Article: Condition Based Predictive Maintenance Systems For Brake System

Share with your network

ABSTRACT: Trains has been a crucial factor in the development of mankind. The key function of a train, that is to transport people / goods over distances, lies at the heart of the usefulness of it. Passenger trains in India clock the highest annual distance, with about 79,835 passenger coaches clocking an average of 770 million kilometres over 1 year. A good and effective maintenance strategy for Brake system is a must have to ensure continuous and efficient operation of the passenger rail network. There are various maintenance strategies currently employed; reactive maintenance, preventive maintenance and scheduled maintenance to name a few. There can be no one maintenance strategy applicable for all parts of a system.

It is highly beneficial to employ a predictive maintenance strategy for function and safety critical parts. One of the ways to achieve predictive maintenance is to monitor the condition of the various components of the brake system, diagnose failures by identifying root cause and deriving the relationship between operational hours and condition of the system at various levels. IoT, Data Analytics, Artificial Intelligence can be used to achieve Condition Based Maintenance. Digital Information about the physical components, such as the Brakes System of coaches, Wheel speed and slippage, is collected and eventually transmitted to computers or server at user premise using IoT. This Digital Information, also known as Data, is analysed by employing various techniques. Using statistical techniques, machine learning algorithms and neural networks, in conjunction with fundamental understanding of the physics of the system, leads to a model of the real-time functioning of the system and achieve predictive maintenance. This will result in a drastic reduction of the unplanned breakdown of crucial components; resulting in increased safety, increased efficiency and reduction in operational costs.

1.0 Introduction:

We live in an age of information. Some truly marvellous inventions and discoveries have happened directly as a result of scientists, engineers and the general knowledge seeker being connected over long distances. This is possible due to what we now call today as the “Internet”, which is a portmanteau of “Interconnected Network”. This was a concept which began in the 1960s and has attained maturity over the past decade. Similarly, Industry 4.0 and its associated technological advancements are occurring today.

1.1 Industry 4.0:

The Industrial Revolution, which began around the year 1760 (almost 260 years ago), is now in the 4th phase of evolution, simply known as ‘Industry 4.0’. Industry 4.0, at its core, utilizes complex Cyber-Physical systems through which huge amounts of information about the physical world is available in a digital format. This is what we now call as ‘data’. These intelligent systems synergize the physical world with the digital world, thus allowing for the computational power of digital technology to dissect, analyse and solve real-life problems yet staying relevant enough for the Human mind to direct the analysis.

Figure 1: The Industrial Revolution Roadmap

1.2 Predictive Maintenance:

Predictive maintenance is a technique that uses condition-monitoring tools and techniques to track the performance of equipment during normal operation to detect possible defects and fix them before they result in failure.

Condition of Brake systems can be monitored by interpreting the data about the various electric, mechanical and pneumatic components. Tracking the performance of equipment detecting possible defects beforehand will allow for the system to remain functional and available, be more efficient and reduce maintenance costs. Predictive maintenance is best realized when it reduces the maintenance frequency without incurring significant associated costs related to implementing this strategy.

2.0 Condition Based Maintenance:

India has the fourth largest rail network in the world. The Ministry of Railways operated Indian Railways (IR) has a total route length of 68,103 km and a track length of 1,26,611 km. Indian Railways has a rolling stock of 39 Steam Locomotives, 5,108 Diesel Locomotives, 7,587 Electric Locomotives for a total of 12,734 Locomotives. Total number of wagons are 3,02,624, with a total number of coaches at 79,835. Passenger coaches alone make up 71,716 of them. 1 These large numbers indicate the sheer volume of rolling stock available with Indian Railways. With such a large quantity of assets, one can clearly see the importance and monetary impact of timely and quality maintenance of these assets.

Remote monitoring and diagnostics a maintenance strategy which employs various Industry 4.0 techniques to implement a Condition based Maintenance and Diagnosis system to carry out maintenance activities not based on time or distance period, but rather on the health and condition of these assets. The benefits of implementing this type of maintenance strategy is numerous. It is with this understanding that Wabtec has invested into researching and implementing Condition based maintenance for its own product lines supplied to Indian Railways, and eventually, provide a value-added service for implementing Condition Based monitoring for key assets of the Indian Railways rolling stock. This is achieved by Condition Monitoring, Data Analytics and Predictive Intelligence.

3.0 Iot Enabled Condition Monitoring:

One of applications of Industry 4.0 technology is the Condition Monitoring System. All the parameters and information related to an asset is collected via various devices, such as sensors, logic controllers, vision systems and more. Various assets parameters are then transmitted from the on-field data collection devices to centralised digital systems. These digital systems will process the data and thus allow engineers and data analysts to understand the health of the asset by establishing trends, predict failures and estimate the remaining life of an asset. Operational inferences can also be made to allow for optimization of energy consumption, optimization of duty cycles, safety assurance, reliability calculations and much more.

