Final Report Summary - MAXBE (INTEROPERABLE MONITORING, DIAGNOSIS AND MAINTENANCE STRATEGIES FOR AXLE BEARINGS)
Executive Summary:
Axle bearing damage initiation, evolution and possible catastrophic failures can cause severe disruptions or even dangerous derailments, potentially causing loss of human life and leading to significant costs for railway infrastructure managers and rolling stock operators. Consequently the axle bearing damage process has safety and economic implications on the exploitation of railways systems that is why it has been the object of intense attention by railway authorities as proved by the selection of this topic by the European Commission in calls for research proposals. Currently, the most used technique for monitoring the axle bearing condition is by measuring and controlling the in-service temperature of rolling stock axle boxes. However the more accurate methods for the bearing health monitoring are by vibration, acoustic or grease analysis
The MAXBE Project (www.maxbeproject.eu) a EU-funded project, appears in this context and its main goal is to develop and to demonstrate novel efficient technologies which can be used for the on-board and wayside condition monitoring of axle bearings. The MAXBE project focuses on detecting axle bearing failure modes at an early stage by combining new with existing monitoring techniques and on characterizing the axle bearing degradation process.
Demonstration of the developed condition monitoring systems was performed in Portugal in the Northern Railway Line with freight and passenger traffic with a maximum speed of 220 km/h, in Belgium in a tram line and in the UK. The monitoring systems consider strain gauges, high frequency accelerometers, temperature sensors and acoustic emission. Both on-board and wayside systems are considered in the project because there is a need for defining the requirement for the on-board equipments and the range of working temperatures of the axle bearing for the wayside systems. Also extensive laboratory tests were performed by the University of Porto to correlate the situ measurements to the status of the axle bearing life. To get a robust technology to support the decision making of the responsible stakeholders synchronized measurements from on-board and wayside monitoring systems will be integrated into a platform. With the MAXBE project concept it will be possible: to contribute to detecting at an early stage axle bearing failures; to create conditions for the operational and technical integration of axle bearing monitoring and maintenance in different European railway networks; to contribute to the standardization of the requirements for the axle bearing monitoring, diagnosis and maintenance. The successful implementation of the project’s key deliverables have positive impact on the the reliability, availability, maintainability and safety of rolling stock and infrastructure with main focus on axle bearing health.
The consortium for the MAXBE project comprises 18 partners from 8 member states, representing operators, railway administrations, axle bearing manufactures, key players in the railway community and experts in the field of monitoring, maintenance and rolling stock. The University of Porto coordinated this research project that kicked-off in November 2012 and was successfully completed on October 2015.
Project Context and Objectives:
The safety of rolling stock and the economic implications of rolling stock maintenance have been of significant concern to the railway industry. In the last few years the maintenance strategy paradigm is significantly changing and the longstanding preventive time based maintenance is being progressively replaced by condition based maintenance. Also the early diagnosis of the rolling stock condition will allow the improvement of the definition of the maintenance strategies for railway vehicles. Condition monitoring technologies offer to the railway operators means to increase reliability and safety. By an earlier detection of potential failures, the railway operators can better plan maintenance actions contributing to achieve financial savings.
The axle bearing damage process and the consequent failures can cause severe delays or even dangerous derailments implicating human lives prejudice and significant costs for railway managers and operators. The usual causes for axle bearing failures are: the ageing and the deterioration of grease; the oil separation from the grease structure; the particle contamination of the lubricant grease due to wear from components; the external particle contamination due to damaged seals and out of balance axle loads. When the axle bearing temperature rises, it can be indicative of possible failure, loss of internal kinematics, lack of lubrication and overloading, factors that can dramatically reduce bearing life.
In railways, the axle bearings condition monitoring can be made directly (on-board) or indirectly (on-track and on-laboratory). Regarding the wayside condition monitoring, hot axle box detectors are the most common track side monitoring systems and are used for the detection of faulty axle bearings in-service. However, these systems cannot detect damage in early stage, because when the alarm temperature is registered the axle bearing is already in an advanced state of degradation. Also, with this type of monitoring systems, overheated bearings may not be detected and false alarms can be triggered during operational conditions. Therefore, more accurate methods for the bearing health monitoring need to be developed considering vibration and acoustic analysis. But these techniques are complex and their application to the railway industry is limited. The on-board condition monitoring systems implies that different types of sensors, generally including temperature, vibration and/or acoustic emission sensors, are installed directly in the axle box, consequently, these systems are more likely to detect an axle bearing fault, especially at an early stage of evolution. But on-board systems may have technical limitations related with power supply, communication system and adaptability of the systems to different type of trains as well as data processing tools that can improve the reliability and effectiveness of these systems.
The goal of the MAXBE project is to provide validated and demonstrated concepts, strategies and guidelines for interoperable axle bearing monitoring and diagnosis. These concepts and guidelines will then support the railway operators and managers that deal with the threats imposed by the existence of axle bearing damage.
This aim is achieved by carrying out the following specific objectives:
- to identify and compare current technologies, commercial systems in operation for monitoring the axle bearing condition, both on-board and track side;
- to identify the requirements and the most suitable indicators for early detection of the axle bearing condition and a combination of COTS sensors for axle bearing monitoring;
- to develop/propose diagnosis methodologies and strategies for the axle bearing condition based on on-board and track side systems;
- to test and benchmark the developed methodologies;
- to develop a condition-based maintenance models for axle bearings;
- to demonstrate the developed methodologies;
- to propose guidelines and protocols for implementing in the European network diagnostic systems for axle bearing condition;
- to disseminate the project results.
To accomplish the above objectives, the MAXBE project comprises nine Work Packages (WP) in which it is proposed: to develop wayside and on-board monitoring systems for axle bearings; to develop an integration platform for the different types of monitoring systems; to develop tools for the maintenance of axle bearings.
Some wayside monitoring systems were developed in the project based on diverse approaches: strain gauges, high frequency accelerometers and acoustic emission. Two on-board monitoring systems composed by temperature sensors, accelerometers and acoustic emission were also build up within the project. To get a robust technology to support the decision making of the responsible stakeholders synchronized measurements from the on-board and the wayside monitoring systems were integrated into a platform. The developed wayside condition monitoring systems were tested in several locations in Europe including the Northern Railway Line in Portugal with freight and passenger traffic with a maximum speed of 220 km/h, De Lijn’s tram line in Antwerp - Belgium and Long Marston, Bescot Depot, Cropredy and Wembly in the UK. The on-board monitoring systems were tested in a train that circulate in the Portuguese Northern Railway Line. The condition monitoring aspects of the research have been complemented through extensive laboratory tests and multibody dynamic modelling used to correlate the in-situ measurements with the status of the axle bearing life. Still within the project, a condition-based maintenance model for axle bearings was developed as well as a smart diagnostics for an early detection of faults and a software tool for the optimal physical distribution of the diagnostic systems.
The detailed description of the Work Packages can be found in the website of the project: www.maxbeproject.eu.
The MAXBE consortium includes 18 partners from 8 European member states: University of Porto (Portugal); REFER (Portugal); ASTS (Italy); UNIGE (Italy); IVE (Germany); COMSA (Spain); University College Cork (Ireland); EVOLEO (Portugal); Nuevas Estrategias de Mantenimiento NEM (Spain); MERMEC (Italy); SKF (Spain); Instituto Superior Técnico (Portugal); Dynamics, Structures and Systems International (Belgium); Vlaamse Vervoersmaatschappij De Lijn (Belgium); NOMADTech (Portugal), I-moss (Belgium); KRESTOS (UK); University of Birmingham (UK).
Project Results:
1 Axle Bearings Failure Modes and Degradation Process
Axle bearing fault initiation can be caused by various factors including friction, wheels flats, poor alignment, poor lubrication, etc. Depending on the factor causing the initiation of the axle bearing fault different, failure types may occur (race or roller bearing related failures, etc.).
