ITALY’s infrastructure manager Italian Rail Network (RFI) launched an ambitious project in 2016 to renew and upgrade its fleet of on-track vehicles (OTV) and road-rail vehicles (RRV) used for measuring and diagnosing infrastructure faults in order to take advantage of the latest technology. The project is part of a wider programme to modernise the fleet of infrastructure maintenance vehicles.

RFI has taken a two-pronged approach to the renewal of its diagnostic fleet and adopted the following short and medium-term actions which started in 2017:

  • construction and commissioning in April 2020 of a new self-propelled vehicle for ultrasound diagnostics of rails, track geometry and wear called DIC-80 supplied by Plasser and Theurer with Nordco ultrasonic technology
  • construction of two new self-propelled vehicles for the diagnostics of track and overhead line geometry and wear, and measuring characteristic parameters of switches, called Falco supplied by Tesmec – the vehicles are now in operation
  • introduction of a self-propelled vehicle called Sirter with systems for diagnosing track and contact line geometry and wear
  • introduction of a vehicle called Aldebaran 2.0 with systems for diagnosing track and contact line geometry faults and wear, as well as telecommunications systems faults
  • refurbishment of the Diamante high-speed ​​diagnostic train, renamed Diamante 2.0, by replacing the existing prototype coaches with coaches compliant with Trenitalia’s Frecciarossa ETR 500 fleet along with installation of the latest diagnostic systems
  • introduction of three K12 vehicles with systems for diagnosing track geometry faults and wear, as well as measuring characteristic parameters of switches, and video inspection supplied by DMA – the track geometry and wear diagnostics systems will be commissioned this month, with the remaining diagnostic systems following in January after AI training has been completed, and
  • introduction of two OBW10 self-propelled vehicles with systems for diagnosing track geometry and wear.

Long-term actions coving the 2023-2024 period are based on the need to:

  • adapt the equipment to technical and regulatory changes
  • the introduction of redundancy technology, correlation and artificial intelligence (AI) for the diagnostic data validation process, with the aim of raising the SIL level and RAMS performance
  • increasing the frequency of surveys to support the transition towards a predictive maintenance policy, and
  • introduction of innovative equipment based on machine learning aimed at mitigating the risk of human error in track supervision activities, the visibility of signals and checking the characteristic parameters of switches.

In addition, RFI wants to reduce operating costs by improving the poor reliability and low availability of its current vehicle fleet and reducing its reliance on renting vehicles and outsourcing transport and diagnostic services. RFI adds that it wants to increase the amount of infrastructure maintenance done in-house while increasing the timeliness of maintenance and improve the ergonomics of work sites. The programme is underway and is structured according to Table 1.

Four types of train are involved in the fleet renewal programme. Type 1 trains comprising five bi-mode electro-diesel trains with a maximum speed of 130-160km/h will be used on the conventional network. Maintenance of the vehicles will be entrusted to the supplier, a temporary partnership of Mermec and Stadler, for the first six years. RFI has chosen a fixed composition with inter-communicating driver’s cabs to allow single driver operation. RFI intends to adopt remote diagnostics which will make it possible to eliminate operators that today man each diagnostic station on the existing trains while at the same time increasing the number of surveys per day as only the drivers’ rosters will have to be planned.

Table 1

Type 2 trains will comprise seven multi-function trains equipped to carry out ultrasonic rail diagnostics and tunnel inspections. Here RFI has chosen rolling stock consisting of a driving trailer and either a towed vehicle with a traction unit or a Type 4 bi-mode vehicle. The trains will be equipped with ultrasonic diagnostic systems designed to guarantee the best connection with the rails and the lowest consumption of connection fluid to improve the identification of internal defects in the rails and prevent their development. These systems will be redundant, with consequent benefits in terms of RAMS parameters. They will also be integrated with eddy current systems to identify surface defects which cannot be detected by ultrasonic technology, as well as video inspection systems to support data analysis. Three trains will also be equipped with automatic tunnel inspection systems, based on machine learning algorithms.

Type 3 trains involve the refurbishment of the existing Diamante high-speed train. Diamante 2.0 will be equipped with innovative diagnostic systems and will be complemented by a second identical train, called Aiace 2.0. The onboard systems will allow the automatic inspection of the contact line and automation of the signal visibility check through automatic inspection systems based on machine learning algorithms.

Finally, there will be 15 bi-mode electro-diesel Type 4 units with a cab at each end. Like the Type 1 trains, these units will be designed for remote diagnostics so will only require a driver on-board. Type 4 trains will be used to monitor junctions and yards and will have a maximum speed of 160km/h.

Diamante 2.0

The diagnostics technology of Type 1, Type 3 and part of Type 4 trains is being developed in partnership with Mermec. Diamante 2.0 will be able to measure track condition at speeds of 330km/h. Multiple laser sensors and cameras working as an integrated system will automatically provide information about the profile of the rail head and overhead line geometry, measuring shape and movement optically. At the same time transducers and accelerometers will mechanically measure the 3D movement of the train as it travels along the track. This data provides information on track geometry - the shape and profile of the rail head, and track twist.

