot be adjusted independently but are a function of the imposed angle of attack of the wheelset and of the applied overall lateral force. Finally, due to the adopted suspension system of the wheelset on the roller rig (usually primary suspensions), the bandwidth of the measuring wheelset that can be explored though the test rig is limited to few Hz and the curvature of the rails modifies (even though slightly) the contacting conditions, thus the contact stress distribution and the contact forces.

Non-rolling test rigs, instead, have a much simpler structure: the wheelset to be calibrated is placed on two dynamometric balances having a minimum of three measuring directions (up to a maximum of six measuring directions, i.e. three forces and three moments, [4],[5]) and a shape that approximates real rails. Wheel-rail contact forces are therefore directly measured thus greatly improving the accuracy of the calibration procedure. The drawback, instead, is that the wheelset is not rolling. Thus, if the output of the measuring set-up of the wheelset is a function of the wheelset angular position, calibration tests have to be repeated for different angular positions.

Moreover, in order to calibrate the wheelset under realistic working conditions, it is useful to carry out numerical simulations of the running behaviour of the wheelset. However, being the "equivalent rails" non moving with respect to the wheelset, it is possible to apply external forces either to the wheelset through the journal bearing or directly to the rails. Thus, the bandwidth of the wheelset that can be assessed during the calibration tests is not determined by the suspension system of the wheelset to the test rig but by the bandwidth of the actuators (that is typically higher). Moreover, if lateral and longitudinal actuators are applied to each dynamometric balance, independent lateral and longitudinal contact forces can be applied to each wheel (obviously, appropriate bounding conditions on the wheelset are required) up to their friction limit.

It is also very easy to test the influence of different gauges (as well as different wheel-rail relative lateral positions) and no rail curvature effect is present.

Looking at the pros and cons, non-rolling test rigs seem more suitable for calibrating instrumented wheelsets except for the fact that the calibration procedure is longer and more complicated since it has to be repeated for different wheelset's angular positions and numerical codes (with their intrinsic approximations) for assessing realistic working conditions are usually required. Challenge E: Bringing the territories closer together at higher speeds Figure 1. Example of rolling test rig (courtesy of Lucchini RS) Figure 2. Example of non-rolling test rig (courtesy of DB and HBM) Rolling test rigs are mainly made up a of two rail discs on which the wheelset to be calibrated is set.

In order to apply known loads on the wheelset, hydraulic/electric load actuators equipped with load cells are used [3]. Typically, two vertical actuators (to simulate the load transfer), one lateral actuator (to reproduce the overall lateral force that is applied to the wheelset through the bogie) and two longitudinal actuators (to impose the desired angle of attack) are used. Moreover, to take into account the relative wheel-rail position, displacement transducers (e.g. laser sensors) at contact level are used. The advantage of such test rigs is that the wheelset can be tested in working conditions that are similar to those encountered during line tests.

Instead, the typical main drawback is due to the fact that no direct measure of wheel-rail contact forces is achievable [2]: wheel-rail contact forces are usually determined either through dynamic equilibria of the whole wheelset, knowing the forces applied by the actuators and the wheelset position, and/or through indirect measures, e.g. the deformation of rail discs near the contact area. Moreover, the lateral forces on the two wheels, as well as the wheel-rail relative lateral position, cannot be adjusted independently but are a function of the imposed angle of attack of the wheelset and of the applied overall lateral force.

Finally, due to the adopted suspension system of the wheelset on the roller rig (usually primary suspensions), the bandwidth of the measuring wheelset that can be explored though the test rig is limited to few Hz and the curvature of the rails modifies (even though slightly) the contacting conditions, thus the contact stress distribution and the contact forces.

Non-rolling test rigs, instead, have a much simpler structure: the wheelset to be calibrated is placed on two dynamometric balances having a minimum of three measuring directions (up to a maximum of six measuring directions, i.e. three forces and three moments, [4],[5]) and a shape that approximates real rails. Wheel-rail contact forces are therefore directly measured thus greatly improving the accuracy of the calibration procedure. The drawback, instead, is that the wheelset is not rolling. Thus, if the output of the measuring set-up of the wheelset is a function of the wheelset angular position, calibration tests have to be repeated for different angular positions.

