MAXBE

Interoperable monitoring, diagnosis and maintenance strategies for axle bearings – MAXBE

Integrating and strengthening the European Area

Co-operative Research Projects

Deliverable 8.1: Technical demonstration of the developed systems and methodologies

Issue no. 0.9

Start date of project: 2012-11-01 Duration: 36 months Organisation name of lead contractor for this deliverable: Nomad Tech Lda

______- 1 -

Revision history Version Date Author Description/Remarks/Reasons for change 0.1 2015-10-16 NOMADTECH First issue (with IMOSS, DL and D2S inputs) 0.2 2015-10-19 NOMADTECH Modifications in section 2.1 0.3 2015-10-19 UPORTO Comments and contributions – section 2.2.3 0.4 2015-10-19 SKF Comments and contributions – section 2.2.4 0.5 2015-10-22 MERMEC Comments and contributions – section 2.1.1 0.6 2015-10-23 MERMEC Comments and contributions – section 2.2.2.3 0.7 2015-10-29 NOMADTECH / NEM Contributions in all sections / Contributions in section 2.1.2 0.8 2015-10-30 ASTS Contributions in section 2.1.1 0.9 2015-10-31 UoB / KRESTOS Contributions in section 2.2.2 and 2.2.3

______- 2 -

INDEX

1. introduction ...... 4 2. Demonstration activities ...... 5 2.1. Room Presentations – Public Workshop ...... 6 2.1.1. On-board and Trackside Common Interfaces and Data Fusion Algorithms ...... 14 2.1.2. Tools for optimal physical distribution of the diagnostic systems and monitoring interval rates ...... 25 2.1.3. Smart Diagnostics and Information Integration ...... 30 2.2. In Site Presentations...... 33 2.2.1. Antwerp, Belgium ...... 34 2.2.2. Estarreja, ...... 45 2.2.3. ...... 60 2.2.4. Depot, Portugal...... 68 3. Conclusion ...... 77

______- 3 -

1. INTRODUCTION

The current document presents the demonstration activities performed within MAXBE project. It provides information related to the public workshop held in Oporto as well as the in loco demonstrations of the MAXBE systems deployed in Portugal, Belgium and United Kingdom.

______- 4 -

2. DEMONSTRATION ACTIVITIES

The demonstration activities were divided into room presentations and in site presentations.

On October 27th there was a public workshop help at University of . The main goal of the workshop was to demonstrate the developments made within the MAXBE project to the rail industry including not only end-users but also manufacturers, other R&D entities and leaders of other European projects.

In site presentations took place in different locations: Antwerp, Belgium; Estarreja, Portugal; Long Marston, United Kingdom and Pendolino Depot, Oporto, Portugal. In these in site presentations, there was the possibility to see “in loco” the systems deployed and in operation mode.

In the following sections, it is shown the demonstration activities within the MAXBE Project.

______- 5 -

2.1. ROOM PRESENTATIONS – PUBLIC WORKSHOP

The public workshop was divided into sessions where all partners had the possibility to present the developments performed within MAXBE. In order to fulfil the main goal of the workshop, several entities/organizations were invited. In the following list, it possible to observe all the invited entities/organizations:

AICEP IVE Cablotec Portugal KRESTOS Steconfer ANI LNEC Univ Minho Ansaldo STS MERMEC LCW Consult APNCF Metro de Lisboa BETAR CARRIS Metro do Porto Ideia.M COMSA Mota-Engil Ferrovias SISCOG CP NEM CARD4B D2S Nomad Tech SUNVIAUTO DL OE EC OERN PoliCabos EMEF Queen's University Belfast Thales ERRAC SKF LNEC EVOLEO SNCF IT FCT Technical University Braunschweig GECAD FERTAGUS TRA 2016 EFACEC Heriot-Watt University UCC INEGI IMOSS UNIFE Amorim INEGI University of Genoa FEUP Infraestruturas de Portugal University of Huddersfield CP ISEP UoB IST ISQ UPORTO Evoleo Tech IST AMBISIG Alma Design

______- 6 -

As stated before the workshop was divided into sessions. In Erro! A origem da referência não foi encontrada. it is shown the agenda of the event.

Figure 1 – Workshop Agenda

Figure 2 – Public Workshop Snapshot

______- 7 -

 Session 1: Axle bearings failure modes, degradation process, laboratory tests, modelling and analysis

Figure 3 – Session 1, speaker Prof. Jorge Seabra (UPORTO)

Figure 4 – Session 1, speaker Ms. Marisa Lima (IST)

______- 8 -

 Session 2: On-board Systems

Figure 5 – Session 2, speakers Ms Elena Pinto and Mr. Ivan Rochford (SKF)

Figure 6 – Session 2, speaker Mr. Carlos Silva (EVOLEO)

______- 9 -

 Session 3: Estarreja Wayside System

Figure 7 – Session 3, Mr. Rodolfo Martins (EVOLEO)

Figure 8 – Session 3, Mr. Francisco Ganhão (IP)

______- 10 -

 Session 4: Integrated LRV wayside monitoring system

Figure 9 – Session 4, Mr. Frederik Vermeulen (IMOSS)

 Session 5: UK Wayside System

Figure 10 – Session 5, Mr. Mayorkinos Papaelias

______- 11 -

 Session 6: On-board and track-side common interfaces and data-fusion algorithms

Figure 11 – Session 6, Mr. Stefano Terrible (ASTS) and Mr. Emanuele Fumeo (UNIGE)

 Session 7: Tools for optimal physical distribution of the diagnostic systems and monitoring interval rates

Figure 12 – Session 7, Mr. Carlos Saborido (COMSA)

______- 12 -

 Session 8: Maintenance Optimizer

Figure 13 – Session 8, Mr. Diarmuid Grimes (UCC)  Session 9: Smart Diagnostics and Information Integration

Figure 14 – Session 9, Mr. Tim Smith (NEM)

In the workshop software demonstrations took place. It was possible to observe software tools developed within MAXBE. In the following sections there is an overview of the developed tools as well as related pictures of the tools.