Based on data provided from the operators (for COMSA, DL and EMEF) it has been established that the maintenance cycle is kilometre-driven. The bearings provided from COMSA and EMEF are lubricated with grease (Shell Alvania Grease 2760) and the bearings from DL are lubricated with oil (Total Carter EP 220).
Deterioration of the structural integrity of axle bearings can be detected using temperature, vibration and low and high-frequency acoustic emission measurements using sensors installed either on each axle bearing or wayside.
As part of Task 2.1 an interactive and friendly-user database in Microsoft Excel (optimise for version 2010) has been created, A screenshot of the developed database is shown in Figure 1.
Figure 1 – Screenshot of the interactive database developed in Excel.
2 Assessment of existing wayside equipment and on-board systems and operational requirements
In the project, information from the partner’s expertise domain has been used to assess the current state-of-the-art and evolutions. Where available, the installation and operational cost of the equipment has been taken into account. A comparison of on-board and wayside systems has been carried out, also w.r.t. financial viability. One should also note as a side constraint that all systems should be as easy as possible to install and require minimum maintenance during their operation.
3 Laboratory tests
The analysis of operating lubricants allows knowing the state of the machine and following the evolution of any change related to wear of the various lubricated elements, and detecting the changes of physical and chemical characteristics that occur in the lubricant.
Analysis was carried out under laboratory conditions by comparing samples of fresh, used and artificially aged (mechanically and thermally) lubricants of oil from a PCC Antwerpen unit and grease from an UME 3400 unit, using Rheological Characterisation, Direct Reading Ferrograph (DRIII), Analytical Ferrograph (FMIII), Particle Counting (ISO 4406:99; NAS 1638, SAE AS 4059), VIBRO Viscometer (SV-10) and Fourier Transform Infrared Spectroscopy (FTIR).
Rolling bearing surface analysis was performed through visual inspection, microscopic surface analysis and roughness measurements analysis on axle bearing elements (inner race, outer race, cage and rollers) from a PCC Antwerpen unit.
Friction torque tests on rolling bearings were performed and compared for samples of fresh, used and artificially mechanically aged lubricants of oil from a PCC Antwerpen unit and grease from an UME 3400 unit. An experimental setup was used to perform vibration measurements. The experimental setup consists of a test rig, which supports a DC servomotor with a speed controller, a shaft drive, a test bearing housing which accommodates an axle bearing and a hydraulic load applicator.Vibration data were collected for two different load conditions: unloaded and loaded.
Considering the importance of the laboratory work developed in task WP2 to the overall results of the MAXBE project, UPORTO decided to continue the laboratory tests and its analysis in collaboration with the partners interested in providing damaged axle bearings and oil samples. The results attained were submitted in one annex attached to the D2.4 in the website of the MAXBE project, in month 24. In this extended work an exhaustive analysis of the failure modes and the associated degradation process of several types of axle bearings was performed. Several techniques were applied in axle bearings from PCC Antwerpen unit (tram that circulates in Antwerp) and UME 3400 unit (Urban train of Porto), such as Ferrography analysis, viscosity analysis and surface analysis (visual inspection, microscopic surface analysis and roughness measurements). Regarding the rolling bearing surface analysis, two used axle tapper roller bearings removed from a PCC Antwerpen unit were analysed in order to identify different types of surface damage in the inner raceways, outer raceways and rollers that could lead to potential bearing failure. Furthermore, rolling bearing tests were performed with a modified Four-Ball Machine, where both types of axle bearings (PCC Antwerp and UME3400) were analysed considering the specific load conditions and lubricants. Vibration analysis in a faultless axle bearing from the UME3400 unit was also performed considering two different load conditions. This additional task allowed to gather very sophisticated information of laboratory tests, with very accurate numerical models and state-of-the art condition monitoring systems based on diverse technologies of monitoring. An annex was attached to the initial document of D2.4 on month 24, and it is available in the MAXBE project website.
Other type of tests were carried out by UoB on healthy and defective bearings using a customised test rig with capable of rotating sample bearing from 100 up to 1000 Revolutions Per Minute (RPM). An R50A resonant AE piezoelectric sensor procured from Physical Acoustics Corporation (PAC) was mounted on top of the bearing case using a magnetic hold-down. A universal general purpose accelerometer was also used to replace the acoustic emission sensor in each test. The bearing samples used in the laboratory rig tests were PFI Inc, model PW29530037CSHD Ford wheel bearing with dimensions of 28 x 53 x 37 mm. These bearings were disassembled; defects were induced and put back on the rig and include a healthy, outer race and roller defective.
The overall results from the vibration analysis shows that both outer race and roller defects have significantly higher peak to peak and RMS values compared to the healthy one. On the other hand, although Crest Factor confirms the defective bearings, it also shows a high value for the healthy bearing. Therefore Crest Factor does not seem to be a suitable method to analyse the vibration data. Kurtosis rises from healthy to the defective bearing which was expected. The overall values for the AE measurement are similar to the vibration measurement. Increasing in peak to peak and RMS values can be observed however Crest factor and Kurtosis do not provide sensible values that can describe the condition of the bearings. The final information regarding the laboratory tests performed in UK was submitted in D9.1 within WP9 in month 36.
4 On-board systems
Two on-board monitoring systems composed by temperature sensors, accelerometers and acoustic emission were a build-up within the project.
The first system corresponds to a monitoring system with wireless communication technology and Acoustic Emission capability, INSIGHT TM from SKF. Due to its structure this system is dedicated to the bogies of the Pendolino train, Pendulino CPA4000. This prototype system installed on the CP Alfa Pendular train was tested in two phases; the first phase ran from November 2013 – January 2014; and the second phase ran from February 2015 – November 2015. The first phase of testing highlighted key areas for improvement including the low power operation of the wireless sensor nodes and stability of the wireless mesh network. The second phase of testing provided the opportunity to validate these design improvements. During the course of testing provision was made for integration of the data into the Ansaldo STS data fusion, facilitating cross-verification of the on-board SKF Insight™ system with the wayside systems installed at the Estarreja site in Portugal.
The second designed on-board monitoring system is composed by a set of sensors installed in the axle box, a GPS and an acquisition system connected with a data processor unit. It aims to measure the temperature and the vibration levels in three distinct directions. A general overview of the on-board monitoring system configuration is presented in Figure 2. The acquisition system receives the raw data from the sensors and stores it into an on-board CPU. The data management, such as the correlation of the data from the sensors with the GPS information, the data processing, is also performed in the on-board unit. A communication module allows the access to the raw and processed data, enables the remote access to the system and monitors the state of the equipment.
The raw data or the processed data converted into key performance indicators (KPIs) are then transferred to an on-shore server, which enables the access of the data in the data fusion algorithms and feeds the smart diagnostic tool. Considering the benefits of the correlation of on-board and wayside measurements, a radio-frequency identification system (RFID) was considered and installed in Alfa-Pendular trains that cross the Estarreja test site. The RFID system is composed by the tags installed in the trains which send a signal to the reader system when the train crosses the wayside monitoring system.
Figure 2 - General overview of the on-board monitoring system in Alfa-Pendular train (CPA4000)
5 Wayside monitoring systems
5.1 Vibration monitoring system
The vibration wayside monitoring system is installed in the Portuguese Northern Railway Line at Estarreja and it is composed by a set of sensors (strain gauges) installed in the track, connected to an acquisition module and to a data processor module. Sensors are installed at the web of rail along an equivalent wheel perimeter length. The system has a total of 28 strain gauges, divided in groups: 12 sensors installed in the external side of each rail and 2 sensors in the internal side of each rail. The installation scheme is presented in Figure 3.
Figure 3 - Wayside system: (a) installation scheme; (b) location of Estarreja test site
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.
In Figure 4, the acquisition system, the data processor unit and the communication system installed in a cabinet in the side of the track is shown, as well as the installation of the sensors in the track.