Moreover, in order to calibrate the wheelset under realistic working conditions, it is useful to carry out numerical simulations of the running behaviour of the wheelset. However, being the "equivalent rails" non moving with respect to the wheelset, it is possible to apply external forces either to the wheelset through the journal bearing or directly to the rails. Thus, the bandwidth of the wheelset that can be assessed during the calibration tests is not determined by the suspension system of the wheelset to the test rig but by the bandwidth of the actuators (that is typically higher). Moreover, if lateral and longitudinal actuators are applied to each dynamometric balance, independent lateral and longitudinal contact forces can be applied to each wheel (obviously, appropriate bounding conditions on the wheelset are required) up to their friction limit.

It is also very easy to test the influence of different gauges (as well as different wheel-rail relative lateral positions) and no rail curvature effect is present.

Looking at the pros and cons, non-rolling test rigs seem more suitable for calibrating instrumented wheelsets except for the fact that the calibration procedure is longer and more complicated since it has to be repeated for different wheelset's angular positions and numerical codes (with their intrinsic approximations) for assessing realistic working conditions are usually required. Challenge E: Bringing the territories closer together at higher speeds Figure 1. Example of rolling test rig (courtesy of Lucchini RS) Figure 2. Example of non-rolling test rig (courtesy of DB and HBM) Rolling test rigs are mainly made up a of two rail discs on which the wheelset to be calibrated is set.

In order to apply known loads on the wheelset, hydraulic/electric load actuators equipped with load cells are used [3]. Typically, two vertical actuators (to simulate the load transfer), one lateral actuator (to reproduce the overall lateral force that is applied to the wheelset through the bogie) and two longitudinal actuators (to impose the desired angle of attack) are used. Moreover, to take into account the relative wheel-rail position, displacement transducers (e.g. laser sensors) at contact level are used. The advantage of such test rigs is that the wheelset can be tested in working conditions that are similar to those encountered during line tests.

Instead, the typical main drawback is due to the fact that no direct measure of wheel-rail contact forces is achievable [2]: wheel-rail contact forces are usually determined either through dynamic equilibria of the whole wheelset, knowing the forces applied by the actuators and the wheelset position, and/or through indirect measures, e.g. the deformation of rail discs near the contact area. Moreover, the lateral forces on the two wheels, as well as the wheel-rail relative lateral position, cannot be adjusted independently but are a function of the imposed angle of attack of the wheelset and of the applied overall lateral force.

Finally, due to the adopted suspension system of the wheelset on the roller rig (usually primary suspensions), the bandwidth of the measuring wheelset that can be explored though the test rig is limited to few Hz and the curvature of the rails modifies (even though slightly) the contacting conditions, thus the contact stress distribution and the contact forces.

Non-rolling test rigs, instead, have a much simpler structure: the wheelset to be calibrated is placed on two dynamometric balances having a minimum of three measuring directions (up to a maximum of six measuring directions, i.e. three forces and three moments, [4],[5]) and a shape that approximates real rails. Wheel-rail contact forces are therefore directly measured thus greatly improving the accuracy of the calibration procedure. The drawback, instead, is that the wheelset is not rolling. Thus, if the output of the measuring set-up of the wheelset is a function of the wheelset angular position, calibration tests have to be repeated for different angular positions.

Moreover, in order to calibrate the wheelset under realistic working conditions, it is useful to carry out numerical simulations of the running behaviour of the wheelset. However, being the "equivalent rails" non moving with respect to the wheelset, it is possible to apply external forces either to the wheelset through the journal bearing or directly to the rails. Thus, the bandwidth of the wheelset that can be assessed during the calibration tests is not determined by the suspension system of the wheelset to the test rig but by the bandwidth of the actuators (that is typically higher). Moreover, if lateral and longitudinal actuators are applied to each dynamometric balance, independent lateral and longitudinal contact forces can be applied to each wheel (obviously, appropriate bounding conditions on the wheelset are required) up to their friction limit.

It is also very easy to test the influence of different gauges (as well as different wheel-rail relative lateral positions) and no rail curvature effect is present.

Looking at the pros and cons, non-rolling test rigs seem more suitable for calibrating instrumented wheelsets except for the fact that the calibration procedure is longer and more complicated since it has to be repeated for different wheelset's angular positions and numerical codes (with their intrinsic approximations) for assessing realistic working conditions are usually required.