______- 13 -

2.1.1. On-board and Trackside Common Interfaces and Data Fusion Algorithms

 Introduction

The aim of WP5 (“Integration of systems”) has been the design and development of an integration software platform able to integrate the several wayside and on-board axle bearings monitoring systems using a data fusion algorithm. In order to achieve the WP5 objectives, the research activities have been divided into four main phases, which are resumed below. The first phase, as defined in the Deliverable 5.1 (“Common Interfaces”), consisted in the definition of common interfaces to allow the integration of heterogeneous multi-vendor sensors for both on-board and wayside axle bearings condition monitoring systems. The common representation of the data has been realized through the design of two XML file formats, agreed with the project partners and tailored to the needs of on-board and wayside monitoring systems. All the partners that provided data coming from the installed monitoring systems shared their measurements, alarms and KPIs with the integration software platform by means of these XML files respecting the common agreed format. The common data format has been designed in order to identify univocally both the axle bearing under examination and the kind of measurement or alarm detected by the monitoring system that generated it. This objective has been achieved by identifying the following parameters:

- ; - Kind of wayside or on-board axle bearing condition monitoring system; - Point of Installation of the system (for wayside monitoring systems); - Wagon; - Bearing; - Alarms, measures and KPIs associated to each bearing.

An example of one of these XML files is presented in Figure 15 and Figure 16, which shows how and what information are included in the format.

______- 14 -

Figure 15 – Example of XML file respecting common data format (part 1/2)

______- 15 -

Figure 16 – Example of XML file respecting common data format (part 2/2)

The communication from wayside and on-board monitoring systems to the integration software platform used the File Transfer Protocol (FTP); for this purpose, Ansaldo STS provided a central common FTP server as a storage solution. The second phase, as described in the Deliverable 5.2 (“Data Fusion Algorithms: Design”), referred to the design of the Data Fusion Algorithm. This task has been divided in the following steps:

- Sensing and Signal Processing; - Feature Extraction (including univocal identification of data and measurements related to a particular axle-bearing); - Data Processing; - Data Fusion Results

Starting from the aforementioned general model, the third phase, as described in the Deliverable 5.3 (“Data Fusion Algorithms: Implementation”), implemented the Data Fusion Algorithm into a software tool able to integrate data coming from different heterogeneous sources and to generate high level and high informative patterns to be used for decision- making purposes (such as for maintenance actions planning). In this phase, the functional architectures and the implementation issues related to the development of the Data Fusion Tool have been taken into account.

______- 16 -

The fourth phase, as described in the D5.4 (“Data Fusion Algorithms: Validation”), included the validation of the Data Fusion Algorithm (DFA) and Data Fusion Tool (DFT), which has been performed and completed according to the WP5 objectives and the requirements defined in the previous phases. At the end of these four phases, the integration software platform has been completed. The following section introduces the architecture of the entire developed system.

 Integration Software Platform

The architecture, shown in Figure 17, represents the integration software platform that has been realized in the MAXBE project. It can be divided into three main levels: the “lower” level, called Monitoring Systems Level, includes all the monitoring systems installed along a railway line and on-board train; the “medium” level, named the Storage and Data Management Level, is responsible for the storage and management of data coming from the lower level of the architecture; the “top” level, called the Applications Level, includes all the software technologies developed for data processing and visualization purposes.

Figure 17 – Overall Integration Software Platform architecture

In the MAXBE project, the Application layer is composed by the Data Fusion Tool (DFT) and the Human-Machine Interface (HMI), which represented the main development activities for ASTS and UNIGE.

______- 17 -

 Data Fusion Tool

The Data Fusion Tool is the software tool able to combine data about axle bearing condition coming from both wayside and on-board monitoring systems that implements the Data Fusion algorithm designed in the context of MAXBE project. It is a modular tool (see Figure 18) composed of several logic blocks:

- A “Trigger Interface” and its “Manager”, which provide a solution for triggering the start of the data collection and processing of the data - An “FTP-I/O” module responsible for the communication with input and output FTP servers - An “XML Manager” module which manages the import, the transformation, the treatment of both XML input and output files - A “Local Database”, mainly used for storing historical data and used as a support for data processing - A “Data Fusion” module, which is the core of the application, communicating with the other components (especially with the local database) and processing data through an ad-hoc data fusion algorithm developed for this project

Figure 18 – Data Fusion Tool architecture

______- 18 -

The application has been developed using common programming and database management tools, in particular Microsoft .NET programming languages and Microsoft SQL server. The Data Fusion Algorithm is divided in three different levels (see Erro! A origem da referência não foi encontrada.) aiming at reducing the dimension of data with subsequent levels of processing. The first level performs sensor data fusion and correlates data related to a particular axle bearing which has been collected by similar systems and sensors. The second level performs temporal data fusion, which is essentially a temporal alignment between data. This is important since the two different sides of monitoring systems, namely on-board and wayside systems, are inherently different from the collection rate point of view. It is obvious, but important to point out that an on-board monitoring system can sample data from the sensors continuously, while wayside monitoring systems can only sample data when the train is passing through the monitoring area installed along tracks. Finally, the third level implements a Maximum a Posteriori (MAP) Bayesian approach able to perform the final correlation between data coming from the two different sides of the monitoring systems, which has been processed by the two previous data fusion levels. This last layer essentially generates two final results for each axle bearing:

- “Alarm Severity Level”, which indicates whether the axle bearing presents problems or not; - “Alarm Veracity”, a probability value that expresses a trusting index of the final alarm level generated by the algorithm; in other words, it expresses the likelihood of the final alarm level and it aims at giving to operators an indication of the uncertainty related to the axle bearing condition.