Figure 4 - Wayside system in Portugal: data processor unit and system installation in the track
The raw data acquired from the sensors installed in the track is processed in the data processor unit through the data processing algorithms briefly described in Figure 5 that includes information from the infrastructure manager data base (for instance general information about the trains - geometry, references). The settings of the data processor unit, as the definition of the KPIs, trending limits and sensitivity can be remotely defined/modified.
Figure 5 - Data processing algorithms
5.2 High Frequency Vibration monitoring system for Light-Rail Vehicle axle bearing monitoring
A high-frequency vibration-based axle bearing fault detection system jointly developed by De Lijn, D2S, and I-moss is installed in the Antwerp depot of De Lijn, the Flemish public transportation company. The system is installed on the exit track of the facility, monitoring all vehicles exiting from the depot onto the Antwerp tram network. Figure 6 shows some pictures of the wayside installation in Antwerp.
The installation on an LRV (Light Rail Vehicle) 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 these requirements into account.
The complete system is composed of 8 high frequency accelerometers, mounted on the rail feet, field side. The accelerometers are connected to a processing unit, located in the vicinity of the accelerometer in a weatherproof casing. The accelerometers are specified for high frequency vibration measurement and have a sensitivity of 10 mV/g.A mechanical protection system is installed over the accelerometers for protection against moisture, dirt and EMC and isolating them from external vibration sources. Spacing between the accelerometers is 52 cm, 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 3G connection to allow data upload to a centralized measurement database. The in-house built signal conditioning cards have been tuned for high frequency signals, measuring vibration signatures of up to 40 kHz.
All vehicles are identified by a unique number. A magnetic loop reads the vehicle number. As a second identification system, this prototype also contains a high-definition, high-speed camera providing a picture of each vehicle, that allows to decode its identification number optically with OCR software (Optical Character Recognition). This solution is promising for mainline installation.
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 central measurement database, also trending is possible.
The installation has been in continuous operation for more than two years now and has proven to be robust in a very adverse environment, with braking sand and busses passing from the De Lijn depot. This robustness of the installation is a major advantage of this type of sensors over other methods of detection.
Figure 6- Wayside axle bearing monitoring system for Light Rail Vehicle
5.3 Acoustic emission monitoring system
Another customized wayside monitoring system is based on a set of acoustic emission (AE) sensors mounted on the rails, connected to a data acquisition unit and a data processor module installed trackside. The data were recorded after automatically triggering the system to acquire as the train neared the instrumented section of the track.
Wheel and rail form a direct mechanical path for the acoustic emission signals produced from the bearing to be transmitted to the AE sensors mounted on the rail i.e. the detection zone. Detecting bearing defect signals transmitted via this direct mechanical path provides more accurate results compared to airborne acoustic detection, because it eliminates adverse effects of surrounding noises and other environmental parameters, such as wind and aerodynamic forces. Moreover, the signals acquired contain high frequency information which makes detection of faults more likely.
Figure 7 shows the installation outline of the wayside tests carried out in Long Marston, UK. The tank freight wagon containing the three faulty bearings was towed by a locomotive as shown in Figure 8. The faulty bearings were only at the same side of the second, third and fourth wheelsets of the freight wagon with 2, 4 and 8 mm roller defects, respectively. Tests were carried out at up to a maximum speed of 48 km/h over a straight section of welded track approximately 1000 meters in length. The sampling rate for AE signals was 500 kHz and the duration of the acquisition was set at 12 seconds.
Figure 7 - Simplified outline of the wayside installation configuration
Figure 8 - The Long Marston testing configuration. The yellow locomotive pulling the test freight wagon can be seen in the back
An optical unit capable of measuring the speed of the train and counting the number of wheelsets was employed in order to correlate AE signals with the position of the wheels. The optical unit was also used to trigger the data acquisition unit to acquire data while the train was passing through the detection zone. As the system was also counting the number of axle boxes, it is possible to truncate the AE signal exactly at the time that each wheelset passes through the detection zone. This method saves a vast amount of data as the sampling rate of 500 KHz needs a large volume of physical memory. This also makes the analysis period shorter.
6 Wayside and On-board interface software
In the context of railway asset management, one of the most innovative research areas is the multi-sensor Data Fusion; for this reasons, an innovative software Integration Platform based on Data Fusion approach for axle bearing condition monitoring has been developed. It has been performed by three main modules, i.e. a Data Fusion Application (DFA), a System Integrator (SI) and a visualization system referred as Human Machine Interface (HMI), which are interconnected by interfaces that allow communications through messages in eXtensible Markup Language (XML) format, as shown in Figure 9.
Figure 9 - MAXBE Integration Platform – architecture
The DFA gathers all the data at sensor level and combines them by subsequent processing; it extracts high-level patterns through a probabilistic Bayesian methodology which is able to exploit data to output both: 1) an accurate estimation of the occurrence probability of current and/or incoming potential failures; 2) a confidence interval of this risk, useful for decision makers to assess the reliability of the extracted pattern.
The SI manages the interaction between different software technologies in order to integrate every module into a single system capable of collecting data, performing the data fusion in order to extract new information and visualizing the results of this process in a user-friendly manner through the HMI. It updates asset related information disseminated to all the parts of the system as soon as new data is available or new data fusion processing has been completed, so that, for example, the visualization system always shows the most updated information possible.
The HMI (shown in Figure 10) is a web-based visualisation system that allows visualising data and information generated by the DFA, such as potential alarms, in a user-friendly graphical way. The most important characteristics of such an interface are the clearance, the simplicity and ease of use. The visualization system includes pop-out elements that highlight the presence of dangerous situations, and consequently it is able to explore data related to the particular phenomena in order to have, if requested, a more precise view of the situation under examination.
Figure 10 - Example of HMI
The entire software Integration Platform based on Data Fusion approach has been inspired by the JDL generic model, of which it represents a possible implementation for condition monitoring applications.
The system has been able to perform the following functionalities:
• to collect and store measurements and data coming from the WTMSs and OTMSs related to the axle bearing functional conditions;
• to correlate and process these data through a Data Fusion Application (DFA) 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 trains (passenger, freight, high speed, etc.);
• to highlight critical asset condition in order to allow maintenance operators to take proper actions, support decision making by giving a synthetic final alarm level and an associated value (“Veracity”) as the likelihood of the generated alarm level.
The tests and validation of the DFA, and consequently of the algorithm, has been performed and completed with laboratory tests, showing that the system gives satisfying results and fulfils all the requirements. The results show benefits in terms of maintenance costs reduction, risk management and safety, and the system proves to have a positive impact all over the railways systems. Future steps include the realization of a real demonstrator by collecting and processing real data coming from the test sites in Estarreja (Portugal) and in Antwerp (Belgium), officially scheduled for the final phase of the MAXBE project.
7 Tool for optimal physical distribution of the monitoring systems
The monitoring systems are expensive and the investment and the maintenance of this type of systems have to be taken into account by the Infrastructure Managers. Moreover, since there is no available tool to help deciding the distribution layout of the monitoring systems in a railway network within the MAXBE project a tool to identify the optimized physical distribution of the condition monitoring systems and its monitoring interval rates is developed.
The tool considers historical and statistical data of the railway network and particularly data from the railway line in analysis. The developed tool takes into account the risk associated to certain indicators and the importance assigned to each one of the pre-defined indicators, in order to be able to identify the most critical aspects regarding the axle bearing failure in a railway network.
Although within MAXBE project, several wayside systems using different types of technology were developed, the most widespread system already existing in several railway networks is still 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, United Kingdom, Germany, Belgium and France. 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 methodology of the software tool is indicated in Figure 11.
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, by weighting the several risk assessment indicators defined in the software tool.
Figure 11 - Methodology of the software tool.