DESIGN OF A NEW FULL SCALE TEST-RIG FOR THE CALIBRATION OF INSTRUMENTED WHEELSETS

Challenge E: Bringing the territories closer together at higher speeds Before introducing the proposed full scale test rig for the calibration of instrumented wheelsets, it is important to briefly describe how a measuring wheelset works: by correlating measured (using traditional resistive or innovative optical/piezoelectric strain gauges) strains on the axle and/or wheel webs to the forces at wheel-rail contact during the calibration phase, it is possible to determine a correlation matrix [2] that, during the line tests, allows to pass from the measured strains to wheel-rail contact forces.

However strains are a function of both contact forces and reaction forces at axle box. It is therefore of great importance to reproduce on the wheelset's calibration test rig the real contact conditions (rail profile) and boundary conditions (axlebox connecting elements). 2. SPECIFICATIONS AND DESIGN OF THE TEST BENCH According to the considerations done in the previous paragraph, it was decided to design a non-rolling test-rig for the calibration of instrumented wheelsets since it allows to directly measure contact forces. Moreover, in order to be able to exactly reproduce both the contact and boundary conditions, and thus obtain an instrumented wheelset that is able to measure contact forces with high accuracy, it was decided to design a test-rig that allows to accommodate a complete bogie.

In this case, in fact, the instrumented wheelset has boundary conditions that are exactly those it will be experience during inline tests. Note that, in order to reduce the complexity (and cost) of the test-rig, only one instrumented wheelset can be calibrated at a time. If the bogie is equipped with more than one instrumented wheelset, the calibration procedure has to be repeated for each instrumented wheelset. This will be explained better later on.

In Europe different kind of bogies are adopted on vehicles designed for passenger or good transportation. Bogie characteristics in terms of dimensions and architecture vary based on load requirements and type of transportation. The main parameters to be taken into account are:  number of axles:  bogie wheelbase;  rail gauge:  wheel diameter;  maximum values of wheel rail contact forces. With regard to the architecture of the bogie it is possible to find in service bogieless vehicles up to four-axle bogies, when heavy weight vehicles are involved. Bogie wheelbase is strictly related to the architecture, usually the larger the number of wheelsets the smaller the wheelbase.

The rail gauge used is unified for interoperability purposes to 1435 mm, but different values are found all over the railways, for example the gauge of the “Circumvesuviana” (a narrow gauge railway near Naples) is set to 950 mm, whereas values up to 1668 mm are adopted in Spanish and Portuguese railways. Even the diameter of the wheels changes in order to satisfy the requirements, the range of variation is from 360 mm for low-flatcar wagons up to 1200 mm for high-speed trains or locomotives. The maximum axle load can vary between 17 tonnes (for passenger vehicles) and 22.5 tonnes (for loaded freight wagons).

Usually vertical loads are limited to those values in order to prevent track damages. In order to define the maximum lateral load applied to the wheel it has to be taken into account that the maximum Nadal coefficient Y/Q (lateral load over vertical load on the wheel), without obtaining derailment, depends on track geometry and on wheel and rail actual profiles and can reach values up to 1.8 [3]. As far as the longitudinal load is concerned the maximum achievable load is approximately equal to vertical load acting at wheel-rail contact.

The project specifications of the test-rig are defined based on the range of variation of the parameters previously described, in order to guarantee the maximum flexibility of use. The structure of the test-rig is essentially composed by two parts, the base and the portal. The base (shown in green in Figure 3) is constrained to the ground and the rails (grey) are installed on the upper part, having the possibility of varying the rail gauge. Each rail is equipped with the standard UIC60 profile and is mounted with a 1/20 cant, it is divided into two parts, the first one is used only to support the bogie whereas the second part has a measuring function; in fact this part is connected by means of a shaped plate (orange) to the load cells in order to obtain a direct measure of wheel-rail contact forces.

Each plate is also connected to both lateral and longitudinal actuators (black) in order to apply directly forces to the rail in these directions.

DESIGN OF A NEW FULL SCALE TEST-RIG FOR THE CALIBRATION OF INSTRUMENTED WHEELSETS

Challenge E: Bringing the territories closer together at higher speeds Figure 3. Solid model of the test rig for the calibration of instrumented wheelsets The base is also composed by two lateral beams (green), placed parallel to the rails, where the portal is fixed. On the surface of the beams several holes are arranged in order to fix the portal in suitable longitudinal positions with respect to the test-rig. Since the portal acts as ground for the vertical actuators, its longitudinal position is chosen for keeping aligned the vertical actuators with respect to the vertical secondary suspensions.