______- 19 -

Figure 19 – Data Fusion Algorithm Implementation

 Human Machine Interface

The “Human Machine Interface” (HMI) is a web-based visualization system that can be accessed through a web browser. It is used to visualize the Data Fusion output in a user- friendly way and it has been designed to support the decision-making process of the maintenance operators. The main panel (see Erro! A origem da referência não foi encontrada.) is divided in the following parts:

- Sites Filter: it is a drop-down menu that allows filtering the Data Fusion Output List by the selected monitoring site; - Last Data Fusion Output: it provides several information about the last Data Fusion Output received by the HMI; - Data Fusion Output List: it is the list of Data Fusion results sorted in a chronological order; - Data Fusion Output Detail: it is specific table that is updated according to the selected Data Fusion Output, showing details about the rolling stocks composing the train; - Graphical panel: it is the graphical visualization of the entire train and related different kind of vehicles (urban, freight, high speed, etc.), the track and the travel direction.

______- 20 -

Figure 20 – Example of the human Machine Interface: Main Panel

______- 21 -

Figure 21 – Pop-up window for assessing axle bearing condition: alarm severity and veracity

______- 22 -

Clicking on “Details” button of the Data Fusion Output Detail panel, a pop-up (see Figure 21) is displayed with a schematic representation of rolling stock; the alarmed axle bearing are displayed in four different colors (green, yellow, orange and red) in function of the alarm level, as well as its related veracity. Finally, an important feature of the system is that it is able to deal with partial data. For instance, since it is highly likely that data coming from the two different monitoring systems could be uploaded to the FTP server at different times, it is important to be sure that in case of the presence of alarms, all the possible information could be presented to the final user, even if partial. For this reason, the system is allowed to override the results previously given in order to integrate the most updated information.

 Demonstration

The final demonstration activities have been done in two different days: the first, during the official MAXBE workshop (on 27th October 2015), based on in-room video presentation in order to show the main functionalities of the Integration Software Platform and the second (on 28th October 2015) based on the installation of the real system into the shelter located in Estarreja demo site (Portugal) – see section 2.2.1 Antwerp, Belgium The demonstration has been performed taking into account the data provided by the axle bearing monitoring systems, described in the next section, installed both in Estarreja site (Portugal) and in Antwerp site (Belgium).

 Data for demonstration

Below, the monitoring systems installed in the Estarreja site (Portugal) are reported:

- Weight in Motion monitoring system (wayside, by EVOLEO) - Hot-box detector system (wayside, by REFER) - Acceleration, acoustic emission, temperature and speed sensors installed onto high speed train (on-board, by SKF)

Regarding the second demonstration site, the wayside monitoring systems (vibration and vehicle identification) have been developed and installed by I-MOSS, D2S and DE LIJN in the Punt-Aan-De-lijn depot in Antwerp (Belgium).

______- 23 -

 Conclusions

In the context of the MAXBE project, an Integration Software Platform based on a Data Fusion Algorithm has been successfully developed. The system aims at providing to the maintenance operators synthetic and synchronized diagnostic information (such as alarms, KPIs, etc.) about the functional condition of the railway assets under examination, and supporting them in the decision-making process. The Integration Software Platform has been able to perform the following functions:

- To collect and store measurements and information coming from the wayside and on-board monitoring systems related to the axle bearing functional conditions; - To correlate and process these data through a Data Fusion Tool (DFT) in order to provide useful aggregated information about the status of each axle bearing under examination; - To visualize through a Human Machine Interface (HMI) events (such as defects, anomalies, failures, etc.) related to axle bearings of different kinds of (passenger, freight, high speed, etc.).

As a final remark, this flexible and scalable solution could be used in the context of condition-based maintenance in order to integrate many other kinds of monitoring and diagnostics sensors, to optimize the selection of technologies and to improve the knowledge about the degradation processes of any train asset (not only axle bearings but also wheels, brake pads, pantograph, etc.).

______- 24 -

2.1.2. Tools for optimal physical distribution of the diagnostic systems and monitoring interval rates

Within MAXBE project, several wayside systems using different types of technology were developed, the most widespread system already existing in several railway networks are the Hot Axle Box Detectors (HABD) and therefore, the guidelines and requirements already available are for this type of devices. In this context, the developed software tool considers the requirements for the installation of HABD, taking into account particularly, the different criterion considered in Portugal, UK, Germany, Belgium and . The tool is a decision-aid support system, which assists the infrastructure manager in the decision of the physical distribution of Wayside Diagnostic Devices (WDD) within the railway network considering their own criteria regarding safety, quality of service and also taking into account the main guidelines for the installation of these devices in the railway network of each country. The tool is available in excel format programmed with Visual Basic and therefore is user-friendly, easy to implement and very flexible in order to be adapted to the end-users needs. At the end, the tool is able to suggest the most adequate places to install a wayside monitoring system, considering a priority list that results from the historical data and also from the infrastructure managers experience and the risk assessment, which is included in the definition of the risk criteria and the importance attributed to each one of the defined indicators.

 Tool Description

The software tool is elaborated in an excel file and programmed with Visual Basic in order to be an easy-to-learn and easy-to-use tool, which can be straightforwardly implemented by potential users, given that it will not require an specific software and due to the familiarity of users with excel sheets. Besides, by constructing the SW tool in excel, it represents a flexible tool which can be easily adapted to satisfy end user’s needs. The main objective of the tool is to help the user to identify the most suitable locations to install wayside devices (WDD) to detect axle bearing failures developed within the project MAXBE, although it can refer to any type wayside device, such as Hot Axle Box Detectors. However, the tool also has other advantages for the user such as a compilation of the requirements demanded by the European norm and other national standards of application in the selected country. The main potential users of the SW tool are Infrastructure Managers, firstly, because they count with the information required to run the SW tool and secondly, because they are usually the ones that decide where to install these systems.

______- 25 -

The methodology to follow to use the software is presented in Figure 22.

Figure 22 – Methodology of the SW tool

This methodology is summarized and described in the following section with the IP (former REFER) use case.