8 Maintenance tool
The increase in condition-based maintenance (CBM) of assets such as axle bearings should have a number of positive benefits for the train operating company, such as a reduction in in-service failures, a reduction in maintenance costs as maintenance for certain components can be moved from periodic planned maintenance to condition-based, increase in the lifetime of the assets, etc. However, this also creates certain challenges for the maintenance operating company, as there is greater uncertainty regarding what depot resources will be needed for performing maintenance and when they will be needed. Indeed there is already significant uncertainty regarding the duration of planned maintenance exams, as the exact maintenance requirements are often only known after performing a manual inspection in the depot.
The maintenance schedule optimizer software tool that has been developed within the MAXBE project was implemented such that it can be deployed in a number of different maintenance depot scenarios, e.g. single versus multiple fleets, passenger versus freight fleets, with or without overtime staff, periodic maintenance with kilometric or time-based limits, etc. The tool considers time-variable fleet demand profiles, periodic exam due dates, resource/staff depot capacity constraints at an hourly level, over a fixed rolling horizon; while also incorporating proactive and reactive methods for handling the uncertainty inherent in the maintenance scheduling problem.
The problem is solved using dedicated Constraint Programming methods, and the optimizer is implemented in Python in Google’s or-tools solver, which is a free, open-source set of operations research tools. All accompanying software are implemented in Python, including a graphical user interface for the solver and a tool for automated distribution fitting of maintenance exam durations based on historical data, and are open-source and freely available. While the tool can primarily be used for scheduling of daily activities in a rolling horizon capacity, it can also be used to identify bottlenecks in the depot specification. Cost-analysis of alternative depot setups (e.g. increasing or decreasing tracks/resources, altering staff shifts) can then be performed through simulation experiments using the tool.
9 Recommendations and conclusions
Through the MAXBE results and the consultation with potential end users, a number of recommendations can be made.
• Axle bearing monitoring should be seen as an extension to Wheel Flat and Out-of-Roundness Monitoring systems.
• The monitoring of the axle bearing requires the knowledge of the frequency response of the healthy bearings as a necessary data for the understanding of any abnormal condition. To obtain this knowledge extensive experimental tests in axle bearings are needed as well as defining proven computational tools for the assessment of reliable and fundamental dynamic response of the axle bearings in realistic operation conditions.
• More accurate methods (than measuring and controlling the in-service temperature of rolling stock axle boxes), such as the vibration, the acoustic and the grease analysis, should be brought to the design of axle bearing monitoring systems.
• The monitoring of axle bearings should consider the possibility of adopting in some cases on-board condition monitoring systems that offer the advantage of collecting measurements during a train journey and under different operating conditions. Also the combination of wayside with on-board systems with the integration of the measurements from both systems may be advantageous in some situations.
• Due to the limited scientific and technological knowledge available in the railway industry regarding the on-board monitoring systems and also considering the lack of experience in the development, installation, maintenance and knowledge extraction of on-board systems, there is a need to contribute to the development of the Technical Specifications for Interoperability of the rolling stock and to the requirements and specifications of on-board equipment for axle bearing condition monitoring systems. Also, the requirements regarding the monitoring performance, trends, thresholds, and limits for alarm levels, operation requirements and interface with other systems have to be clearly established for this novel monitoring systems.
• The MAXBE partners have developed and tested a data integration architecture integrating all state-of-the-art and newly developed axle bearing condition monitoring techniques, both for on-board and for wayside monitoring. The architecture contains a specification for the data representation layer, detailing data representation for data on DUT (device under test), i.e. vehicles and axle box and for the test equipment, i.e. the sensors and the monitoring system (data acquisition and processing). The architecture also specifies the protocol to be used at the data communication layer. The standard specification is available as MAXBE deliverable 5.1. To realise the vision of interoperable maintenance and an open market for condition-based maintenance, we propose adding the above standard for interoperable axle bearing condition monitoring data to the Rolling Stock TSI. By adding this specification to the Rolling Stock TSI, an uptake for the LRV market and for wayside systems can also be envisaged.
• Tools for the optimal physical distribution of wayside monitoring systems should be adopted by railway administrations in order to: optimize the number of systems in a railway line without compromising the detection of axle bearing failures at an early stage; contribute for the standardization and the interoperability of the systems’ installation.
• A shift from traditional periodic-based maintenance to a condition-based and predictive maintenance approach can bring clear benefits to the maintenance organisation in terms of costs and efficiencies, and to the train operator in terms of improved reliability, increased fleet availability and most importantly improved safety. However, it is not at all trivial to effectively schedule condition-based maintenance tasks (which could be related to any train sub-system) alongside planned preventive maintenance tasks for all train units in a fleet or indeed across multiple fleets. It is therefore recommended that maintenance and operator organisations make use of planning and optimisation tools such as the one developed in MAXBE to assist with the task of maintenance scheduling and to optimise key objectives within the organisation. This way, tasks can be scheduled automatically and updated as soon as there is a perturbation to the planned schedule. Without an 'optimiser' tool, planning decisions are harder to make and resources may not be managed in the most effective and efficient manner. The need for computationally-assisted planning will increase as organisations rely more heavily on automated asset monitoring systems (such as axle bearing monitoring) and more maintenance tasks are planned based on monitored asset condition, rather than pre-determined frequencies for inspection and maintenance.
Potential Impact:
With the project results it is possible to contribute for detecting at an early stage axle bearing failures; to create conditions for the operational and technical integration of axle bearing monitoring and maintenance in different European railway networks; for the axle bearing diagnosis and maintenance. The project has potential at several levels as mentioned below.
Impact on railway vehicles reliability
The monitoring systems developed within this project have a profound impact in improving the reliability of rolling stock and in increasing the cost-efficiency of maintenance activities based on condition-based maintenance rather than corrective.
Impact on the interoperability
The systems developed in the project both wayside and on-board may be equally adoptable to any type of rail network worldwide thus increasing the exploitation and dissemination potential of the technology implemented during the project.
Impact on railway operation
It is expected that the systems developed in MAXBE (on-board and wayside monitoring systems, wayside and on-board interface software, tool for the optimal physical distribution of wayside monitoring systems, condition-based maintenance software and smart diagnosis tool) will be of general interest in the railways community worldwide and are of direct interest for operators and infrastructure companies. The consortium anticipates that the deliverables of the project may be commercialised within 6-12 months after the end of the project.
Socio-economic impact
The research undertaken in MAXBE has positive economic impact due to: i) the increase of the life time of the axle bearings as consequence of an earlier diagnosis of the bearing condition, ii) the reduction of false alarms as result of using advanced monitoring and sensing technologies as well as wayside and on-board interface, iii) the optimization of the maintenance of rolling stock, iv) the optimization of the physical distribution of wayside monitoring systems and v) the reduction of rolling stock failures. The increase of safety in operation together with the increase of the availability of railway vehicles for operation contribute for a positive societal effects.
Impact on health and safety
The development of reliable and interoperable systems for monitoring, diagnosis and maintenance of axle bearings allows to reduce the risk of derailment and to prevent damage of the rail track.
Impact on environment
By an optimal maintenance of rolling stock components, it is possible to increase the life time of the axle bearings which is a contribution for the environmental sustainability through a reduction of material waste.
Moreover, the successful implementation of the developed MAXBE systems will contribute to the minimisation of the number of derailments associated with rolling stock carrying hazardous substances that can cause significant pollution in the surrounding area where the derailment occurred.
As indicated in D 9.2 several exploitable products were produced by the partners of the project: softwares, tools, monitoring systems.
The knowledge resulting from the project will be part of lectures in the University of Porto – FEUP as well as in the other universities involved in the project. The developed monitoring systems are considered that reach a TRL 5 -6. With further developments and testing expected to occur in short-term period it is possible to reach a TRL 7/8 thus hitting an internship potential placing on the market. Most of the software developed in the project may be commercialised in the next months.
The knowledge resulting from the project was exploited through the following media:
- project web site that will continue after the project with the project webpage to maintain interest in the project;
- presentations;
- press articles;
- publications, technical reports, which are made available for public by downloadable from the project website;
- paper publication in conferences.