The portal is made up of two lateral columns (blue) connected to the base and a transversal beam (blue) linking the columns which can be fixed in different positions. the vertical actuators (black) are installed on the transversal beam through spherical joints and their vertical position depends on the can be modified acting on the transversal beam. Additionally their lateral position can be set in order to match the lateral distance of the vertical secondary suspensions. In order to avoid large bogie displacements, bumpstops in lateral and longitudinal directions are used. With regard to the actuators, hydraulic ones were chosen due to their capability to generate huge forces (greater than electric drives for equal volume) and apply small displacements around the same working position without the development of backlashes in the transmission.

To obtain a pass-band as high as possible, it has been decided to place the hydraulic actuators, where possible (i.e. for longitudinal and lateral motions), at wheel-rail interface and equip them with high performance servovalves. In the vertical direction, instead, it was decided to place the actuators on top of the bogie so to apply the forces through the secondary suspension. This decision was taken in order to keep the height of the test rig as small as possible for safety reasons. Thus, although high performance servovalves were chosen also for the vertical actuators, the pass-band in vertical direction is limited by the pass-band of primary and secondary suspensions (where available).

To be able to calibrate almost any kind of instrumented wheelset, it was decided to adopt vertical actuators that could generate a load up to 250kN each. However, due to the flexible configuration of the test rig, even more powerful and, eventually, more than two vertical actuators should be able to fit into the test rig. For what concerns the longitudinal and lateral actuators, although only two actuators for each dynamometric balance would be sufficient (one in longitudinal and one in lateral direction), it was decided to use three actuators (due to the available space two placed longitudinally and one placed laterally) in order to be able to generate any in-plane motion and thus to also assess the influence of the contact point position on the calibration matrix.

To achieve the required flexibility target, also for both longitudinal and lateral actuators it was decided to adopt actuators able to generate a force of 50kN. As for the vertical actuators, if higher forces are required, both longitudinal and lateral actuators should easily be replaced by more powerful ones.

Challenge E: Bringing the territories closer together at higher speeds Figure 3. Solid model of the test rig for the calibration of instrumented wheelsets The base is also composed by two lateral beams (green), placed parallel to the rails, where the portal is fixed. On the surface of the beams several holes are arranged in order to fix the portal in suitable longitudinal positions with respect to the test-rig. Since the portal acts as ground for the vertical actuators, its longitudinal position is chosen for keeping aligned the vertical actuators with respect to the vertical secondary suspensions.

The portal is made up of two lateral columns (blue) connected to the base and a transversal beam (blue) linking the columns which can be fixed in different positions. the vertical actuators (black) are installed on the transversal beam through spherical joints and their vertical position depends on the can be modified acting on the transversal beam. Additionally their lateral position can be set in order to match the lateral distance of the vertical secondary suspensions. In order to avoid large bogie displacements, bumpstops in lateral and longitudinal directions are used. With regard to the actuators, hydraulic ones were chosen due to their capability to generate huge forces (greater than electric drives for equal volume) and apply small displacements around the same working position without the development of backlashes in the transmission.

To obtain a pass-band as high as possible, it has been decided to place the hydraulic actuators, where possible (i.e. for longitudinal and lateral motions), at wheel-rail interface and equip them with high performance servovalves. In the vertical direction, instead, it was decided to place the actuators on top of the bogie so to apply the forces through the secondary suspension. This decision was taken in order to keep the height of the test rig as small as possible for safety reasons. Thus, although high performance servovalves were chosen also for the vertical actuators, the pass-band in vertical direction is limited by the pass-band of primary and secondary suspensions (where available).

To be able to calibrate almost any kind of instrumented wheelset, it was decided to adopt vertical actuators that could generate a load up to 250kN each. However, due to the flexible configuration of the test rig, even more powerful and, eventually, more than two vertical actuators should be able to fit into the test rig. For what concerns the longitudinal and lateral actuators, although only two actuators for each dynamometric balance would be sufficient (one in longitudinal and one in lateral direction), it was decided to use three actuators (due to the available space two placed longitudinally and one placed laterally) in order to be able to generate any in-plane motion and thus to also assess the influence of the contact point position on the calibration matrix.

To achieve the required flexibility target, also for both longitudinal and lateral actuators it was decided to adopt actuators able to generate a force of 50kN. As for the vertical actuators, if higher forces are required, both longitudinal and lateral actuators should easily be replaced by more powerful ones.