______- 26 -

 Portuguese Test Case

A case-study of the application of the tool in a realistic scenario based on the Portuguese Northern Railway line has been performed. The information is based on realistic data and the knowledge provided by REFER, the Portuguese Railway Infrastructure Manager.

This test case was used to validate and improve the SW tool for the assessment of the optimal physical distribution of wayside diagnostic systems (HBD) and the monitoring intervals.

Figure 23 – Portuguese test case – Input phase

Figure 24 – Parameters to taken into account and weighted for decision support

______- 27 -

 Demonstration

The tools for optimal physical distribution of the diagnostic systems and monitoring interval rates have been presented on session 7 during the MAXBE public workshop.

Figure 25 – Tool presentation at the MAXBE workshop

Figure 26 – Software tool case study demonstration

______- 28 -

Figure 27 – Software tool case study demonstration

Figure 28 – Software tool case study demonstration

______- 29 -

2.1.3. Smart Diagnostics and Information Integration

 Maintenance Schedule Optimisation (MSO) Tool

NEM Solutions was involved in design scope of the tool thanks to its knowledge and experience of working with rolling stock maintenance organisations as well as direct contact through a number of clients. Together with IVE this assisted UCC greatly in the modelling of the maintenance task realisation and the principal factors that contribute to task scheduling. During the second half of the project NEM Solutions has engaged with potential end users and has demonstrated the scheduling tool with two objectives; firstly to contribute to the validation of the tool and the identify opportunities and barriers to industrialisation; and secondly to demonstrate the work achieved within the MAXBE European project. This was done both with the final prototype and user interface, as well as with earlier iterations of the software in order to gain early feedback from potential users. Initial discussion between work package participants and potential end users resulted in the generation of over 25 scenarios or 'use cases' for which there was perceived or real interest in being able to represent using the tool. These were then prioritised and filtered to align with the project objectives, available resources and capabilities. As a result of intermediate project feedback, there were significant corrections and changes in the optimisation tool. Of particular note was the interest of what factors could be modelled using the tool, and its flexibility to cope with different planning constraints and demands.

A selection of the following developed capabilities were demonstrated: - Scheduling of periodic maintenance based on time (days), mileage, or combination of both. - Updating of schedule at any time (e.g. this can be done weekly, daily, or as soon as any known change takes place) - Rescheduling of plan due to introduction of unforeseen maintenance task; either corrective or predictive maintenance (for example from diagnostic system recommendations). - Fully flexible generation of train demand profile: hourly definition of train demand, on each day of the week. - Scheduling including 'one-off' changes to the normal constraints or resources, such as: o reduced human resource (e.g. planned strike, or public holiday), o additional train demand (e.g. due to popular sporting event),

______- 30 -

o reduced depot resource (e.g. depot track out-of-use for civil maintenance work) - Updating of schedule to reflect an overrun of a current maintenance task - Categorisation of staff into two competencies (e.g. mechanical, electrical, or both) - Generation of 'WHAT-IF' scenarios to simulate the change in dimensioning of depot or staff resources, or a future increase7decrease in train demand etc. - Capability to split maintenance between two depots (provided that each defined maintenance exam can only be performed in one specific depot)

Feedback from demonstration sessions included the following:

- The scheduling tool was considered very fast in its execution. This allos the user to run the scheduler at any time when required, to either update the schedule with current or future events, or to run various planning scenarios in quick succession. - The scheduler can quickly demonstrate the organisation's resource usage and the capacity for maintenance throughout the scheduling time window. Amendments can be made to the model to decrease resource availability to account for anticipated corrective maintenance workload. - The tool is flexible enough to adjust input parameters to reflect the situation of the maintenance organisation, as well as adjust the weighting on objective functions, such as meeting train demand, keeping overtime staff hours to a minimum, or satisfying as closely as possible exam due dates. - It was noted that it was not possible to schedule the event in which a maintenance exam is not fully completed, with remaining tasks being performed a following day (after train has spent time in revenue service) [This would be a huge effort to introduce as it would require the definition of individual tasks of an exam and which can be performed separately from each other] - It was not possible to assign staff resource dynamically (as can happen in reality in some organisations), in the situation where staff can be pulled off of one maintenance exam to assist in the completion of another activity which is being performed in parallel on a different train. - It was found that some difference in opinion existed between potential users regarding the merging or substituting of periodic exams when they fall close to one another, and how it should best be modelled. It was concluded that this would have to be analysed on an individual case basis. For example if exam B is due shortly after exam A, how can the benefit (if any) of performing B immediately after A be modelled. This could be in terms of time saved due to reduced train preparation and logistics, or maintenance tasks that do not need to be repeated in both exams.

______- 31 -

 Demonstration

The demonstration of smart diagnostics was made to the owner of the original data, a maintainer of a high-speed fleet of over 20 trains. However, as the data was confidential, the results could not easily be shown to other potential end users. Feedback was very positive and in the future it is anticipated that for the high speed fleet, the smart diagnostic techniques will replace the current rule-based diagnostics used for safety-critical warnings. It is also expected that the diagnostic techniques are used for diagnostics of other systems outside the axle bearing and railway environment, such as for analysis of SCADA data from wind turbines where the reliability of bearings and rotating machinery have a huge influence on overall system performance. The demonstration of information integration in web-based user interfaces has been done on a few end users to gain initial feedback at the beginning of what is an area of research and development work which extends far beyond the scope of the MAXBE project. For full- scale demonstration it will be necessary to work more closely with numerous maintenance organisations to develop full functionality with real maintenance data and information provided through communication links with monitoring systems, maintenance schedulers and asset management systems such as SAP and IBM.

Figure 29 – Maintenance Schedule Optimisation (MSO) Tool Demonstration

______- 32 -

2.2. IN SITE PRESENTATIONS

At the in site presentations, which took place in different locations, it was possible to present the deployed systems to rail industry. During the in site presentations, the systems were demonstrated in operation mode. In the following sections there is an overview of the deployed systems as well as related pictures of the demonstrations.