List of Websites:
www.maxbeproject.eu
Axle bearing damage initiation, evolution and possible catastrophic failures can cause severe disruptions or even dangerous derailments, potentially causing loss of human life and leading to significant costs for railway infrastructure managers and rolling stock operators. Consequently the axle bearing damage process has safety and economic implications on the exploitation of railways systems that is why it has been the object of intense attention by railway authorities as proved by the selection of this topic by the European Commission in calls for research proposals. Currently, the most used technique for monitoring the axle bearing condition is by measuring and controlling the in-service temperature of rolling stock axle boxes. However the more accurate methods for the bearing health monitoring are by vibration, acoustic or grease analysis
The MAXBE Project (www.maxbeproject.eu) a EU-funded project, appears in this context and its main goal is to develop and to demonstrate novel efficient technologies which can be used for the on-board and wayside condition monitoring of axle bearings. The MAXBE project focuses on detecting axle bearing failure modes at an early stage by combining new with existing monitoring techniques and on characterizing the axle bearing degradation process.
Demonstration of the developed condition monitoring systems was performed in Portugal in the Northern Railway Line with freight and passenger traffic with a maximum speed of 220 km/h, in Belgium in a tram line and in the UK. The monitoring systems consider strain gauges, high frequency accelerometers, temperature sensors and acoustic emission. Both on-board and wayside systems are considered in the project because there is a need for defining the requirement for the on-board equipments and the range of working temperatures of the axle bearing for the wayside systems. Also extensive laboratory tests were performed by the University of Porto to correlate the situ measurements to the status of the axle bearing life. To get a robust technology to support the decision making of the responsible stakeholders synchronized measurements from on-board and wayside monitoring systems will be integrated into a platform. With the MAXBE project concept it will be possible: to contribute to detecting at an early stage axle bearing failures; to create conditions for the operational and technical integration of axle bearing monitoring and maintenance in different European railway networks; to contribute to the standardization of the requirements for the axle bearing monitoring, diagnosis and maintenance. The successful implementation of the project’s key deliverables have positive impact on the the reliability, availability, maintainability and safety of rolling stock and infrastructure with main focus on axle bearing health.
The consortium for the MAXBE project comprises 18 partners from 8 member states, representing operators, railway administrations, axle bearing manufactures, key players in the railway community and experts in the field of monitoring, maintenance and rolling stock. The University of Porto coordinated this research project that kicked-off in November 2012 and was successfully completed on October 2015.
Project Context and Objectives:
The safety of rolling stock and the economic implications of rolling stock maintenance have been of significant concern to the railway industry. In the last few years the maintenance strategy paradigm is significantly changing and the longstanding preventive time based maintenance is being progressively replaced by condition based maintenance. Also the early diagnosis of the rolling stock condition will allow the improvement of the definition of the maintenance strategies for railway vehicles. Condition monitoring technologies offer to the railway operators means to increase reliability and safety. By an earlier detection of potential failures, the railway operators can better plan maintenance actions contributing to achieve financial savings.
The axle bearing damage process and the consequent failures can cause severe delays or even dangerous derailments implicating human lives prejudice and significant costs for railway managers and operators. The usual causes for axle bearing failures are: the ageing and the deterioration of grease; the oil separation from the grease structure; the particle contamination of the lubricant grease due to wear from components; the external particle contamination due to damaged seals and out of balance axle loads. When the axle bearing temperature rises, it can be indicative of possible failure, loss of internal kinematics, lack of lubrication and overloading, factors that can dramatically reduce bearing life.
In railways, the axle bearings condition monitoring can be made directly (on-board) or indirectly (on-track and on-laboratory). Regarding the wayside condition monitoring, hot axle box detectors are the most common track side monitoring systems and are used for the detection of faulty axle bearings in-service. However, these systems cannot detect damage in early stage, because when the alarm temperature is registered the axle bearing is already in an advanced state of degradation. Also, with this type of monitoring systems, overheated bearings may not be detected and false alarms can be triggered during operational conditions. Therefore, more accurate methods for the bearing health monitoring need to be developed considering vibration and acoustic analysis. But these techniques are complex and their application to the railway industry is limited. The on-board condition monitoring systems implies that different types of sensors, generally including temperature, vibration and/or acoustic emission sensors, are installed directly in the axle box, consequently, these systems are more likely to detect an axle bearing fault, especially at an early stage of evolution. But on-board systems may have technical limitations related with power supply, communication system and adaptability of the systems to different type of trains as well as data processing tools that can improve the reliability and effectiveness of these systems.
The goal of the MAXBE project is to provide validated and demonstrated concepts, strategies and guidelines for interoperable axle bearing monitoring and diagnosis. These concepts and guidelines will then support the railway operators and managers that deal with the threats imposed by the existence of axle bearing damage.
This aim is achieved by carrying out the following specific objectives:
- to identify and compare current technologies, commercial systems in operation for monitoring the axle bearing condition, both on-board and track side;
- to identify the requirements and the most suitable indicators for early detection of the axle bearing condition and a combination of COTS sensors for axle bearing monitoring;
- to develop/propose diagnosis methodologies and strategies for the axle bearing condition based on on-board and track side systems;
- to test and benchmark the developed methodologies;
- to develop a condition-based maintenance models for axle bearings;
- to demonstrate the developed methodologies;
- to propose guidelines and protocols for implementing in the European network diagnostic systems for axle bearing condition;
- to disseminate the project results.
To accomplish the above objectives, the MAXBE project comprises nine Work Packages (WP) in which it is proposed: to develop wayside and on-board monitoring systems for axle bearings; to develop an integration platform for the different types of monitoring systems; to develop tools for the maintenance of axle bearings.
Some wayside monitoring systems were developed in the project based on diverse approaches: strain gauges, high frequency accelerometers and acoustic emission. Two on-board monitoring systems composed by temperature sensors, accelerometers and acoustic emission were also build up within the project. To get a robust technology to support the decision making of the responsible stakeholders synchronized measurements from the on-board and the wayside monitoring systems were integrated into a platform. The developed wayside condition monitoring systems were tested in several locations in Europe including the Northern Railway Line in Portugal with freight and passenger traffic with a maximum speed of 220 km/h, De Lijn’s tram line in Antwerp - Belgium and Long Marston, Bescot Depot, Cropredy and Wembly in the UK. The on-board monitoring systems were tested in a train that circulate in the Portuguese Northern Railway Line. The condition monitoring aspects of the research have been complemented through extensive laboratory tests and multibody dynamic modelling used to correlate the in-situ measurements with the status of the axle bearing life. Still within the project, a condition-based maintenance model for axle bearings was developed as well as a smart diagnostics for an early detection of faults and a software tool for the optimal physical distribution of the diagnostic systems.
The detailed description of the Work Packages can be found in the website of the project: www.maxbeproject.eu.
The MAXBE consortium includes 18 partners from 8 European member states: University of Porto (Portugal); REFER (Portugal); ASTS (Italy); UNIGE (Italy); IVE (Germany); COMSA (Spain); University College Cork (Ireland); EVOLEO (Portugal); Nuevas Estrategias de Mantenimiento NEM (Spain); MERMEC (Italy); SKF (Spain); Instituto Superior Técnico (Portugal); Dynamics, Structures and Systems International (Belgium); Vlaamse Vervoersmaatschappij De Lijn (Belgium); NOMADTech (Portugal), I-moss (Belgium); KRESTOS (UK); University of Birmingham (UK).
Project Results:
1 Axle Bearings Failure Modes and Degradation Process
Axle bearing fault initiation can be caused by various factors including friction, wheels flats, poor alignment, poor lubrication, etc. Depending on the factor causing the initiation of the axle bearing fault different, failure types may occur (race or roller bearing related failures, etc.).