Data from the various components of the Brake systems will be collected via various sources and transmitted to servers via an IoT Gateway. An IoT (Internet of Things) gateway is a programmable hardware controller which enables communication between devices or between devices and a software platform. It serves as a connection between the server and the physical devices, usually by “pulling” the data from various sources. The data is then segregated, stored and made usefully available via dashboards, inferences and alerting systems.

4.0 Data Analysis:

The data collected from the components of the Brake system will be thoroughly analysed using various statistical techniques and fed into Machine Learning Models. Statistical analysis techniques, such as descriptive and inferential techniques may be used for producing various results. Primarily, we aim to describe the nature of the data collected, explore the relation of the sample data to the population and create a summary model explaining the various relationships and inferences derived. Techniques such as the t-test, z-test, ANOVA and tailed tests are used.

Figure 2: Snippets of Analysis Done

5.0 Predictive Maintenance as a Reality:

Wabtec has always strived to be ahead of competition. One of the key aspects is being the forerunner in innovation and adoption of new technology and best practices. Therefore, Wabtec has implemented Remote Monitoring and Diagnostics in its freight locomotives supplied to Indian Railways and are striving for the same in the passenger coaches. Faiveley Transport, a wholly owned part of Wabtec, supplies Air Brakes systems, HVAC systems, Couplers, Pantographs and Doors to the railway industry and car builders. Systems are being developed for implementing Remote Monitoring and Diagnostics for Brakes and HVAC systems. In a broad perspective, there are 3 steps employed to achieve this, as seen in below figure.

Figure 3: Wabtec’s 3-step process

5.1 Scope of Predictive Maintenance in ILS Brake System:

The purpose of this document is to detail the specification and requirements of the implementation of Predictive maintenance technology for the ILS Brake System. IoT gateway will be used to continuously log data about the sub-systems we supply. There are 3 major components to the Predictive maintenance Implementation for ILS, as shown in the figure below:

Figure 3: Interfacing, Data Collection, Transmission, Storage and Dashboards

Our Brakes systems have controllers which have data about the components of the systems available in them. These data will be pulled via a programmable and robust IoT gateway, which has been developed internally, and sent via a private mobile network stream to our processing servers. We ensure that the data transmission is reliable, fast and completely secure. Failures are also minimized by minimizing the data collection points and having redundant gateways for the crucial data collection points.

Data logged will be about various key parameters such as compressor status, cylinder pressure, valve position and similar information. The Web application is the most important segment of the web platform. It is the key back-end of the web-portal. The web portal is where users will access the data via visualization on dashboards. It will be a highly secure platform, with ability to view data, download it, receive alerts and more features.

5.2 Benefits of Predictive Maintenance:

There are multiple benefits of implementing predictive maintenance. On an average, Maintenance costs are down by up to approximately 30% for critical components, unexpected failures can be reduced by up to 40%, repair and overhaul time can come down by up to 40%, spare parts inventory may be reduced by up to 30%, general increase of up to 30% in mean time between failures (MTBF) and a general increase of up to 30% increase in system uptime. Studies also show that up to 30% of components having preventive maintenance strategy may be replaced by predictive maintenance.

  • Wabtec is looking to achieve the following benefits, namely in the aspect of maintenance optimization and service optimization:
    • Maintenance Schedule to be optimized based on prediction of failures
    • Life Cycle Costs to be made more accurate based on On-Field failure data
    • Database of Failure rates, failure modes, design changes and their respective impact on system failure to be dynamically maintained
    • Services response time can be synchronized with a predicted failure to ensure availability of service engineer and spare parts
    • Spare parts can be optimized based on prediction of failures
    • Storage of spare parts and dispatch of the same can be optimized based on failure predictions.

5.3 Data Analytics And Machine Learning For Failure Prediction:

Various statistical analysis and machine learning models are employed to achieve predictive maintenance. The Data will be fed into various Machine Learning Models for various results. Supervised learning techniques (e.g. regression, Naïve Bayes, ensemble techniques) are used for failure prediction of components over time and hidden patterns. Unsupervised learning techniques (e.g. kNN, clustering, random forest allocation) are used for anomaly detection associating failure with possible root causes and hidden relationships.

Figure 5: Machine Learning Model Development

Maintenance Schedule is optimized based on prediction of failures. Life Cycle Costs estimations are made more accurate based on On-Field failure data. Database of Failure rates, failure modes, design changes and their respective impact on system failure are dynamically maintained.

Services response time can be synchronized with a predicted failure to ensure availability of service engineer and spare parts. Spare parts can be optimized based on prediction of failures. Storage of spare parts and dispatch of the same can be optimized based on failure predictions.

Figure 6: Detecting condition degradation early on

6.0 Advancements of Predictive Maintenance:

Wabtec locomotives which is supplied to India railways is equipped with Remote monitoring and diagnostics at locomotive level which tracks important parameters of locomotive, use the data to process and built intelligent prediction models through which major service affecting failures were predicted way ahead to give an alert to customer to take appropriate actions. This has led to save millions of INR with respect to penalties and other associated risk.