______- 33 -

2.2.1. Antwerp, Belgium

This section describes the high-frequency vibration-based axle bearing fault detection system jointly developed by De Lijn, D2S, and I-moss and installed in the Antwerp depot of De Lijn.

 Installation Site at Punt Aan De Lijn Antwerp

Punt-Aan-De-Lijn is the northern and bus storage facility of the Flemish public transportation company ‘De Lijn’ at Antwerp. The system is installed at the exit of the facility, indicated on the picture below.

Figure 30 – Entrance Punt-Aan-De-lijn, installation site

The installation on an LRV network has some specific characteristics: embedded track, car and bus traffic, speed and weights different from mainline, etc. The hardware setup chosen in the De Lijn demonstrator takes this requirement into account.

 System specifications

______- 34 -

The complete system is composed of 8 high frequency accelerometers, mounted on the rail feet, field side. All these accelerometer are connected to a processing unit through a cable duct. The Processing unit is located in the vicinity of the accelerometer, in a weatherproof casing. A raw schematic of the setup is given here:

Figure 31 – Installation site

The accelerometers are high frequency and have a sensitivity of 10 mV/g and are equipped with an integrated cable, to protect them from moisture and dirt. A mechanical protection system is installed over the accelerometer for protection and isolating it from other vibrations.

Spacing between the accelerometers is 52 cm, this is approximately one fourth of the wheel circumference. The processing unit is composed of signal conditioning cards, a data acquisition card and an embedded computer. The system also features a GPRS system to allow remote connection. The in-house built signal conditioning cards have been tuned for high frequency signals, since they will be measuring signals of up to 40 kHz.

______- 35 -

At the installation site the rails are embedded for mixed tram and bus traffic, so special precautions are required to ensure a well-functioning and stable system. This includes: underground cable ducts, accelerometers with integrated cabling, mechanical protection systems.

Figure 32 – Wheel dimensions

 Vehicle Identification

loop

decoder

loop

Figure 33 – Vehicle identification through magnetic loop and high-definition camera

All vehicles are identified by a unique number. A magnetic loop reads the vehicle number. As a backup system, also a synchronized picture is taken from a high-definition, high-speed camera.

______- 36 -

 Automated Processing and Fault Criterion

Each wheel’s high-frequency vibration is measured by 4 sensors sampled at 80 kHz. This time signal is enveloped and on the spectra a peak-to-average norm is taken as a basis for anomaly alerting. Based on the measurement database, also trending is possible. Figure 34 shows the raw measured signals for all 4 sensors. Figure 35 shows a detail of the enveloped signal.

Figure 34 – Raw vibration signals of a vehicle passage measured by the system

______- 37 -

Figure 35 – Enveloped signal (detail)

______- 38 -

 Reporting

In the next paragraph we will discuss the interface to the WTMS, which allows for an integrated reporting of the measurements from this individual wayside monitoring system. In this paragraph, we discuss the GUI built-in as a webserver in the vibration-based systems and which allows this system to be used in a stand-alone operation. This will be the case for smaller (e.g. urban) operators or in a degraded mode operation when the integrated system is not functional. The vibration-based system is based on a generic vibration monitoring system developed by I-moss. It continuously monitors all vibration on all sensors. This can be reported for example in a 24-hours overview graph. Each peak corresponds to a vehicle passage. The software allows for a flexible configuration of loggings and alarms. The I-moss Wheel Flat detection and Out-of Roundness Monitoring (WORM) system is also based on this platform and is in integrated in the system presented here.

Figure 36 – IMOSS monitoring system GUI, 12h measurement

______- 39 -

Figure 37 presents the MAXBE GUI developed for the vibration-based system. Each measurement is shown as a single line on the screen, with a schematic drawing of the train. Axle bearings that reach warning levels are shown orange, when they reach alarm levels they are show red. (Note: levels on the screenshots are mock values and are not in relation with actual vehicles). Raw signals of the sensors can be downloaded, and it is possible to listen to the vibration signal by clicking on the loudspeaker icon. The GUI allows also to select only measurements that have found anomalies. It is also possible to view the history of one vehicle and do trend analysis. The data can also be downloaded as CSV for to do custom interpretation in Excel or to develop interface to an automated maintenance management system.

Figure 37 – GUI for vibration-based axle bearing fault detection system

______- 40 -

 WTMS Synchronization

The system stores the processed measurements in an SQL relational database. The two main tables are the “passage” table and the “axle bearing” table. The passage table identifies each measurement of a passing vehicle. It corresponds to the “Measurement_identification” XML file that is exchanged with the MAXBE WTMS server as defined in D5.1. The axle bearing table defines the measurement of an individual bearing during its passage (and is therefore related to the passage table. It corresponds to the “Bearing_event” XML file as sent to the MAXBE WTMS server.

 Demonstration

The system as described above has been demonstrated in Antwerp. Below are some commented pictures of the event.

Figure 38 – Mr. De Donder, De Lijn tram workshop manager, welcomes MAXBE partners to the Antwerp depot for the system demonstration

______- 41 -

Figure 39 – Mr. De Donder introduces the audience to the specificities of the tram running gear, specifically with respect to the axle bearings.

Figure 40 – Dr. Vermeulen, I-moss, explains the sensor setup installed on the track for the purpose of axle bearing monitoring. ______- 42 -

Figure 41 – A Hermelijn vehicle is being measured during pass-by over the system. On the left side of the picture, the cabinet with the data acquisition can be seen.

 Conclusion

The final system is in continuous operation for more than six months at the time of this writing. The installation has thus proven to be robust in a very adverse environment, with braking sand and busses passing from the De Lijn depot. This robustness of sensors is a major advantage over other methods of detection. The wayside electronics containing the data acquisition, processing and transmission to the wayside has been proven in the field for more than 1 year. The system installation at De Lijn depot in Antwerp has been demonstrated to the project partners at the occasion of the M30 consortium meeting.

