Digital Transformation in the Transportation Sector: Unleashing the Power of Data

AECOM’s experience has been that in many cases across the transportation sector, engineering data are often owned by the delivery supply chain and only partially delivered to the sponsoring organisation. As more digital engineering software and tools emerge to the market, the ability for asset owners to manage the data across these tools gives opportunity for large efficiency savings, visualisation, and insight.

One recent road infrastructure project federated all engineering and environmental data and designs into a single system, allowing the sponsoring organisation to manage all digital information, amounting to over 2,000 layers. It was reported that time-saving efficiencies of up to 70% were recording in accessing design information, such as GIS or BIM data, photos, or survey information. Aside from efficiency and transparency, access to this data allows for innovative value-generating exercises. In this project example, 13,000 images from the photogrammetry surveys were processed into a 3D reality model of the project site, which gives engineers, surveyors, and other stakeholders a different perspective of the site and provides a backdrop to other datasets. Storing data in the cloud brings opportunities for streaming datasets to an individual’s device without the need for expensive or expansive memory. 1.4 TB of data in this project are available to stream, leaving it now possible for a wider collection of project members to perform tasks such as creating dynamic cross-sections of the site, without needing to request such an action from the design team. This functionality further enhances efficiency savings. Finally, bringing digital engineering information and data into a single environment gives the managing organisation an audit trail for data access controls, interaction, and version control.

The project showcased how digital technologies can transform the conventional highway construction process by seamlessly blending with existing legacy systems and their data. This fusion of digital and legacy systems not only enhances operational efficiency but also leads to substantial cost reduction. Active stakeholder involvement played a pivotal role during this project. Engaging stakeholders ensured that their perspectives, concerns, and expertise were considered throughout the project's lifecycle. This collaborative approach fostered transparency, alignment of goals, and ultimately contributed to the project's success.

The transportation industry includes many datasets with large unstructured text fields, with examples such as pdf documents, project issues or reviews, or health and safety reports. One category of artificial intelligence (AI), natural language processing (NLP), allows large-scale analysis of text-based information that may have previously been inaccessible, as well as translation services or the extraction of specific information [2].

Data were recently processed with NLP for the purpose of classifying driving incidents, based on text descriptions captured by Traffic Officers in a mobile application. Approximately 18,000 of these records were processed, noting causes, behaviours, actions, and locations regarding the incident. Each of these were classified, with a dashboard describing the insights for each output.

A manual process of reviewing the notebooks of Traffic Officers that was previously unfeasibly expensive for analysis was therefore completed in seconds, for tens of thousands of entries. This is only possible due to the storage of the data in digital format and the advanced analytics of NLP. AECOM’s transportation engineers were then able to perceive insights from the results, drawing and presenting conclusions to the client organisation that will feed forward into policy design and targeted areas for safety improvements for road transportation. This example of collaboration between subject matter experts and data professionals has become the default at AECOM, creating new opportunities for insight that were previously unfeasible.

One example where advanced data engineering techniques have opened new possibilities is with telematics and big-data analysis. Current methods for assessing the distribution of fuel consumption and vehicle emissions average measured vehicle speed over road links that may be kilometres long and disregard the impact of acceleration [3]. AECOM have demonstrated the use of telematic information from real road users to provide an enhanced level of granularity in fuel consumption analysis.

One example dataset encompasses over 6,000 vehicles over two months and includes 0.5 billion datapoints, describing every three seconds the motion, geolocation, and vehicle information for every active vehicle. Insights at national, regional, road, road section, or even specific geolocation therefore are all possible. Each journey was identified and assigned a unique ID, thus allowing the investigation of specific journeys or routes.

The telematic assessment of fuel consumption has three advantages to previous road-link level methodologies: 1) Data are collected at a scale orders of magnitude larger than current methods, simultaneously, and without the requiring costly equipment or roadside in-person measurements; 2) acceleration data add an extra dimension to existing standard road emission calculations in which this crucial factor is not currently considered, giving particular detail to stop/start traffic – a source that is poorly represented in current methodologies; 3) GPS locations of individual timestamps lead to a level of spatial granularity that may be tailored to the research question, rather than an assessment being limited to adhere to traditional road links.

A new scale of data provides opportunity for data science tools that enhance insight and can discover previously unidentified efficiencies. For example, driving styles may be classified with AI to reduce the impact of biases that may exist in the data. The power of this new way of considering emissions opens the potential for future investigations such as identifying, classifying, and understanding national emissions hotspots, or analysing and visualising the distribution of relative fuel consumption for different road, junction, or infrastructure types. However, as with the NLP analysis in Sect. 2.2, analysis is only useful when coupled with the domain understanding of what brings value to organisations and end-users within the industry. It is widely understood that cutting road traffic emissions as part of sustainability is a key goal of highways authorities [4], but the input of subject matter experts allows targeting this to key roads, regions, or environments to align with more precise organisation goals.

Predictive maintenance (PM) is a technique that forecasts the remaining lifespan of critical components through inspections or diagnostics, allowing these parts to be utilised up to their maximum service duration [5]. Employing maintenance scheduling based on real-time requirements, thereby lowering operational costs the technique finds extensive use in Industry 4.0 applications. Research also shows that predictive maintenance extends useful life and reduces failure rate [6]. These are illustrated in Fig. 1.

Bild vergrößern

Electric motors were selected as the equipment of choice for this project due to their extensive utilisation in transportation infrastructure, including applications such as baggage handling systems at airports, traction systems of rolling stock, and electric buses. The primary goal of the project was to leverage real-time sensory data acquired from these motors to determine the optimal timing for maintenance or replacement, thereby preventing substantial performance degradation and unexpected failures.

Highly sensitive vibration and temperature sensors were installed on the motors. These sensors transmitted sensory data to mini edge servers via nearby base stations. After pre-processing the data was sent through 4G communication to the cloud, where virtual models were deployed. These virtual models were created through data analytics and machine learning techniques applied to operational data, essentially representing the P-F curves or signatures of the motors. Real-time monitoring of the motors’ condition against these signatures was conducted. The initial results of the tests demonstrated that the developed models could effectively offer early warnings, detect anomalies, provide short-term predictions of outages, and optimise performance.

The project highlighted the promise of applying predictive maintenance in transportation infrastructure, particularly in mission-critical sectors. Furthermore, it is instrumental in shifting the maintenance business model away from traditional time and material-based approaches towards more cost-effective performance and contract-based models. Additionally, predictive maintenance is frequently linked to the concept of digital twins, which is a prominent focus of many governments. However, when it comes to implementing large-scale deployments, numerous challenges persist. Transportation infrastructure often operates in demanding environments characterized by high temperatures, radioactivity, water exposure, and magnetic fields, all of which can substantially affect data quality. Addressing these challenges necessitates substantial domain expertise and experience.