There are multiple projects globally with Wabtec remote monitoring and diagnostics technique implemented as part of Brake system. To state some examples: J151 project in Singapore, Grand Paris Express in Europe which has Metroflexx Brake system.

6.1 Metroflex Overview:

Metroflexx is an integrated brake control system that features a service brake, emergency brake, wheel slide protection, communication with TCMS, advanced diagnostic function including Condition Based Maintenance.

  • Key Benefits:
    • Unique versatility: one single part for multiple brake configurations (control per car / per bogie / per axle).
    • Very low weight thanks to advanced manufacturing process (additive)
    • Extended MTBO, up to 10 years
    • Very short response time for reduced dwell time
    • A patented remote release ensures maximum train availability (no undue train immobilisation in the case of a single failure).
    • Includes our latest adaptive wheel slide protection and brake compensation for shorter braking distance guaranteed in all adhesion conditions, and best wheel slide protection.

As part of introducing this concept in Indian metro, discussions with DMRC in progress to implement Metroflexx Brake system with remote monitoring with the below key aspects;

  • METROFLEXX has data storage & calculation capabilities to provide data analytics and therefore maintenance recommendations, condition status, MTBO prediction
  • These data can be transmitted to authorized personnel trough internet, with no need to hold trains and retire them from commercial service
  • METROFLEXX is in fact provided with a WiFi IoT (Internet of Things) module, where transmission antenna is embedded
  • The communication between the module and METROFLEXX is ensured to be SAFE by hardware design. There is no possibility to hack METROFLEXX software
  • Data can be collected through a centralized system (composed mainly by a WiFi antenna and a computer) installed into the main depots or stations
  • Data can also be transmitted over internet thanks to WiFi connectivity
  • All data will be collected for data processing, optimized overhaul planning, predictive intervention, REX
  • Said data can be crosschecked automatically with the data collected by test-benches, creating a strong maintenance REX database

6.2 Wabtec Locomotives:

Wabtec locomotives had a great presence across globe. With 17K loco serving 50 different customers around the globe is equipped with condition based predictive maintenance technology provided a technological advantage compared to our competition. Below tables provides some of the reference in India and other parts of world where our system are installed and running successfully in field. Customer – (Country) No. Of Assets
1. IR Wabtec Locomotive, (India) 400
2. AURIZON, (Australia) 65
3. BNSF, (US) 4000
4. CSX, (US) 2000
5. RUMO, (Brazil) 600

Figure 7: Indian & Global references for Condition based and predictive maintenance

Figure 8: Indian assets for Condition based and predictive maintenance

For locomotives, we adopt 5 level technique as shown below with advantages by saving both material and labour

  • Confirm healthy, do not replace
    • Health model or remaining useful life model
    • Data analytics augmented with engineering models
    • Use RM&D data, material usage, service history
    • Most useful in consumables components
  • Condition based work scope
    • Data driven work scope decisions (Lite / heavy / full)
    • Can leverage analysis for decision consultation
    • Most useful in high value, low frequency components
  • Avoid secondary damage / Pinpointed diagnostics
    • Proactive replacement of parts to avoid higher secondary damage
    • Pinpoint component to be replaced
    • Will need domain + analytics

7.0 Conclusion And Further Applications:

Condition monitoring based predictive maintenance is the first step in Wabtec’s vision for revolutionizing the railway industry. The goal is to implement an Artificially Intelligent control system in conjunction with condition monitoring. AI will be able to tune parameters, implement quick containment actions based on failure predictions, request maintenance activities, order spare parts, optimize train timetables and control the train’s operation. This is one of the applications of latest technology in the railway industry. By using Condition monitoring based predictive maintenance for ILS Brake system, the data collected on the dynamics of braking system is used to monitor the health and helps in analyzing the performance of system. A health model or remaining useful life mode will be designed.

  1. By monitoring patterns in real time and looking at data, the machine learning model can identify repeat scenarios which can be used to create rules for failure prediction. With the implementation of CBM on ILS brake system, our aim is to move from preventive or corrective maintenance approach to a condition-based maintenance approach. With real time alerts to operator, the maintenance plan will be optimized. The important parameter of Brake system will be tracked and health of this will be made available on demand. With performance dashboards, a complete visualization of Reliability, failures and other trends can be viewed on a wholistic manner.
  2. There are multiple other applications possible by utilizing the same technology for Brake system. Product traceability at component level through all levels may be achieved, starting from ideation stage through manufacturing to delivery and demonstration during service. This will result in highly useful data being available which can be used with backward traceability for RCA, design improvement, cost reduction etc.
  3. With increasing rail network and considering the existing operational fleet, adopting and implementation of condition-based maintenance will have huge benefits in terms of operational Reliability which will enhance the Availability of trains in fleet which will result in better customer satisfaction.

Source: Wabtec Corporation

Be the first to comment

Leave a Reply

Your email address will not be published.