______- 43 -

______- 44 -

2.2.2. Estarreja, Portugal

The wayside monitoring system installed in Estarreja at the Portuguese Northern Railway line is composed by a vibration system, an acoustic system and a trigger device.

2.2.2.1. Vibration system

This section describes the vibration wayside system jointly developed by UPORTO and EVOLEO and installed in the Portuguese Northern Railway Line managed by IP. A field demonstration was held in October 2015, during a general meeting of the Project organized by Nomad Tech Lda.

 General Description of Installation

The experimental test site in Portugal is located in the Portuguese Railway Northern Line, nearby Estarreja, at PK 291,991 (Figure 42). This location was selected within the Task 2.9 of WP2. In short, the reasons for the selection of this site are: the proximity to a hot box and to a hot wheel detection system; the easy access from people to the equipment installed in the track; the existence of a power supply and a telecommunication interface; the good geometrical conditions of the track in the section. Moreover, in this site it circulates rolling stock from different operators (CP, COMSA), comprising both passenger and freight trains with a wide range of train speed, In the selected test site, the track is ballasted with concrete sleepers and 60E1 rails, and the maximum allowed speed is 220 km/h. Also, at the Estarreja site, the section is in a straight alignment with an insignificant slope and with no interferences nearby, allowing the trains to travel with a constant speed, which are satisfactory conditions to install a condition monitoring device.

______- 45 -

Figure 42 – Location of Estarreja test site

The general configuration of the vibration wayside monitoring system installed at the Estarreja site is presented in Figure 43. This system is composed by strain gages sensors. The raw data acquired from the sensors installed in the track are processed in the data processor unit through processing algorithms that may need to be feed up with information and data from other systems, such as the infrastructure manager data base (for instance, general information about the trains -geometry, references).

Figure 43 – System description

______- 46 -

For the successful detection of wheel defects, sensors are installed along an equivalent wheel perimeter length. In the present case, the instrumented strain gages are placed over a total length of 3.6 m that considers seven sleepers equally spaced in 0.6 m interval. The strain gages were installed at the web of rail for the evaluation of dynamic loads transmitted by the train to the track and have an internal compensation of the variation of the rail temperature. The system has a total of 28 strain gages; they are divided in groups of 12 in the external side of each rail and groups of 2 in the internal side of each rail. The sensors are protected against the railway adverse environment, with dust, ballast, water and all the heavy maintenance activities performed in this type of infrastructure with a robust mechanical system. The cables connecting the sensors to the acquisition system are in underground ducts in order to ensure the safety and the durability of the system considering the maintenance activities. The system configuration allows to weight in motion and to detect wheel defects in a speed range between 5 to 250 km/h, a wheel diameter between 350 and 1000 mm, and it is able to detect and to identify trains up to 300 axles with a range of load between 5 and 400 kN. The identification of each vehicle of the train is carried out with a RFID which is being installed in the Pendulino CPA and EMUs (Electrical Multiple Units) trains that pass at the Estarreja site. This system is composed by tags attached to the train vehicles to be identified. These tags send a signal to the reader system when the carriage crosses over the monitoring system.

Figure 44 – Wayside System – Scheme

______- 47 -

Figure 45 – Vibration wayside monitoring system: installation

The data processor unit responsible for acquiring the data during the passage of a train over the condition monitoring system and for processing the acquired data afterwards based on a data processing algorithm is installed in a cabinet as shown in Figure 46. The data acquisition is performed through a CompactDAQ Chassi from National Instruments, equipped with several modules that are connected to the strain gages sensors.

Figure 46 – Data Processor Unit Overview

______- 48 -

The MAXBE software interface allows the intuitive and easy interaction between the user and the monitoring system through a computer based platform. The front page of the interface allows the identification of the status of the system and the visualization of the information of the most recent event such as the train identification and characteristics and weather information. The software interface also allows the visualization of live data and the historical registers and also to define the settings of the system (Figure 47).

Figure 47 – Software Interface

 Measurements and Results

The wayside condition monitoring system installed in Estarreja is constantly measuring the passenger and freight trains that cross the system, in a 24/7 basis, for about 6 months. In this section, an example of the results registered for some of the train fleets that circulate in the Portuguese railway network are presented. Regarding to the freight trains, the average commercial speed varies between 80 and 120 km/h according with the train characteristics and to the transported material. In what concerns the passenger trains:

- the Portuguese high-speed train (Alfa-Pendular) circulates at around 220 km/h which corresponds to the maximum commercial speed allowed in this Portuguese railway line - the Intercity train circulates between 180 and 200 km/h - the regional and the urban trains of Oporto’s area also pass at the Estarreja site and circulate at around 140 km/h

______- 49 -

The registers with the raw data acquired for the different type of trains are presented in the following figures.

Figure 48 – Results – Freight Train: raw data

Figure 49 – Results – Alfa Pendular: raw data

______- 50 -

Figure 50 – Results – Intercity train: raw data

Figure 51 – Results – UME3400 – Oporto´s urban train: raw data

______- 51 -

2.2.2.2. Trigger device

The aim of this system is to trigger the monitoring devices installed on the track. In the Framework of MaxBe project, its purpose is to trigger and send some needed information to the KRESTOS acoustic wayside system. Regarding the requirement expressed to trigger the KRESTOS wayside monitoring system, the measuring equipment requires being triggered at least 2/3rd of second before train arrival on the measuring area. So, the Mermec system has been installed approximately 80meters away from the measuring equipment on each side of the track. The Mermec triggering system is composed of 2 different kinds of sensors:

- Train detection sensors - Laser barrier

Some equipment are installed in the Electrical cabinet near the track (PC 4U and electronic rack 3U). This cabinet has been installed in a dry and air conditioned area, with 230Vac power supply. The maximum power consumption is 1500 Watts. The Figure 52 presents the triggering system implantation.

