Multi-Agent Remote Predictive Diagnosis of Dangerous Good Transports

Abstract

A real-time remote monitoring system for motor-vehicle fleets of dangerous good transport is proposed for the dynamic fleet management, the predictive diagnosis and the failure prognosis of vehicle wear, operating danger, and fraud on goods. Predictive fault diagnosis and failure prognosis are performed by a hierarchical multi-agent architecture based on 2 levels: (i) fault detection and isolation, by wavelet transforms (signal segmentation), Bayesian logic (feature extraction), and cultural algorithms (fault isolation); and (ii) diagnosis-prognosis, by hybrid structures fuzzy/neuro-fuzzy