Keywords

1 Introduction

The fourth industrial revolution, called Industry 4.0, is based on the digitalization of processes thanks to the “Internet of Things”, so every component of the processes is connected to the Internet and can send and receive information. This revolution is changing not only the ways in which products are designed but also the internal organization of companies and the relationship between them and society [1]. Connecting thousand or millions of devices and processing huge amounts of information requires technologies such as Big Data, Artificial Intelligence, data fusion, etc. In addition, other technologies such as additive manufacturing and Digital Twin are also used in Industry 4.0. The application of these new technologies to traditional predictive maintenance leads to Maintenance 4.0.

The implementation of Maintenance 4.0 in railways is slow and, up to day, focused mainly on the infrastructure. Kaewunruenet al. [2] proposed the use of Building Information Modeling (BIM) to create a Digital Twin embedded in the life cycle of railway infrastructure. It is applied to Taipei Metro to enhance its performance in operation and maintenance. Fuzzy logic is proposed by Karakose [3] for the predictive maintenance of rails, overhead lines and pantographs. The monitoring of track condition through signals collected from multiple trains is proposed by Lederman [4].

The most advanced implementation of Maintenance 4.0 in rolling stock is, probably, the work developed by the company East Japan Railway in the last years [5]. The strategy followed for implementing this paradigm is based on four pillars: condition-based monitoring, asset management, database integration and work supported by artificial intelligence. East Japan Railways is currently operating the series E235 commuter rolling stock, which is able to monitor not only its own systems but also infrastructure systems such as the track and the overhead line.

According to Grieves and Vickers [6], “the Digital Twin is a set of virtual information constructs that fully describes a potential or actual physical manufactured product from the micro atomic level to the macro geometrical level. At its optimum, any information that could be obtained from inspecting a physical manufactured product can be obtained from its Digital Twin”. Although this technology is widely used in product design, life cycle management or logistics, its application to the railway industry is limited and mainly focused on infrastructure or traffic management [7].

This work presents an application case for applying Maintenance 4.0 to a high-speed train designed and built before the advent of the paradigm through the development of a Digital Twin.

2 Digital Twin

The high-speed train studied in this work is composed of two motor cars with two bogies each and 8 articulated cars that rest on 9 trailer bogies. The train opened the first high-speed line in Spain and has been travelling through Spain and the south of France up to 300 km/h since then. We will focus on the trailer bogie of the eighth car which is the nearest to the second motor car. One wheelset of this bogie is equipped with uniaxial accelerometers in the three space directions to measure vibrations.

The bogie’s digital twin presented in this work is developed from technical data to characterise it and develop improvements if needed. The digital twin is composed of three models: a 3D CAD model, a FEM model and a multibody model.

The CAD 3D model is made using the PTC CREO Parametric software. This detailed model (see Fig. 1) includes the bogie frame, two wheelsets, four axle boxes, the brake system, the primary and secondary suspensions, and the car-bogie joints. There are three accelerometers placed in the real bogie for signal acquisition and processing purposes that are also modelled.

The FEM model is developed using ANSYS software from the 3D CAD model. The geometry of key elements such as the bogie frame (see Fig. 2) is imported into ANSYS in order to perform modal analyses and obtain their natural frequencies.

The multibody model is developed with the help of Universal Mechanism software, which includes specific tools for the dynamic simulation of rail vehicles. The multibody model is assembled taking into account the data from the 3D CAD model, paying special attention to the inertial properties of the bodies and the physical relationship between them. Thanks to this multibody model, it is possible to better understand the behaviour of the real bogie under different traffic operating conditions. This model is shown in Fig. 3.

The accelerometers are placed on the axle box, so the performance of the roller bearings is easily accessible. In case a fault in the roller bearings occurs, it should be visible in the spectra. These frequencies are calculated by using the equations proposed by Palmgren [8].

Fig. 1.
figure 1

Bogie 3D CAD model developed in CREO Parametric.

Full size image
Fig. 2.
figure 2

FEM model of the bogie frame in ANSYS.

Full size image
Fig. 3.
figure 3

Multibody model of the bogie in Universal Mechanism.

Full size image

3 Results

Due to the limited space available, only a selection of the characteristic frequencies of the bogie is listed in Table 1. Roller bearing fault and sleeper pass frequencies are computed from theory. Natural modes of the bogie frame, wheelset and axle box are computed by FEM. It should be noted the proximity between the BSF (Ball Spin Frequency) and the sleeper pass frequency, which can lead to mixing up both phenomena and inferring false roller bearing failures.