Figure 52 – Triggering system implantation

______- 52 -

These sensors have been installed on a commercial track in Estarreja (Portugal). The train detection sensors are used to detect the train arrival and then to count the number of axles. The laser barriers are used to calculate the speed of the train. These sensors are installed on the each side of the track. This is composed of an emitter and a receiver; there are 2 sets on each sensor modules. The speed will calculated using the time and the distance between the 2 sensors of 1 module.

Figure 53 – Laser Sensors Installation

The train detection is done using the redundant sensors, once the train presence has been validated, a relay can be switched for each measuring system. If several monitoring system would be triggered, the electrical signal can be customized for each measuring system, as Mermec will only switch a relay.

Figure 54 – Laser and magnetic sensors

______- 53 -

The information about the speed, and axle count, are sent to the monitoring systems via an Ethernet connexion. The connexion uses UDP protocol and Mermec application is a server, sending messages to all connected clients. The information about speed is updated and sent for each vehicle entering the measurement area. About axle counting, the system is able to count entering and exiting wheels, which means the total number of detected axle, is validated after the train leaves the area. The Ethernet message sent via UDP to all clients contains the needed information about the train (ID, speed, length, axle number…) currently present in measuring area embedded in an xml formatted structure.

Figure 55 – Laser barrier and magnetic sensor

The physical installation of the system on the Estarreja site has been performed in April 2015.

2.2.2.3. Acoustic System

A KRESTOS AE system was installed by UOB in collaboration with colleagues from NOMAD TECH and IP (former REFER) at Estarreja. The system consisted of two AE channels and was interfaced with a magnetic and optical trigger system manufactured by MER MEC. The AE data were sampled at 500 kSamples/s for a duration of 12 seconds. The system was remotely monitored using a 3G mobile connection. The AE signals were amplified through two amplification stages (pre-amplification 40 dB and main stage amplification 7 dB) with at total gain of 47 dB. Two R50A resonant AE sensors manufactured by PAC were attached on opposite sides of the web of the two rails. The photograph in figure 1 shows the installed system at Estarreja.

______- 54 -

Figure 56 – The KRESTOS AE system at Estarreja installed by UOB in collaboration with NOMAD TECH and IP (former REFER).

The screenshot in Figure 57 shows the adjustment control screen for the KRESTOS AE system.

Figure 57 – Adjustment control screen for the KRESTOS AE system

 Conclusion

The final system is in continuous operation and it comprises a vibration system, an acoustic one and a trigger device. The installation as well as the wayside electronics have proven to be robust.

______- 55 -

 Demonstration

The wayside system described above has been demonstrated in 28th October 2015. Below are some pictures of the demonstration.

Figure 58 – Mr. Helder Fonseca explaining Vibration System (left); Vibration System output (right)

Figure 59 – Mr. Helder Fonseca explaining Vibration System

______- 56 -

Figure 60 – Mr. François Defossez explaining Trigger System

Figure 61 – Mr. Guilhem Villemin presenting trigger system output

______- 57 -

During Estarreja wayside demonstration it was possible to see the On-board and Trackside Common Interfaces and Data Fusion Algorithms developed by ASTS and UNIGE. Below there are some pictures illustrating this demonstration action.

Figure 62 – Mr. Stefano Terrible (ASTS) and Mr. Emanuele Fumeo (UNIGE) explaning Common Interfaces and Data Fusion Algorithms

Figure 63 – Mr. Stefano Terrible (ASTS) and Mr. Emanuele Fumeo (UNIGE) explaning Common Interfaces and Data Fusion Algorithms

______- 58 -

Figure 64 – Detail of the PC and the monitor running both Data Fusion Tool and Human Machine Interface

______- 59 -

2.2.3. United Kingdom

Several installations have been carried out in the UK both for short and long term period by UOB and KRESTOS in collaboration with Network Rail. The installations included sites at the Long Marston depot, at Bescot Yard, Wembley West Coast Mainline and Cropredy West Coast line. Cropredy is the main UK site where an integrated AE and vibration system triggered by a treadle whenever a train passes by has been installed. UOB submitted to Network Rail the safety case which was finally approved following rigorous checks by the responsible engineers. Subsequently a Certificate of acceptance has been issued by Network Rail enabling installations to be carried out on the UK rail network (figure 3). The results of the installations are to be reported every 3 months to the Network Rail responsible staff in the form of written reports.

Figure 65 – Certificate of Acceptance issued by Network Rail enabling the installation of the integrated AE-vibration system on the UK network.

______- 60 -

The following photographs show the various sites instrumented.

Figure 66 – Long Marston long term installation at the depot entry

______- 61 -

Figure 67 – Bescot Yard installation monitoring freight wagons

______- 62 -

Figure 68 – Wembley installation. Maximum line speed at the instrumented section is 200 km/h

At Cropredy West Coast Line a 10-channel capable system has been installed. At the moment 4 AE channels, 2 vibration channels and 1 trigger channel are currently being used. The integrated systems has been installed adjacent to the Croperdy Hot Axle Box Detector (HABD) which will be used as reference. Up to 300 trains each day are currently being monitored by the UOB/KRESTOS system, including passenger and freight trains. The line has a maximum speed of 160 km/h. The installation was carried out by a combined team from UOB, KRESTOS and Network Rail.

The schematic in Figure 69 shows the system description.

Figure 69 – Croprerdy AE/vibration system installation by UOB, Krestos and Network Rail

______- 63 -

The photographs in Figure 70 show aspects of the installation.

Figure 70 – Photographs showing the installation of the AE/vibration system and triggering unit in Croperdy.

______- 64 -

The server of the system was installed at the same box where the server of the Croperdy HABD is placed, therefore working in parallel. Data from the HABD can be retrieved upon request to Network Rail for comparison purposes. Furthermore, all train IDs and types are also available via the same route. Figure 71 shows the server unit together with some of the rest of the hardware of the AE/vibration system. The entire system is customised and software has been designed exclusively for this application.

