IAV Hirundo

Gain insight into future fleet performance and associated failures. Probabilistic life prediction enables accurate estimates while providing information on data reliability. This method analyzes vehicle-specific mileage and current failure statistics to predict future failures. It uses probabilistic models that combine machine learning and statistics, accounting for all data uncertainties

Key Features:

• Automatic Data Cleansing: Removes invalid, duplicate, and non-monotonic entries for reliable processing.
• Automatic Model Selection: Identifies the optimal model structure for failure data without manual intervention.
• Vehicle-Specific Mileage Estimate: Creates individual aging models for each vehicle for more accurate fleet mileage distribution.
• Consideration of Uncertainties: Uses probabilistic models to account for data uncertainties, providing probability of occurrence alongside estimates.

IAV Hirundo's approach allows for predicting future failure rates, which helps in estimating warranty costs and planning spare parts. The process is fully automated, making it accessible even for non-experts. It is implemented as a machine learning pipeline on Microsoft Azure and is being adapted as a software-as-a-service application on AWS