Based on data provided from the operators (for COMSA, DL and EMEF) it has been established that the maintenance cycle is kilometre-driven. The bearings provided from COMSA and EMEF are lubricated with grease (Shell Alvania Grease 2760) and the bearings from DL are lubricated with oil (Total Carter EP 220).
Deterioration of the structural integrity of axle bearings can be detected using temperature, vibration and low and high-frequency acoustic emission measurements using sensors installed either on each axle bearing or wayside.
As part of Task 2.1 an interactive and friendly-user database in Microsoft Excel (optimise for version 2010) has been created, A screenshot of the developed database is shown in Figure 1.
Figure 1 – Screenshot of the interactive database developed in Excel.
2 Assessment of existing wayside equipment and on-board systems and operational requirements
In the project, information from the partner’s expertise domain has been used to assess the current state-of-the-art and evolutions. Where available, the installation and operational cost of the equipment has been taken into account. A comparison of on-board and wayside systems has been carried out, also w.r.t. financial viability. One should also note as a side constraint that all systems should be as easy as possible to install and require minimum maintenance during their operation.
3 Laboratory tests
The analysis of operating lubricants allows knowing the state of the machine and following the evolution of any change related to wear of the various lubricated elements, and detecting the changes of physical and chemical characteristics that occur in the lubricant.
Analysis was carried out under laboratory conditions by comparing samples of fresh, used and artificially aged (mechanically and thermally) lubricants of oil from a PCC Antwerpen unit and grease from an UME 3400 unit, using Rheological Characterisation, Direct Reading Ferrograph (DRIII), Analytical Ferrograph (FMIII), Particle Counting (ISO 4406:99; NAS 1638, SAE AS 4059), VIBRO Viscometer (SV-10) and Fourier Transform Infrared Spectroscopy (FTIR).
Rolling bearing surface analysis was performed through visual inspection, microscopic surface analysis and roughness measurements analysis on axle bearing elements (inner race, outer race, cage and rollers) from a PCC Antwerpen unit.
Friction torque tests on rolling bearings were performed and compared for samples of fresh, used and artificially mechanically aged lubricants of oil from a PCC Antwerpen unit and grease from an UME 3400 unit. An experimental setup was used to perform vibration measurements. The experimental setup consists of a test rig, which supports a DC servomotor with a speed controller, a shaft drive, a test bearing housing which accommodates an axle bearing and a hydraulic load applicator.Vibration data were collected for two different load conditions: unloaded and loaded.
Considering the importance of the laboratory work developed in task WP2 to the overall results of the MAXBE project, UPORTO decided to continue the laboratory tests and its analysis in collaboration with the partners interested in providing damaged axle bearings and oil samples. The results attained were submitted in one annex attached to the D2.4 in the website of the MAXBE project, in month 24. In this extended work an exhaustive analysis of the failure modes and the associated degradation process of several types of axle bearings was performed. Several techniques were applied in axle bearings from PCC Antwerpen unit (tram that circulates in Antwerp) and UME 3400 unit (Urban train of Porto), such as Ferrography analysis, viscosity analysis and surface analysis (visual inspection, microscopic surface analysis and roughness measurements). Regarding the rolling bearing surface analysis, two used axle tapper roller bearings removed from a PCC Antwerpen unit were analysed in order to identify different types of surface damage in the inner raceways, outer raceways and rollers that could lead to potential bearing failure. Furthermore, rolling bearing tests were performed with a modified Four-Ball Machine, where both types of axle bearings (PCC Antwerp and UME3400) were analysed considering the specific load conditions and lubricants. Vibration analysis in a faultless axle bearing from the UME3400 unit was also performed considering two different load conditions. This additional task allowed to gather very sophisticated information of laboratory tests, with very accurate numerical models and state-of-the art condition monitoring systems based on diverse technologies of monitoring. An annex was attached to the initial document of D2.4 on month 24, and it is available in the MAXBE project website.
Other type of tests were carried out by UoB on healthy and defective bearings using a customised test rig with capable of rotating sample bearing from 100 up to 1000 Revolutions Per Minute (RPM). An R50A resonant AE piezoelectric sensor procured from Physical Acoustics Corporation (PAC) was mounted on top of the bearing case using a magnetic hold-down. A universal general purpose accelerometer was also used to replace the acoustic emission sensor in each test. The bearing samples used in the laboratory rig tests were PFI Inc, model PW29530037CSHD Ford wheel bearing with dimensions of 28 x 53 x 37 mm. These bearings were disassembled; defects were induced and put back on the rig and include a healthy, outer race and roller defective.
The overall results from the vibration analysis shows that both outer race and roller defects have significantly higher peak to peak and RMS values compared to the healthy one. On the other hand, although Crest Factor confirms the defective bearings, it also shows a high value for the healthy bearing. Therefore Crest Factor does not seem to be a suitable method to analyse the vibration data. Kurtosis rises from healthy to the defective bearing which was expected. The overall values for the AE measurement are similar to the vibration measurement. Increasing in peak to peak and RMS values can be observed however Crest factor and Kurtosis do not provide sensible values that can describe the condition of the bearings. The final information regarding the laboratory tests performed in UK was submitted in D9.1 within WP9 in month 36.
4 On-board systems
Two on-board monitoring systems composed by temperature sensors, accelerometers and acoustic emission were a build-up within the project.
The first system corresponds to a monitoring system with wireless communication technology and Acoustic Emission capability, INSIGHT TM from SKF. Due to its structure this system is dedicated to the bogies of the Pendolino train, Pendulino CPA4000. This prototype system installed on the CP Alfa Pendular train was tested in two phases; the first phase ran from November 2013 – January 2014; and the second phase ran from February 2015 – November 2015. The first phase of testing highlighted key areas for improvement including the low power operation of the wireless sensor nodes and stability of the wireless mesh network. The second phase of testing provided the opportunity to validate these design improvements. During the course of testing provision was made for integration of the data into the Ansaldo STS data fusion, facilitating cross-verification of the on-board SKF Insight™ system with the wayside systems installed at the Estarreja site in Portugal.
The second designed on-board monitoring system is composed by a set of sensors installed in the axle box, a GPS and an acquisition system connected with a data processor unit. It aims to measure the temperature and the vibration levels in three distinct directions. A general overview of the on-board monitoring system configuration is presented in Figure 2. The acquisition system receives the raw data from the sensors and stores it into an on-board CPU. The data management, such as the correlation of the data from the sensors with the GPS information, the data processing, is also performed in the on-board unit. A communication module allows the access to the raw and processed data, enables the remote access to the system and monitors the state of the equipment.
The raw data or the processed data converted into key performance indicators (KPIs) are then transferred to an on-shore server, which enables the access of the data in the data fusion algorithms and feeds the smart diagnostic tool. Considering the benefits of the correlation of on-board and wayside measurements, a radio-frequency identification system (RFID) was considered and installed in Alfa-Pendular trains that cross the Estarreja test site. The RFID system is composed by the tags installed in the trains which send a signal to the reader system when the train crosses the wayside monitoring system.
Figure 2 - General overview of the on-board monitoring system in Alfa-Pendular train (CPA4000)
5 Wayside monitoring systems
5.1 Vibration monitoring system
The vibration wayside monitoring system is installed in the Portuguese Northern Railway Line at Estarreja and it is composed by a set of sensors (strain gauges) installed in the track, connected to an acquisition module and to a data processor module. Sensors are installed at the web of rail along an equivalent wheel perimeter length. The system has a total of 28 strain gauges, divided in groups: 12 sensors installed in the external side of each rail and 2 sensors in the internal side of each rail. The installation scheme is presented in Figure 3.
Figure 3 - Wayside system: (a) installation scheme; (b) location of Estarreja test site
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.
In Figure 4, the acquisition system, the data processor unit and the communication system installed in a cabinet in the side of the track is shown, as well as the installation of the sensors in the track.