Challenge E: Bringing the territories closer together at higher speeds Figure 3. Solid model of the test rig for the calibration of instrumented wheelsets The base is also composed by two lateral beams (green), placed parallel to the rails, where the portal is fixed. On the surface of the beams several holes are arranged in order to fix the portal in suitable longitudinal positions with respect to the test-rig. Since the portal acts as ground for the vertical actuators, its longitudinal position is chosen for keeping aligned the vertical actuators with respect to the vertical secondary suspensions.

The portal is made up of two lateral columns (blue) connected to the base and a transversal beam (blue) linking the columns which can be fixed in different positions. the vertical actuators (black) are installed on the transversal beam through spherical joints and their vertical position depends on the can be modified acting on the transversal beam. Additionally their lateral position can be set in order to match the lateral distance of the vertical secondary suspensions. In order to avoid large bogie displacements, bumpstops in lateral and longitudinal directions are used. With regard to the actuators, hydraulic ones were chosen due to their capability to generate huge forces (greater than electric drives for equal volume) and apply small displacements around the same working position without the development of backlashes in the transmission.

To obtain a pass-band as high as possible, it has been decided to place the hydraulic actuators, where possible (i.e. for longitudinal and lateral motions), at wheel-rail interface and equip them with high performance servovalves. In the vertical direction, instead, it was decided to place the actuators on top of the bogie so to apply the forces through the secondary suspension. This decision was taken in order to keep the height of the test rig as small as possible for safety reasons. Thus, although high performance servovalves were chosen also for the vertical actuators, the pass-band in vertical direction is limited by the pass-band of primary and secondary suspensions (where available).

To be able to calibrate almost any kind of instrumented wheelset, it was decided to adopt vertical actuators that could generate a load up to 250kN each. However, due to the flexible configuration of the test rig, even more powerful and, eventually, more than two vertical actuators should be able to fit into the test rig. For what concerns the longitudinal and lateral actuators, although only two actuators for each dynamometric balance would be sufficient (one in longitudinal and one in lateral direction), it was decided to use three actuators (due to the available space two placed longitudinally and one placed laterally) in order to be able to generate any in-plane motion and thus to also assess the influence of the contact point position on the calibration matrix.

To achieve the required flexibility target, also for both longitudinal and lateral actuators it was decided to adopt actuators able to generate a force of 50kN. As for the vertical actuators, if higher forces are required, both longitudinal and lateral actuators should easily be replaced by more powerful ones.

DESIGN OF A NEW FULL SCALE TEST-RIG FOR THE CALIBRATION OF INSTRUMENTED WHEELSETS

Challenge E: Bringing the territories closer together at higher speeds As already pointed out, the advantage of non-rolling test-rigs with respect to rolling ones is that wheelrail contact forces can be directly measured. Therefore, in the test-rig under each wheel of the instrumented wheelset, a seven component dynamometric balance is positioned having a piece of rail placed on top of it. Besides the three load cells on top of the in-plane hydraulic (longitudinal and lateral) actuators, four load cells should be placed on top of four rods with spherical joints at the extremities that connect the balance to the base of the test-rig.

This configuration of the dynamometric balance allows to measure not only the three contact force components (longitudinal, lateral and vertical) but also the resultant torques (pitch, roll and spin torques) that can be profitably used to determine the wheel-rail contact point position as will be explained later on. Load cells should be chosen to measure forces up to the highest possible value, i.e.

the maximum load applied by the actuators (for the longitudinal and lateral ones) times, eventually, a safety factor, and  one fourth of the sum of maximum load applied by the two vertical actuators plus the weight of the bogie divided by the number of wheels plus the weight of the wheelset divided by two times, eventually, a safety factor. Obviously, due to the fact that also the static force component is of interest, strain gauge load cells should be used. The engineering solutions adopted in the design of the test-rig guarantee the flexibility of use requested by the specifications, in fact:  the rail gauge can range from 600 mm to 1700 mm;  bogies with different architecture and wheelbase can be mounted into the test-rig;  wheelsets with different wheel radii up to 1200 mm can be tested;  the maximum total vertical force which can be applied to the bogie through the secondary suspensions is set to 500 kN;  the maximum lateral force applicable to each plate is set to 50 kN;  the maximum longitudinal force is set equal to two times the lateral one.

