Predictive diagnosis and preventive maintenance

The Predictive Diagnosis and Preventive Maintenance of vehicles powertrain subsystems consist in a set of predictive and diagnostic models with the capability of modelling the normal behaviour of the subsystem and predict the anomalies evolution.


The main goal of this project is to develop time-saving and cost-effective tools, which for the industrial vehicles companies means to develop solutions able to:

  • Reduce repair time at workshops,
  • Avoid replacement of healthy components,
  • Prevent severe failures.

Product capabilities

The diagnostic and predictive models achieve these objectives performing:

  • Normal behaviour modelling,
  • Subsystem health index computation,
  • Early detection of anomalies,
  • Estimate of residual time and km to potential critical event,
  • Root cause identification.


Predictive Capabilities

The developed models are capable of detecting the relevant anomalies before their occurrence and to provide a rough estimate of time and distance remaining to a future potential breakdown

Root cause identification

The Fault Isolation Module was deployed, integrating the set of diagnostic models developed for each of the powertrain subsystems, to allow identifying the anomaly root cause

Reduction of repair time

The early detection of the anomalies together with the root cause identification, allow operators to send the driver to the workshops with the component to be replaced already in stock or to notify the workshop to be ready for a specific repair

Reduction of maintenance cost

Especially in cases of ordinary maintenance, when components are replaced regardless of their condition, the diagnostic model avoids unnecessary replacements of healthy components


System engineering approach

This approach encompasses all the aspects related to physical systems modelling, fault detection and isolation, data analysis, algorithms definition, improvement and customisation, software design and development for the integration with on-board (ECU) or off-board platforms

Machine learning

The subsystems normal behaviour models are developed based on Machine Learning techniques (e.g. Neural Networks), exploiting a set of telemetries sent from a fleet of vehicles to the monitoring service

Patterns extraction

Consistency maps were developed based on patterns extraction, with the purpose of identifying sensors inconsistencies and isolate these from subsystems’ faults

Fault isolation

The Fault Isolation strategy had to deal with more than 70 different root causes and more than 20 diagnostic models (models of normal behaviour of powertrains’ components)