Figure 71 – The AE/vibration system server installed at Cropredy. The system can be monitored remotely via 3G connection.

The photograph in Figure 72 shows the personnel from UOB and KRESTOS discussing the installation details and system capabilities with Mr Patrick Vallely and the rest of the Network Rail team.

Figure 72 – Discussion of the installation details at Cropredy between UOB, KRESTOS and Network Rail personnel

______- 65 -

Various signal processing methodologies have been under development including spectral kurtosis. An example of this analysis is shown in Figure 73.

10

5

0

Amplitude / V / Amplitude -5

-10 0 1 2 3 4 5 6 7 8 9 10 Time / s

200

1000 150

100 500

50 250 Kurtosis/Arbitrary unit Kurtosis/Arbitrary 0 150 200 100 0 2 4 50 0 6 8 10 0 Frequency / kHz Time / s Figure 73 – Raw signal and analysis using spectral kurtosis

______- 66 -

Figure 74 shows the display of the system onsite at Cropredy after one of the trains has been measured.

Figure 74 – System display after measurement of one of the trains.

______- 67 -

2.2.4. Pendolino Depot, Portugal

SKF Insight™ On-Board System

The prototype SKF Insight ™ on-board condition monitoring system developed as part of the MAXBE project is designed specifically for monitoring axle-bearings in passenger rail applications and consists of wireless sensor nodes, an acquisition system and integration into industrial SKF condition monitoring software systems.

 Wireless Sensor Nodes

The wireless sensor nodes are self-powered and designed to be installed inside the axlebox, on the axle bearing seal. Each sensor node is equipped with an array of sensors including vibration, acoustic emission, temperature and speed sensors. The specific measurements taken by the wireless sensor nodes are:

- SKF Acceleration Enveloping Band 3 (vibration) - SKF Acoustic Emission Enveloping (acoustic emission) - Temperature - Speed

These measurements were selected due to their proven capabilities in detecting bearing damage and damaging bearing conditions. Measurement intervals, sampling frequency and sample length are all user-configurable and the measurement data once acquired is transmitted wirelessly back to the acquisition system.

Figure 75 – Sensor mounted on a bearing seal (left) and visible during installation on the CP Alfa Pendular (right)

______- 68 -

Figure 76 – View of the sensor installation once completed

 Acquisition System

The sensor nodes by their nature do not require a wired infrastructure to transmit data but rather take advantage of wireless mesh network technology to facilitate transmission. Data is sent to a wireless gateway that is mounted on the underside of the carriage. This gateway is connected to a PC housed in a control cabinet on board the carriage. SKF Insight™ software processed and buffers the data prior to its transmission via a 3G router to a remote database.

Figure 77 – The PC and UPS system on board the CP Alfa Pendular train

______- 69 -

Figure 78 – System Architecture

 Condition Monitoring Software

Once the data has been imported into the database it is monitored by SKF @ptitude Observer software for any indications of bearing damage, damaging bearing condition or any anomalous activity. This software facilitates both automated monitoring and in-depth analysis. In addition to SKF monitoring software, has been developed a web application for demonstration purposes as a tool for rolling stock maintenance.

______- 70 -

Figure 79 – Global View in SKF web application

This application shows, in a schematic way, important data to understand the health state of the monitored components. The health of each component is shown according to alarm settings made in SKF @ptitude Observer.

Figure 80 – Alfa Pendular view in SKF web application

______- 71 -

Some functionalities such as GPS location, lubrication or kilometres counter are disabled for this project but shown as an example of possible capabilities of the system. In the application trend KPI data can be seen for each component: wheel, bearing rollers, bearing outer ring and bearing inner ring. In addition trends for bearing temperature and train speed are presented.

Figure 81 – Axlebox view in SKF web application

______- 72 -

 Installation on CP Alfa Pendular

In conjunction with Nomad Tech, a prototype SKF Insight™ system was installed on an operational CP Alfa Pendular train. Four axleboxes on a single carriage were instrumented with wireless sensor nodes. Data was collected over the course of various trials with Nomad Tech resulting in a significant longitudinal corpus of data. The probability of damage detection in the trials, given only four wheels were instrumented, was low. Given the optimal location of the sensors in relation to the axle bearings, coupled with extensive experience in rail applications, it is expected that bearing damage or damaging activity (if present) would be clearly detectable. Activity from one axle bearing was identified as showing symptoms of bearing damage with increases in trend data combined with harmonic and side-band activity at approximate bearing frequencies in both acoustic emission and vibration. In-depth analysis has identified that the activity is most likely due to electrical discharge at the location of very minor localised damage in the bearing. This bearing remains the subject of ongoing monitoring by SKF and is thought to be at significantly higher risk of developing further bearing damage.

 Contribution to data-fusion algorithm

Following the same criteria as for alarm settings in SKF software and KPIs defined for the data-fusion tool, a Matlab script has been developed to extract data from the SKF database, and to convert to .xml format for the data-fusion tool. According to KPIs specification, a set of data coming from the train have been sent to Ansaldo STS to be integrated in the data-fusion tool.

Figure 82 – Example of generated data

______- 73 -

Figure 83 – XML data generation for data fusion algorithm

 Demonstration

The system described above has been demonstrated in 28th October 2015. Below are some pictures of the demonstration.

Figure 84 – Demonstration at Pendolino workshop

______- 74 -

Figure 85 – Mr. Ivan Rochford presenting onboard system

Figure 86 – Ms. Elena Pinto presenting onboard system

______- 75 -

Figure 87 – Mr. Ivan Rochford during onboard system demonstration

______- 76 -

3. CONCLUSION

The main goal of WP8 – Demonstration was achieved. During the actions undertaken within WP8, there was a close contact with rail industry which gave the opportunity to present all the developments performed during MAXBE project.

In all demonstration actions the systems were successful demonstrated showing the viability, technical feasibility and interoperability of the MAXBE.

______- 77 -