3. REALIZATION OF THE TEST BENCH AND SENSOR LAYOUT As described in the previous paragraph, the specifications and requirements of the test rig impose to have quite a bulky structure that is mainly made by a basement that contains the dynamometric balances with the longitudinal and lateral actuators and on top of which the complete bogie with the instrumented wheelset(s) to be calibrated is placed. To be able to apply the vertical loads through hydraulic actuators on the bogie's secondary suspensions, a portal is used. Figure 3 shows a picture of the developed test rig.

Before producing the test rig, stresses and deformations in the test rig structure were investigated through a finite element model in order to verify that no failures nor excessive deformations would occur. Fe350 steel was considered for the structure. Figures 4 and 5 show von Mises stresses (maximum value equal to 55MPa) and the vertical displacement component (maximum value equal to 3mm) respectively when both vertical actuators are applying a force of 250kN. Both values are considered acceptable. As previously described, to obtain a high pass-band, high performance servo-valves were used.

Figure 6 shows the transfer function of the servo-valves, produced by MOOG, adopted for the longitudinal and lateral actuators. It can be seen that the pass-band reaches almost 130Hz with a phase lag of about 60°. Note that the phase lag between the input (reference signal) and the output (generated force/displacement) of the actuators is of no importance. The phase lag that should be considered is the one being the applied displacement and the measured contact force. However, due to the fact that the load cells are placed on top of the actuators and that the actuators are rigidly coupled with the rail piece below the wheel, this phase lag is always equal to zero in the frequency range of interest.

DESIGN OF A NEW FULL SCALE TEST-RIG FOR THE CALIBRATION OF INSTRUMENTED WHEELSETS

Challenge E: Bringing the territories closer together at higher speeds Figure 4. Von Mises stresses in the test rig structure due to the application of 250kN through the vertical actuators Figure 5. Vertical displacement component in the test rig structure due to the application of 250kN through the vertical actuators Figure 6. Transfer function of the servo-valves adopted for the longitudinal and lateral hydraulic actuators (courtesy of MOOG) Figure 7. Transfer function of the servo-valves adopted for the vertical hydraulic actuators (courtesy of MTS) For what concerns the vertical actuators an MTS servo-valve (model 252.25), having pass-band reported in Figure 7, is adopted.

Note that the servo-valve has a pass-band of more than 30Hz. However, as already pointed out, the effective pass-band in vertical direction is much more limited due to the filtering effect of primary and secondary suspensions (where available). Each actuator is equipped with a load cell for force feedback. These load cells are dimensioned on the maximum force the actuator is able to generate: vertical actuators are equipped with MTS load cells having full scale equal to 250kN while longitudinal and lateral actuators are equipped with HBM load cells having full scale equal to 50kN. Finally, the (four) vertical rods of the dynamometric balances are equipped with HBM load cells having full scale equal to 100kN.

Figure 8 shows two pictures of the dynamometric balance adopted. The four instrumented rods below the balance are clearly visible as well as the lateral actuator (with its load cell on top) and one of the two longitudinal actuators (with its load cell on top). Both rods and actuators are connected to the rail piece and to the ground through spherical joints. Thus, the absolute position of the rail piece is fully defined by knowing the lengths of the rods and of the longitudinal and lateral actuators. Challenge E: Bringing the territories closer together at higher speeds Figure 4. Von Mises stresses in the test rig structure due to the application of 250kN through the vertical actuators Figure 5.

Vertical displacement component in the test rig structure due to the application of 250kN through the vertical actuators Figure 6. Transfer function of the servo-valves adopted for the longitudinal and lateral hydraulic actuators (courtesy of MOOG) Figure 7. Transfer function of the servo-valves adopted for the vertical hydraulic actuators (courtesy of MTS) For what concerns the vertical actuators an MTS servo-valve (model 252.25), having pass-band reported in Figure 7, is adopted. Note that the servo-valve has a pass-band of more than 30Hz. However, as already pointed out, the effective pass-band in vertical direction is much more limited due to the filtering effect of primary and secondary suspensions (where available).

Each actuator is equipped with a load cell for force feedback. These load cells are dimensioned on the maximum force the actuator is able to generate: vertical actuators are equipped with MTS load cells having full scale equal to 250kN while longitudinal and lateral actuators are equipped with HBM load cells having full scale equal to 50kN. Finally, the (four) vertical rods of the dynamometric balances are equipped with HBM load cells having full scale equal to 100kN. Figure 8 shows two pictures of the dynamometric balance adopted. The four instrumented rods below the balance are clearly visible as well as the lateral actuator (with its load cell on top) and one of the two longitudinal actuators (with its load cell on top).