Table 1. Selection of characteristic frequencies calculated theoretically and using FEM.
Full size table

Irregularities on the track and the wheels can be simulated in the multibody model, trying to represent the actual conditions of the train. For this, harmonic irregularities are introduced both in the two wheels of the front axle of the bogie and on the track. In the wheels, these defects have an amplitude of 0.01 mm and a total of 100 waves along the wheel perimeter. In the track, very short wave defects (wavelength of 0.04 m and amplitude of 0.05 mm) are combined with random irregularities generated according to the indications of the European Rail Research Institute (ERRI) [9].

Random irregularities are created from Eqs. (1) to (3), which define the frequency spectra for the track horizontal irregularities, half the sum of the vertical irregularities and half the difference of the vertical irregularities, respectively. The values of the parameters used are listed in Table 2.

$$\Phi \left(\Omega \right)=\frac{{a}_{h}\cdot {\Omega }_{c}^{2}}{\left({{\Omega }^{2}+\Omega }_{R}^{2}\right)\cdot ({{\Omega }^{2}+\Omega }_{c}^{2})},\Omega >0$$
(1)
$$\Phi \left(\Omega \right)=\frac{{a}_{v}\cdot {\Omega }_{c}^{2}}{\left({{\Omega }^{2}+\Omega }_{R}^{2}\right)\cdot ({{\Omega }^{2}+\Omega }_{c}^{2})},\Omega >0$$
(2)
$$\Phi \left(\Omega \right)=\frac{1}{{b}_{A}^{2}}\cdot \frac{{\Omega }^{2}}{{{\Omega }^{2}+\Omega }_{s}^{2}}\cdot \frac{{a}_{v}\cdot {\Omega }_{c}^{2}}{\left({{\Omega }^{2}+\Omega }_{R}^{2}\right)\cdot ({{\Omega }^{2}+\Omega }_{c}^{2})},\Omega >0$$
(3)
Table 2. Values for generating frequency spectra according to [9].
Full size table

Once the parameters of the irregularities have been established, a simulation of the bogie is carried out under a load of 15 t/axle at a speed of 300 km/h. The accelerations computed by Universal Mechanism in the left end of the front wheelset, in the vertical direction, are recorded. This measurement point is chosen because it corresponds to the location of the accelerometers on the actual train.

In order to compare the simulated and the measured signals, the simulation step is set in a way that the simulated and measured signals have the same sampling frequency. Figure 4 shows the comparison, in frequency terms, of the simulated signal and the actual acceleration measured in the train also at 300 km/h.

Comparing both spectra, it can be seen that the frequency components of the simulated signal have much less noise than those of the real signal. This affects the amplitudes of the main components of the simulated spectrum, which are considerably larger than the ones of the real spectrum. These differences can be minimized by adjusting the model accordingly and always keeping in mind its limitations.

The qualitative analysis of the spectra in Fig. 4 is much more interesting because, if attention is paid to the active zones of the spectra, it can be observed that these zones coincide. In both cases, three active zones are distinguished: 0 Hz–200 Hz, 800 Hz–1,100 Hz and 1,900 Hz–2,200 Hz. By selecting one type of defect in the multibody model, it is possible to study which areas of the spectrum are stimulated by each type of defect and, extrapolating to actual signals, qualitatively determine the possible physical origin of the spectrum components. In this way, it is possible to advance in the detection of the condition of the state of the mechanical system. Specifically, the frequency regions higher than 800 Hz are activated when the very short wave defect into the track is introduced.

Fig. 4.
figure 4

Comparison of power spectra of the simulated and real signals.

Full size image

4 Conclusions

This work presents an application case for applying Maintenance 4.0 to a high-speed train designed and built before the advent of the paradigm through the development of a Digital Twin.

The Digital Twin of the high-speed train bogie has been developed, which consists of three interrelated models: a 3D CAD model developed in CREO Parametric, a finite element model in ANSYS and a multi-body model in Universal. Mechanism. From these models, it is possible to determine the characteristic frequencies of the actual system and identify them in the frequency spectrum of the measured vibratory signals. In addition, they allow studying different operating situations of the real train.

Feedback from the measured data helps to improve the tunning of the multibody model and, therefore, to build up a Digital Twin which can be used within Maintenance 4.0 to identify the probable causes of the frequency components that appear on the spectra and establish the condition of the high-speed bogie. Significant frequency components have been identified that correspond to phenomena such as defects in the track, or the harmonic coupling of the shaft rotation frequency with certain natural frequencies.