Downer – Enabling Predictive Maintenance

Summary of the Project

Downer Rollingstock Services (Downer RSS) embarked on a Predictive Maintenance Project in 2017. Apart from having to define the project, create a team and compete for budget; the engineering challenge in-itself was monumental – to use data from trains to predict the need for maintenance before it impacted service. Downer RSS expected to significantly reduce contract life-cycle costs, through reduced penalties for delays in-service and by changing its regular maintenance regime.

Typically, a project of this nature is treated as an IT project. However, the framework created by the Asset Strategy & Innovation team, consisted of 4 key areas: an Asset Management Strategy, a technical study to determine data required from the train, a data analytics platform and a Reliability-Centered Maintenance model.

The result has been a successful implementation of a complex project that is expected to save Downer RSS millions of dollars across its long-term maintenance contracts.

Description of the Project

Downer Rolling Stock Services (Downer RSS) built the Waratah Train fleet between 2010 and 2014, with the intention to remotely download large amounts of train condition data to perform condition-based maintenance. Downer RSS produced a number of applications that would be able to manage the large data-sets generated by the train, with train sub-system data transferred to an ICT Shore system to interpret and alert of faults.

Over the next few years, Downer RSS embarked on several projects to evolve those applications to the next generation of analytics and data-interpretation. The first of these attempts was driven by the engineering team to combine MMIS data from IBM Maximo with Passenger Train Condition Data in a project called MaxiMart. This project was driven by the need to address engineering investigation and data exploration for engineering change.

Concurrently, Downer RSS also identified the need to develop an analytics platform that would combine several data sources into a singular platform on locomotives that they maintained, with the ability to write comprehensive business rules to manage and alert on asset condition. This project was called DownerWorld.

This was a clear demonstration of sporadic technology initiatives that resulted from the traditional innovation approach. Neither project was able to be fully funded or supported within the business to achieve its stated promise.

Having identified that the traditional innovation approach was not yielding true business value – Downer RSS changed its approach for the Predictive Maintenance Project in 2017.

 Use of Best Practice Asset Management Principles

The Asset Strategy & Innovation team identified that the Predictive Maintenance project needed to be driven through Asset Management Principles. This involved reviewing the Asset Management Framework to ensure that the appropriate Top-Down and Bottom-Up approaches were resulting in a strategic grouping of initiatives to drive the right outcome. The strategic grouping would serve several purposes:

  • To clearly detail and group together the complete programme of works necessary to achieve pre-defined business improvement, including enabler projects (that provide little value on their own) and respective key projects (that leverage on enablers within the business)
  • To align with the business strategy detailed in the Strategic Asset Management Plan
  • To highlight gaps, overlaps and interdependence of projects run throughout the business.

This was shown to be a gap in the Asset Management Framework, and the Asset Strategy & Innovation team led a business change to add Business Improvement Plans to the framework.

Figure 1 – An excerpt from Downer Rolling Stock Services’ Asset Management Framework that shows alignment between the Business Improvement Plans and the Strategic Asset Management Plan

Applying Downer RSS’ Asset Management Framework, the predictive maintenance project was identified as an Asset Management Objective and was translated into Downer RSS’ Strategic Asset Management Plan (SAMP). This work drew on Downer RSS’ information flow charts and identified that there were clear gaps in three specific areas of the business.

  1. Clarity on what asset condition data is needed to perform condition or predictive based maintenance
  2. A flexible, scalable and supported Data Analytics Platform that would allow condition monitoring of assets, build long-term trends and sophisticated statistical models
  3. Comprehensive RCM models to provide defensible changes to long-term maintenance regimes and drive informed Life-Cycle Cost Models

  Degree of Originality & Ingenuity of Solution

Figure 2 – Example of level of detail of Train Cab and Saloon Air Conditioning Unit MADe model, showing the subsystem components and transfer of energy between components

Although Downer RSS received significant amounts of data from its trains – this data was geared towards managing the current health of the train and not to predict maintenance requirements. There was a clear need to identify the specific data required to monitor asset condition at a sub-system level, as well as the context and operating environment of the asset.

This exercise had not been previously completed in the Australian Rail industry and Downer RSS had to look at other industries to understand how this could be developed.

Downer RSS found that the defence industry was most developed in this space – and leveraged a local Victorian small business to help develop causation models for complex sub-systems. These models would help determine the exact sensors and data required to identify the possibility of each failure-mode occurring on a sub-system. This formed defensible justification to procure, install and/or change sensors on its trains.