Figure 4 - Wayside system in Portugal: data processor unit and system installation in the track
The raw data acquired from the sensors installed in the track is processed in the data processor unit through the data processing algorithms briefly described in Figure 5 that includes information from the infrastructure manager data base (for instance general information about the trains - geometry, references). The settings of the data processor unit, as the definition of the KPIs, trending limits and sensitivity can be remotely defined/modified.
Figure 5 - Data processing algorithms
5.2 High Frequency Vibration monitoring system for Light-Rail Vehicle axle bearing monitoring
A high-frequency vibration-based axle bearing fault detection system jointly developed by De Lijn, D2S, and I-moss is installed in the Antwerp depot of De Lijn, the Flemish public transportation company. The system is installed on the exit track of the facility, monitoring all vehicles exiting from the depot onto the Antwerp tram network. Figure 6 shows some pictures of the wayside installation in Antwerp.
The installation on an LRV (Light Rail Vehicle) 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 these requirements into account.
The complete system is composed of 8 high frequency accelerometers, mounted on the rail feet, field side. The accelerometers are connected to a processing unit, located in the vicinity of the accelerometer in a weatherproof casing. The accelerometers are specified for high frequency vibration measurement and have a sensitivity of 10 mV/g.A mechanical protection system is installed over the accelerometers for protection against moisture, dirt and EMC and isolating them from external vibration sources. Spacing between the accelerometers is 52 cm, 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 3G connection to allow data upload to a centralized measurement database. The in-house built signal conditioning cards have been tuned for high frequency signals, measuring vibration signatures of up to 40 kHz.
All vehicles are identified by a unique number. A magnetic loop reads the vehicle number. As a second identification system, this prototype also contains a high-definition, high-speed camera providing a picture of each vehicle, that allows to decode its identification number optically with OCR software (Optical Character Recognition). This solution is promising for mainline installation.
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 central measurement database, also trending is possible.
The installation has been in continuous operation for more than two years now and has proven to be robust in a very adverse environment, with braking sand and busses passing from the De Lijn depot. This robustness of the installation is a major advantage of this type of sensors over other methods of detection.
Figure 6- Wayside axle bearing monitoring system for Light Rail Vehicle
5.3 Acoustic emission monitoring system
Another customized wayside monitoring system is based on a set of acoustic emission (AE) sensors mounted on the rails, connected to a data acquisition unit and a data processor module installed trackside. The data were recorded after automatically triggering the system to acquire as the train neared the instrumented section of the track.
Wheel and rail form a direct mechanical path for the acoustic emission signals produced from the bearing to be transmitted to the AE sensors mounted on the rail i.e. the detection zone. Detecting bearing defect signals transmitted via this direct mechanical path provides more accurate results compared to airborne acoustic detection, because it eliminates adverse effects of surrounding noises and other environmental parameters, such as wind and aerodynamic forces. Moreover, the signals acquired contain high frequency information which makes detection of faults more likely.
Figure 7 shows the installation outline of the wayside tests carried out in Long Marston, UK. The tank freight wagon containing the three faulty bearings was towed by a locomotive as shown in Figure 8. The faulty bearings were only at the same side of the second, third and fourth wheelsets of the freight wagon with 2, 4 and 8 mm roller defects, respectively. Tests were carried out at up to a maximum speed of 48 km/h over a straight section of welded track approximately 1000 meters in length. The sampling rate for AE signals was 500 kHz and the duration of the acquisition was set at 12 seconds.
Figure 7 - Simplified outline of the wayside installation configuration
Figure 8 - The Long Marston testing configuration. The yellow locomotive pulling the test freight wagon can be seen in the back
An optical unit capable of measuring the speed of the train and counting the number of wheelsets was employed in order to correlate AE signals with the position of the wheels. The optical unit was also used to trigger the data acquisition unit to acquire data while the train was passing through the detection zone. As the system was also counting the number of axle boxes, it is possible to truncate the AE signal exactly at the time that each wheelset passes through the detection zone. This method saves a vast amount of data as the sampling rate of 500 KHz needs a large volume of physical memory. This also makes the analysis period shorter.
6 Wayside and On-board interface software
In the context of railway asset management, one of the most innovative research areas is the multi-sensor Data Fusion; for this reasons, an innovative software Integration Platform based on Data Fusion approach for axle bearing condition monitoring has been developed. It has been performed by three main modules, i.e. a Data Fusion Application (DFA), a System Integrator (SI) and a visualization system referred as Human Machine Interface (HMI), which are interconnected by interfaces that allow communications through messages in eXtensible Markup Language (XML) format, as shown in Figure 9.
Figure 9 - MAXBE Integration Platform – architecture
The DFA gathers all the data at sensor level and combines them by subsequent processing; it extracts high-level patterns through a probabilistic Bayesian methodology which is able to exploit data to output both: 1) an accurate estimation of the occurrence probability of current and/or incoming potential failures; 2) a confidence interval of this risk, useful for decision makers to assess the reliability of the extracted pattern.
The SI manages the interaction between different software technologies in order to integrate every module into a single system capable of collecting data, performing the data fusion in order to extract new information and visualizing the results of this process in a user-friendly manner through the HMI. It updates asset related information disseminated to all the parts of the system as soon as new data is available or new data fusion processing has been completed, so that, for example, the visualization system always shows the most updated information possible.
The HMI (shown in Figure 10) is a web-based visualisation system that allows visualising data and information generated by the DFA, such as potential alarms, in a user-friendly graphical way. The most important characteristics of such an interface are the clearance, the simplicity and ease of use. The visualization system includes pop-out elements that highlight the presence of dangerous situations, and consequently it is able to explore data related to the particular phenomena in order to have, if requested, a more precise view of the situation under examination.
Figure 10 - Example of HMI
The entire software Integration Platform based on Data Fusion approach has been inspired by the JDL generic model, of which it represents a possible implementation for condition monitoring applications.
The system has been able to perform the following functionalities:
• to collect and store measurements and data coming from the WTMSs and OTMSs related to the axle bearing functional conditions;
• to correlate and process these data through a Data Fusion Application (DFA) 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 trains (passenger, freight, high speed, etc.);
• to highlight critical asset condition in order to allow maintenance operators to take proper actions, support decision making by giving a synthetic final alarm level and an associated value (“Veracity”) as the likelihood of the generated alarm level.
The tests and validation of the DFA, and consequently of the algorithm, has been performed and completed with laboratory tests, showing that the system gives satisfying results and fulfils all the requirements. The results show benefits in terms of maintenance costs reduction, risk management and safety, and the system proves to have a positive impact all over the railways systems. Future steps include the realization of a real demonstrator by collecting and processing real data coming from the test sites in Estarreja (Portugal) and in Antwerp (Belgium), officially scheduled for the final phase of the MAXBE project.
7 Tool for optimal physical distribution of the monitoring systems
The monitoring systems are expensive and the investment and the maintenance of this type of systems have to be taken into account by the Infrastructure Managers. Moreover, since there is no available tool to help deciding the distribution layout of the monitoring systems in a railway network within the MAXBE project a tool to identify the optimized physical distribution of the condition monitoring systems and its monitoring interval rates is developed.
The tool considers historical and statistical data of the railway network and particularly data from the railway line in analysis. The developed tool takes into account the risk associated to certain indicators and the importance assigned to each one of the pre-defined indicators, in order to be able to identify the most critical aspects regarding the axle bearing failure in a railway network.
Although within MAXBE project, several wayside systems using different types of technology were developed, the most widespread system already existing in several railway networks is still 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, United Kingdom, Germany, Belgium and France. 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 methodology of the software tool is indicated in Figure 11.
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, by weighting the several risk assessment indicators defined in the software tool.
Figure 11 - Methodology of the software tool.