Both rods and actuators are connected to the rail piece and to the ground through spherical joints. Thus, the absolute position of the rail piece is fully defined by knowing the lengths of the rods and of the longitudinal and lateral actuators. Challenge E: Bringing the territories closer together at higher speeds Figure 4. Von Mises stresses in the test rig structure due to the application of 250kN through the vertical actuators Figure 5. Vertical displacement component in the test rig structure due to the application of 250kN through the vertical actuators Figure 6. Transfer function of the servo-valves adopted for the longitudinal and lateral hydraulic actuators (courtesy of MOOG) Figure 7.

Transfer function of the servo-valves adopted for the vertical hydraulic actuators (courtesy of MTS) For what concerns the vertical actuators an MTS servo-valve (model 252.25), having pass-band reported in Figure 7, is adopted. Note that the servo-valve has a pass-band of more than 30Hz. However, as already pointed out, the effective pass-band in vertical direction is much more limited due to the filtering effect of primary and secondary suspensions (where available). Each actuator is equipped with a load cell for force feedback. These load cells are dimensioned on the maximum force the actuator is able to generate: vertical actuators are equipped with MTS load cells having full scale equal to 250kN while longitudinal and lateral actuators are equipped with HBM load cells having full scale equal to 50kN.

Finally, the (four) vertical rods of the dynamometric balances are equipped with HBM load cells having full scale equal to 100kN. Figure 8 shows two pictures of the dynamometric balance adopted. The four instrumented rods below the balance are clearly visible as well as the lateral actuator (with its load cell on top) and one of the two longitudinal actuators (with its load cell on top). Both rods and actuators are connected to the rail piece and to the ground through spherical joints. Thus, the absolute position of the rail piece is fully defined by knowing the lengths of the rods and of the longitudinal and lateral actuators.

DESIGN OF A NEW FULL SCALE TEST-RIG FOR THE CALIBRATION OF INSTRUMENTED WHEELSETS

Challenge E: Bringing the territories closer together at higher speeds Figure 8. Picture of one of the two actuated dynamometric balances used to directly measure wheelrail contact forces. Through simple equilibria and assuming negligible friction in the spherical joints that suspend the rail piece, it is possible to determine both wheel-rail contact force components as well as the position of the geometric wheel-rail contact point. Figure 9 shows a scheme of the dynamometric balance with the considered reference system. Fx1 and Fx2, Fy and Fz1, Fz2, Fz3 and Fz4 are the measured longitudinal, lateral and vertical forces respectively while X, Y and Q are the longitudinal, lateral and vertical wheel-rail contact force components.

a1 is the lateral distance between the longitudinal actuators while a2 and a3 are the longitudinal and lateral distance between the vertical actuators. Finally, xc, yc and zc are the wheel-rail contact point coordinates with respect to the adopted reference system.

Fz4 Fx1 yc zc xc Fx2 a1/2 a1/2 Fz2 Fz2 Fy a2/2 a2/2 a3/2 a3/2 Figure 9. Scheme of the dynamometric balance with the considered reference system. Thus, through static equilibrium equations along x, y and z directions, the longitudinal, lateral and vertical wheel-rail contact force components can be determined: 2 1 xi (1) y Y F  (2) Challenge E: Bringing the territories closer together at higher speeds Figure 8. Picture of one of the two actuated dynamometric balances used to directly measure wheelrail contact forces.

Through simple equilibria and assuming negligible friction in the spherical joints that suspend the rail piece, it is possible to determine both wheel-rail contact force components as well as the position of the geometric wheel-rail contact point.

Figure 9 shows a scheme of the dynamometric balance with the considered reference system. Fx1 and Fx2, Fy and Fz1, Fz2, Fz3 and Fz4 are the measured longitudinal, lateral and vertical forces respectively while X, Y and Q are the longitudinal, lateral and vertical wheel-rail contact force components. a1 is the lateral distance between the longitudinal actuators while a2 and a3 are the longitudinal and lateral distance between the vertical actuators. Finally, xc, yc and zc are the wheel-rail contact point coordinates with respect to the adopted reference system.