Figure 3 – Homepage of Downer’s TrainDNA data analytics platform

Once Downer RSS had identified the right data that needed to be downloaded at the right frequency from the trains – Downer RSS needed to build a data analytics platform to ingest, analyse, interpret and alert its staff of identified predicted asset failures. Downer RSS had previously attempted building these platforms, with some attempts resulting in developing completely bespoke in-house developed software (very functional but difficult to support) versus others that were commercial-off-the-shelf (COTS) products with little to no customisation (high level of support compromising on the end-product which is expensive to customise).  Learning from its previous attempts, Downer RSS invested in the development of a platform that would be an appropriate blend of COTS products and internally supported customisation catering for flexibility in functionality and multiple asset types. The result was TrainDNA.

TrainDNA is a highly-scalable asset-agnostic data-analytics platform. Its foundation is Microsoft’s software stack, that is packaged into a cloud-based, scalable, SAAS solution called Neuroverse by an internal Downer team: Downer Digital & Data Services (Downer D3S). Neuroverse hosts and manages the highly customised TrainDNA, which provides the functionality to manipulate big-data into real-time insights in a structured and well-designed user-interface for various user-groups: operations personnel that need to be alerted of a pending train failure, front-line engineers that need to investigate an endemic issue, and maintenance and design engineers who need to explore maintenance and design change options respectively. The figure below shows the drilling down of reporting from fleet level to component-level analysis.

Figure 4 – Drilling down from fleet-level to component-level in TrainDNA

Apart from identifying asset data requirements using new methods in the Australian Rail industry and developing one of the most comprehensive analytics platforms currently available – Downer RSS ran into an additional challenge: it didn’t employ any data scientists! TrainDNA addressed this by creating an ‘app-store’ service environment, to allow external data scientists to access Downer RSS’ data in a closed and IP secure method to implement analysis techniques coded in the most common coding languages used by data scientists (Julia, Python and R). This created a sandbox for data scientists with essentially no limitations imposed by the platform, where simple statistical analysis methods and complex machine learning algorithms can be run through the same system interface. Downer RSS engaged Deakin University through the Rail Manufacturing Cooperative Research Centre (RMCRC) as its first data scientists.

Finally, to realise the benefits provided from the analytics platform, Downer RSS developed comprehensive Reliability-Centred Maintenance models. These models allow Downer RSS to perform detailed cost-benefit analyses to change their maintenance strategy and embed predictive maintenance into their operations.
This was completed by:

These models become a crucial step to realising benefits by using the insights produced by data analytics and embedding them into Downer RSS’ maintenance strategy. Below is an example of the reports run as an output of the RCM models to compare against various maintenance strategies and ensure that each respective change will result in a lower Life-Cycle Cost (LCC).

Figure 5 – Reporting dashboard for RSS executives to consider the effects of different maintenance strategies

Program & Project Management

Setting up and managing a Predictive Maintenance project in Downer RSS did not come without its challenges. These challenges can be summarised into the following categories below:

  1. Successful implementation and delivery of tangible benefits to the Downer RSS business meant that the Operations and Maintenance teams would need to embrace and integrate predictive maintenance into their way of working. However, the fear of automation and reduction in overall jobs became a significant challenge to overcome culturally.
  2. Downer Group has a national footprint – meaning that its contracts and teams are spread across Australia. Managing multiple teams and key stakeholders in Sydney, Brisbane, Melbourne and Perth was a significant logistical challenge to deliver a complex project over a short timeframe.
  3. Lack of appropriate skills – Data Scientists and analytics software experts are not typically employed by Downer RSS. Developing and managing a complex data driven project without these skillsets was a significant challenge to overcome technically.
  4. Downer Group divisions generally provide services on large Mining, Facilities, Transport and Infrastructure projects. These projects typically follow the Waterfall PM methodology. For a group that was unfamiliar, running an Agile project became a significant accounting and reporting challenge.

The key to getting through this was getting the right people to work together, understand the overall project elements and drive the right outcomes. This involved workshops, meetings, information sessions, reports, presentations, trello-boards and all-sorts of communication with multiple stakeholder groups. Elements of this will remain challenging and the team will continue to collaborate and improve understanding across Downer Group.

                Benefit/Value of the project to the Organisation

The result of focusing on all these aspects together has been the critical recipe to success. Essentially, Downer RSS embodied the Asset Management approach to innovation in delivering on a complex and multi-faceted programme. The Asset Management approach aligned the top-down strategy with business requirements identified from the bottom-up analyses – resulting in a strategic grouping that identified 3 key areas of work to deliver the right outcome instead of focusing only on a data analytics platform.

