By applying Artificial Intelligence in transportation, organizations involved in production and management of transportation infrastructures have an effective competitive advantage, thanks to new business models and the optimization of operations. This is a critical factor to appropriately address the new challenges of the economy in the time of Covid-19.
Why Artificial Intelligence in transportation
As pointed out by the CEO of Revelis – Salvatore Iiritano during the Connect for transportation call for partners organized by PWC – “tackling phase 2 by combining productivity and reducing costs should be the goal of companies to be able to start stronger than before”. To succeed in this undertaking in such a critical moment, it is essential to implement Big Data Analytics solutions which, through Machine Learning and Deep Learning techniques, combined with automatic reasoning mechanisms, allow organizations to take advantage of the strategic value of the large amount of information available, through functionalities for:
- data stream acquisition
- phenomena prediction (an in particular fault prediction)
- behavioural profiling for users and machinery
- optimized planning of spare parts and workforce scheduling
- real-time monitoring of transport infrastructures
- multidimensional analysis
Artificial Intelligence enables effective decision support which, combined with process automation logics, allow a radical change in the dynamics of large organizations, allowing on the one hand the reduction of costs and errors related to routine activities, and on the other the use of human resources on more interesting and challenging issues.
Predictive maintenance in transportation
One of the most interesting applications of Artificial Intelligence in transportation sector is represented by predictive maintenance, which, starting from a forecast of the residual operating time (“time to failure”) of critical components (for example the doors of a train carriage or the brake system of a car), allows maintenance and repair activities just in time to avoid stopping the vehicle (predictive maintenance train, predictive maintenance car), ensuring optimized management of the resources involved in maintenance activities (warehouse of the components of spare parts, repair staff).
Predictive maintenance is opposed to “reactive” maintenance, which provides for repairs only after failure has occurred, and “preventive” maintenance, which involves interventions at constant time intervals, regardless of the state of health of the components and the actual risk of failure . It is clear that the paradigm shift brought by predictive maintenance determines a dramatic reduction in costs as well as the maximization of plant operating times.
But how exactly does it work?
Thanks to the diffusion of the Internet of Things, which makes it possible to have increasingly sophisticated and connected sensors, the means of transport (trains, cars) can transmit real-time data streams relating to the measurement of a series of physical and mechanical parameters that represent the “picture” of the operating state of the vehicle itself.
These data streams:
- acquired on big data storage systems
- used by powerful inductive motors to train machine and deep learning models aimed at identifying anomalies
- processed directly on board the vehicle (“edge computing”), applying the induced models in real time
The outcome of the application of the models is the identification of an anomaly, which can be better analyzed through the use of explanation techniques, up to the forecast of the residual life time of the components involved, and consequently to the prediction of faults.
The results of the fault forecasts is the input to optimize the repair shops, both in terms of personnel to be allocated and in terms of spare parts to be made available.
FaultPredictor: Revelis’ solution for predictive maintenance
Revelis offers a predictive maintenance solution based on its Rialto ™ platform which implements the process described in the previous paragraph, enabling:
- monitoring of data acquired by sensors on board the means of transport
- the acquisition and processing of data streams
- the training of anomaly detection models
- the explanation of the phenomena
- the optimization of repair processes
The proposed solution also exploits an automatic reasoning engine, which works in a complementary way to inductive models, allowing a control of “canonical” failure scenarios, through the dynamic definition of rules for the recognition of malfunction scenarios.
The FaultPredictor solution can be applied not only in the transport sector but also in other industrial application areas. Revelis has the expertise necessary for the verticalization of the solution, in order to fit the needs of its customers; Revelis data scientists apply the most modern project management and data analysis methodologies, as well as DevOps approaches, to allow the end customer to progressively achieve its objectives, evaluating the results in itinere and taking full advantage of the business value contained in the data.