A machine vision system will use a series of lasers and cameras, and AI to detect faulty track and overhead line components. Image analysis software will use an algorithm to compare what the cameras see with an as-built image repository of the infrastructure. For example, the systems will automatically identify missing rail fastenings, which is a common but significant fault, and immediately notify the problem to the maintenance team for remedial action.

Stadler and Mermec were awarded a €130m turnkey contract in May to supply five new
bi-mode diagnostics trains.

Other examples involve surface defects and sleeper crack detection, which are two of the most complex tasks for this kind of system. The use of machine learning technology guarantees performance optimisation during the project’s lifecycle as the system learns automatically from validated and confirmed faults, which translates into maximising the overall detection rate while concurrently minimising false positives.

Wheel-rail interaction is another important measurement that has been implemented by RFI on its measurement vehicles. The specific acquisition system allows the monitoring and measurement of the parameters representative of the dynamic behaviour of the vehicle subjected to the stresses imposed by the track, providing useful information to characterise safety levels for trains in operating conditions. Measurements are performed using strain gauges installed directly on the wheels.

The vehicles are equipped with several systems for monitoring signalling subsystems and telecommunications. A system has been developed to monitor and analyse the operating condition of track circuits to detect possible anomalies in ground equipment. This type of analysis is fundamental to avoid issues relating to signalling subsystem malfunctions that can jeopardise safety.

An automatic system to recognise lineside signals from the train driver’s perspective has been implemented using a stereoscopic approach. The algorithm recognises and classifies signals in real-time within a stream of images acquired using diagnostic systems. The stereoscopic system consists of two high-resolution cameras arranged on the same floor at a fixed and known distance to detect the type of target and its position in relation to the track centre so that all signals are detected and localised.

These innovative opto-electronic vision and information technologies, while enabling early detection of failures and irregularities, can also generate significant storage and data modelling requirements, which if not addressed properly, can lead to data being under-utilised and even abandoned forming data graveyards. The complexity and the number of systems, which sometimes operate in isolation, don’t always guarantee that railways have new information to support safety and track maintenance management in an efficient and effective manner.

Conceptualising and realising the right data model and architecture enables the exploration of these big data lakes transforming the information into actional deliverables. Modern web platforms, either in the cloud or on premises, give train operators and infrastructure managers access not only to the diagnostic information, but also the tools to analyse and forecast track condition, and finally adopt a successful condition-based and predictive asset maintenance strategy.

The key challenges that big data integration and analysis web platforms address include:

  • multiple data sources
  • massive amounts of data history
  • limited data cleansing and filtering
  • quantifying the impact of operational changes, and
  • minimising train delays without jeopardising safety.

Diamante 2.0 and the type 1 and 4 vehicles are equipped with a series of sensors that not only take care of the diagnosis of infrastructure faults, but also of the diagnostic systems themselves.

The software manages real-time data acquisition while simultaneously looking for problems in data storage and transfer using 4G and 5G connectivity. Specific attention has been given to the data architecture onboard and in the office so that data transfer is performed seamlessly.

A solution has been developed to monitor and control the operation of such systems to increase their reliability and availability and ensure that there are no service interruptions by centralising the diagnosis of system failures and their modular resolution. The analysis of faults and their causes is fundamental for the evolution of the systems themselves and of their design. The diagnostic systems installed onboard the vehicles follow the IoT digital transformation business processes.

As part of this project, it is vital to guarantee very high protection of the stored data and communication through stringent protection from possible cyber-attacks especially for safety-critical signalling systems. Cyber security applies to both the management of infrastructure maintenance and train operation. In this way, the confidentiality, integrity, and availability of the information is maintained, in addition to other characteristics such as authenticity, non-repudiation and reliability.

During the life of the project, 5G technology has been considered as a possible enabler to improve the performance of IoT and deliver the information to the final users. Tests are underway with the main operators: TIM and Vodafone.

RFI in partnership with the Ugo Bordoni Foundation (FUB), TIM, Mermec, Telespazio, the Bruno Kessler Foundation and Marini Impianti Industriali has invested in developing 5G and satellite technologies for the railway, launching the Satellite Diagnostic Integrated Networks and the Dinos5G project. The latter was approved and funded by the European Space Agency (ESA) and aims to integrate 5G and satellite mobile network technology to maximise the efficiency of the network regarding maintenance and to minimise traffic impacts. This is achieved using predictive diagnostics systems capable of processing in real time many signals from sensors installed throughout the rail network and from measurement systems installed on diagnostic trains.

Predictive maintenance will further raise network efficiency standards. In the future it will be possible to monitor the infrastructure using drones, allowing intervention even before an anomaly occurs.

The pilot site for the trials is located at the Bologna San Donato test circuit, which will be equipped with the latest sensor systems, and will be able to use an integrated 5G-satellite communication channel to centralise all diagnostic data related to the infrastructure.