Downer RSS has already seen benefit in the form of managing critical components of its trains through TrainDNA during its testing phase, before its full roll-out in March 2019. However, this is not the only benefit provided to the organisation. In summary, Downer RSS achieved the following:

  1. Successfully enabled the ability to perform predictive maintenance within its business through the delivery of several initiatives driven through an Asset Management focus
  2. Increased awareness and maturity of the Downer RSS Asset Management Framework by the addition of business improvement plans and using that to drive the Predictive Maintenance Program
  3. Established relationships with universities and helped to support local innovative businesses
  4. A step change in embracing technology – where Downer RSS Operations and Maintenance teams can now have a mature conversation about technology, automation and how it affects skills required in the business in the future
  5. Creation of Intellectual Property in the area of analytics, providing
    1. Primary support for Downer RSS operations efficiency
    1. A competitive advantage over other Rollingstock Services organisations in Australia and New Zealand
    1.  A platform and methodology that can be easily used across various asset types that Downer Group maintains across Australia and New Zealand

The asset management principles and lessons learned from successfully acquiring data analytics capability are being reapplied to future planning of digitisation within the organisation. This platform is part of a larger digitisation business strategy and serves as foundational work towards building the digital future in Downer. As a result, the programme of works around this project is strategically grouped with future programmes of work that build directly on top of the platform, such as Geographic Information System (GIS), Internet of Things (IOT) devices, Integrated Operations Centre (IOC), and an optimisation and simulation platform.

Specific contribution made by the nominated Individual/team

The Asset Strategy & Innovation team has been responsible for the Predictive Maintenance Project and is structured for its success. Reporting to the GM Innovation, the team is embedded within the Downer RSS business as three smaller teams; the Asset Strategy team, the Technology Strategy team and the Maintenance Strategy team. These three teams together had responsibility of the individual aspects of the Predictive Maintenance Project – and without these individuals, this project would not have existed. However, the Predictive Maintenance project involved numerous other teams as well – without whom this project could not have been successful.

These teams include:

  • The Downer Digital – Data Services (D3S) team who is responsible for the development of the Neuroverse and TrainDNA platform. This team has grown in maturity, capacity and capability to keep pace with the huge demands of this project in the past 1.5 years. They have consistently punched above their weight and delivered an excellent platform that is revered throughout Downer Group.
  • Deakin University’s Institute for Intelligent Systems Research and Innovation (IISRI) department has helped develop capability within the TrainDNA analytics platform as our first external research division – providing data science capability within the platform. Their engagement, enthusiasm and support has been second-to-none is making this project successful.
  • PHM Technology Solutions – a small business based out of Victoria that has helped the team determine the correct data required from the train. Their support and dedication has been vital to this projects success.
  • Downer RSS Leadership team, who have supported the project from its inception to its delivery in March 2019. This team has not only funded the team, but provided essential support to overcome cultural, technical and integration issues over the past 2 years.
  • Finally, the Operations & Maintenance teams in both the Through-Life-Support (TLS) and Project Management Services (PMS) divisions that have embraced the project and brought it into their normal way of working.

Figure 6 – The Downer RSS, D3S and Deakin IISRI Team conducting a workshop together in the D3S Perth office in November 2018 on the rollout stages and future work for the Predictive Maintenance Project

General Comments

In summary, there are four general comments that are described below:

  • In 2018, Downer RSS pursued ISO55001 certification for its Asset Management System. This involved a pre-audit Asset Management Maturity Assessment followed by an ISO55001 certification audit in July 2018. Downer RSS was awarded its ISO55001 certificate from Bureau Veritas in 2018 with excellent feedback on its systems, processes and culture. Notably though, both the pre-audit report and the final certification assessment called out elements of the Predictive Maintenance Project as strengths:

Figure 7 – Excerpt from the ISO55001 Certification Audit by Bureau Veritas in July 2018

Figure 8 – Excerpt from the Asset Management Maturity Assessment by IQ-AM in May 2018

  • The amount of data managed by the TrainDNA is most certainly considered big data. In 7 years of historical data alone, TrainDNA currently manages over half-a-trillion data points for only one type of data for only the Waratah Fleet! Below are some stats on the scale of the data being captured and analysed on the regular basis.

Figure 9 – Statistics on the amount of Data in TrainDNA as of December 2018

  • The Asset Strategy & Innovation team who have come together are not your typical rail company employees – instead they are a team of talented individuals with diverse backgrounds. Some stats below show the level of diversity of this team.
  • Lastly, the work completed for determining the right data to be captured from the train is a significant piece of work that will increase maturity for the entire industry. This knowledge is set to be shared with Downer RSS’ Tier 2 suppliers, so that they can provide sub-systems that are pre-built to enable predictive maintenance regimes. This knowledge, that so-far has not been determined and articulated concisely in the Rail Industry, will benefit the industry and its clients overall. Below are some screenshots of the types of documents created as a result of this exercise and a clipping from an Institute of Asset Management Article describing the work completed by the Asset Strategy & Innovation Team.
Scroll to Top