--> Cooperating Artificial Neural and Knowledge-Based Systems in a Truck Fleet Brake-Balancing Application

Michael Lawrence Smith, Eaton Corporation

A proprietary air brake-balance analysis system for trucks gathers five sets of data relating air pressure, time, braking force, and temperature. Each test produces a complex, color graph plotted against axes chosen from pressure, time, and temperature. A human expert can make impressive diagnoses about the brake and air systems after studying these graphs. I describe five artificial neural networks that are trained to render a judgment about these graphs and a knowledge-based system that accepts these judgments and combines them with additional information to arrive at a precise problem identification and a procedure to solve the problem. The brake-balance system is innovative because it uses a rare approach to a real problem: cooperative problem solving and diagnostics between a knowledge-based system and a suite of neural networks. Success rates are 90 percent for the neural nets and 100 percent for the knowledge-based system. The annual savings is at least $100,000.

This page is copyrighted by AAAI. All rights reserved. Your use of this site constitutes acceptance of all of AAAI's terms and conditions and privacy policy.