8 Maintenance tool
The increase in condition-based maintenance (CBM) of assets such as axle bearings should have a number of positive benefits for the train operating company, such as a reduction in in-service failures, a reduction in maintenance costs as maintenance for certain components can be moved from periodic planned maintenance to condition-based, increase in the lifetime of the assets, etc. However, this also creates certain challenges for the maintenance operating company, as there is greater uncertainty regarding what depot resources will be needed for performing maintenance and when they will be needed. Indeed there is already significant uncertainty regarding the duration of planned maintenance exams, as the exact maintenance requirements are often only known after performing a manual inspection in the depot.
The maintenance schedule optimizer software tool that has been developed within the MAXBE project was implemented such that it can be deployed in a number of different maintenance depot scenarios, e.g. single versus multiple fleets, passenger versus freight fleets, with or without overtime staff, periodic maintenance with kilometric or time-based limits, etc. The tool considers time-variable fleet demand profiles, periodic exam due dates, resource/staff depot capacity constraints at an hourly level, over a fixed rolling horizon; while also incorporating proactive and reactive methods for handling the uncertainty inherent in the maintenance scheduling problem.
The problem is solved using dedicated Constraint Programming methods, and the optimizer is implemented in Python in Google’s or-tools solver, which is a free, open-source set of operations research tools. All accompanying software are implemented in Python, including a graphical user interface for the solver and a tool for automated distribution fitting of maintenance exam durations based on historical data, and are open-source and freely available. While the tool can primarily be used for scheduling of daily activities in a rolling horizon capacity, it can also be used to identify bottlenecks in the depot specification. Cost-analysis of alternative depot setups (e.g. increasing or decreasing tracks/resources, altering staff shifts) can then be performed through simulation experiments using the tool.
9 Recommendations and conclusions
Through the MAXBE results and the consultation with potential end users, a number of recommendations can be made.
• Axle bearing monitoring should be seen as an extension to Wheel Flat and Out-of-Roundness Monitoring systems.
• The monitoring of the axle bearing requires the knowledge of the frequency response of the healthy bearings as a necessary data for the understanding of any abnormal condition. To obtain this knowledge extensive experimental tests in axle bearings are needed as well as defining proven computational tools for the assessment of reliable and fundamental dynamic response of the axle bearings in realistic operation conditions.
• More accurate methods (than measuring and controlling the in-service temperature of rolling stock axle boxes), such as the vibration, the acoustic and the grease analysis, should be brought to the design of axle bearing monitoring systems.
• The monitoring of axle bearings should consider the possibility of adopting in some cases on-board condition monitoring systems that offer the advantage of collecting measurements during a train journey and under different operating conditions. Also the combination of wayside with on-board systems with the integration of the measurements from both systems may be advantageous in some situations.
• Due to the limited scientific and technological knowledge available in the railway industry regarding the on-board monitoring systems and also considering the lack of experience in the development, installation, maintenance and knowledge extraction of on-board systems, there is a need to contribute to the development of the Technical Specifications for Interoperability of the rolling stock and to the requirements and specifications of on-board equipment for axle bearing condition monitoring systems. Also, the requirements regarding the monitoring performance, trends, thresholds, and limits for alarm levels, operation requirements and interface with other systems have to be clearly established for this novel monitoring systems.
• The MAXBE partners have developed and tested a data integration architecture integrating all state-of-the-art and newly developed axle bearing condition monitoring techniques, both for on-board and for wayside monitoring. The architecture contains a specification for the data representation layer, detailing data representation for data on DUT (device under test), i.e. vehicles and axle box and for the test equipment, i.e. the sensors and the monitoring system (data acquisition and processing). The architecture also specifies the protocol to be used at the data communication layer. The standard specification is available as MAXBE deliverable 5.1. To realise the vision of interoperable maintenance and an open market for condition-based maintenance, we propose adding the above standard for interoperable axle bearing condition monitoring data to the Rolling Stock TSI. By adding this specification to the Rolling Stock TSI, an uptake for the LRV market and for wayside systems can also be envisaged.
• Tools for the optimal physical distribution of wayside monitoring systems should be adopted by railway administrations in order to: optimize the number of systems in a railway line without compromising the detection of axle bearing failures at an early stage; contribute for the standardization and the interoperability of the systems’ installation.
• A shift from traditional periodic-based maintenance to a condition-based and predictive maintenance approach can bring clear benefits to the maintenance organisation in terms of costs and efficiencies, and to the train operator in terms of improved reliability, increased fleet availability and most importantly improved safety. However, it is not at all trivial to effectively schedule condition-based maintenance tasks (which could be related to any train sub-system) alongside planned preventive maintenance tasks for all train units in a fleet or indeed across multiple fleets. It is therefore recommended that maintenance and operator organisations make use of planning and optimisation tools such as the one developed in MAXBE to assist with the task of maintenance scheduling and to optimise key objectives within the organisation. This way, tasks can be scheduled automatically and updated as soon as there is a perturbation to the planned schedule. Without an 'optimiser' tool, planning decisions are harder to make and resources may not be managed in the most effective and efficient manner. The need for computationally-assisted planning will increase as organisations rely more heavily on automated asset monitoring systems (such as axle bearing monitoring) and more maintenance tasks are planned based on monitored asset condition, rather than pre-determined frequencies for inspection and maintenance.
Potential Impact:
With the project results it is possible to contribute for detecting at an early stage axle bearing failures; to create conditions for the operational and technical integration of axle bearing monitoring and maintenance in different European railway networks; for the axle bearing diagnosis and maintenance. The project has potential at several levels as mentioned below.
Impact on railway vehicles reliability
The monitoring systems developed within this project have a profound impact in improving the reliability of rolling stock and in increasing the cost-efficiency of maintenance activities based on condition-based maintenance rather than corrective.
Impact on the interoperability
The systems developed in the project both wayside and on-board may be equally adoptable to any type of rail network worldwide thus increasing the exploitation and dissemination potential of the technology implemented during the project.
Impact on railway operation
It is expected that the systems developed in MAXBE (on-board and wayside monitoring systems, wayside and on-board interface software, tool for the optimal physical distribution of wayside monitoring systems, condition-based maintenance software and smart diagnosis tool) will be of general interest in the railways community worldwide and are of direct interest for operators and infrastructure companies. The consortium anticipates that the deliverables of the project may be commercialised within 6-12 months after the end of the project.
Socio-economic impact
The research undertaken in MAXBE has positive economic impact due to: i) the increase of the life time of the axle bearings as consequence of an earlier diagnosis of the bearing condition, ii) the reduction of false alarms as result of using advanced monitoring and sensing technologies as well as wayside and on-board interface, iii) the optimization of the maintenance of rolling stock, iv) the optimization of the physical distribution of wayside monitoring systems and v) the reduction of rolling stock failures. The increase of safety in operation together with the increase of the availability of railway vehicles for operation contribute for a positive societal effects.
Impact on health and safety
The development of reliable and interoperable systems for monitoring, diagnosis and maintenance of axle bearings allows to reduce the risk of derailment and to prevent damage of the rail track.
Impact on environment
By an optimal maintenance of rolling stock components, it is possible to increase the life time of the axle bearings which is a contribution for the environmental sustainability through a reduction of material waste.
Moreover, the successful implementation of the developed MAXBE systems will contribute to the minimisation of the number of derailments associated with rolling stock carrying hazardous substances that can cause significant pollution in the surrounding area where the derailment occurred.
As indicated in D 9.2 several exploitable products were produced by the partners of the project: softwares, tools, monitoring systems.
The knowledge resulting from the project will be part of lectures in the University of Porto – FEUP as well as in the other universities involved in the project. The developed monitoring systems are considered that reach a TRL 5 -6. With further developments and testing expected to occur in short-term period it is possible to reach a TRL 7/8 thus hitting an internship potential placing on the market. Most of the software developed in the project may be commercialised in the next months.
The knowledge resulting from the project was exploited through the following media:
- project web site that will continue after the project with the project webpage to maintain interest in the project;
- presentations;
- press articles;
- publications, technical reports, which are made available for public by downloadable from the project website;
- paper publication in conferences.
List of Websites:
www.maxbeproject.eu