Fz4 Fx1 yc zc xc Fx2 a1/2 a1/2 Fz2 Fz2 Fy a2/2 a2/2 a3/2 a3/2 Figure 9. Scheme of the dynamometric balance with the considered reference system. Thus, through static equilibrium equations along x, y and z directions, the longitudinal, lateral and vertical wheel-rail contact force components can be determined: 2 1 xi (1) y Y F  (2) Challenge E: Bringing the territories closer together at higher speeds Figure 8. Picture of one of the two actuated dynamometric balances used to directly measure wheelrail contact forces.

Through simple equilibria and assuming negligible friction in the spherical joints that suspend the rail piece, it is possible to determine both wheel-rail contact force components as well as the position of the geometric wheel-rail contact point.

Figure 9 shows a scheme of the dynamometric balance with the considered reference system. Fx1 and Fx2, Fy and Fz1, Fz2, Fz3 and Fz4 are the measured longitudinal, lateral and vertical forces respectively while X, Y and Q are the longitudinal, lateral and vertical wheel-rail contact force components. a1 is the lateral distance between the longitudinal actuators while a2 and a3 are the longitudinal and lateral distance between the vertical actuators. Finally, xc, yc and zc are the wheel-rail contact point coordinates with respect to the adopted reference system.

Fz4 Fx1 yc zc xc Fx2 a1/2 a1/2 Fz2 Fz2 Fy a2/2 a2/2 a3/2 a3/2 Figure 9. Scheme of the dynamometric balance with the considered reference system. Thus, through static equilibrium equations along x, y and z directions, the longitudinal, lateral and vertical wheel-rail contact force components can be determined: 2 1 xi (1) y Y F  (2)

DESIGN OF A NEW FULL SCALE TEST-RIG FOR THE CALIBRATION OF INSTRUMENTED WHEELSETS

Challenge E: Bringing the territories closer together at higher speeds 4 1 zi (3) For the determination of the longitudinal, lateral and vertical wheel-rail geometric contact point position, the static equilibrium equations around x, y and z axis are used: Yx Xy F F ( 4) Qy Yz 5) Qx Xz 6) thus requiring the solution of the following matrix equation: (7) The accuracy of the described methodology to determine the wheel-rail geometric contact point position relies on the accurate measurement of the distances a1, a2 and a3 as well as on the assumption that the measured forces are purely along the corresponding axis.

In case of inclined rods/actuators, in fact, the measured forces are no longer parallel to the axis of the considered reference system (that is fixed to the plate of dynamometric balance supporting the rail piece). Therefore it is necessary to define their inclination with respect to the considered reference axis or, equivalently, the rotation of the shaped plate, to this aim two inclinometers are placed on top of it, whereas the third rotation is determined using the actual displacements of the longitudinal actuators. The above equations become a little more complicated but still wheel-rail contact force components and geometric contact point position can be accurately determined.

Note that each dynamometric balance outputs 16 analogue channels: - 7 forces measured by the seven load cells; - 3 displacements measured by the position sensors integrated into the actuators; - 2 rotations provided by the inclinometers adopted to determine the rotation of the rods/actuators. With regard to the accuracy of the measure of wheel-rail contact forces by means of the dynamometric balance it is possible to give an estimate of the uncertainty associated with the measure. Assuming that each load cell belongs to the same ISO class, having a maximum uncertainty U approximately equal to 0.025% of the full scale and taking into account equations (1), (2) and (3) the standard deviations of the measured wheel-rail contact forces are:   0.018 X kN (8)   0.013 y kN (9)   0.050 Z kN (10) 4.

CALIBRATION EXAMPLE The test-rig previously described has been used for the calibration of the wheelset of a bogie used for metro service. In the following the calibration procedure used to determine the calibration matrix of the instrumented wheelset will be presented together with the results of the calibration process. In order to use the wheelset as a measuring device it is necessary during the calibration phase to correlate the measured strains to the wheel-rail contact forces which generates them. Using the testrig it is possible to apply vertical, lateral and longitudinal forces to the wheels by means of the hydraulic actuators and to measure simultaneously the strains on the axle and on the wheels and also the effective value of the wheel-rail contact force by means of the seven component dynamometric balance.

The calibration process is articulated essentially into four phases. The first phase consists of the definition and the execution of a test schedule, by which it is possible to determine the calibration matrix which relates the inputs (wheel-rail contact forces) of the system (measuring wheelset) to its