EXPERIMENTAL EVALUATION OF A STRUCTURE-BASED CONNECTIONIST NETWORK FOR FAULT-DIAGNOSIS OF HELICOPTER GEAR BOXES

This paper presents the experimental evaluation of the Structure-Based Connectionist Network (SBCN) fault diagnostic system introduced in th e preceding article (Jammu et al., 1998). For this, vibration data fro m two different helicopter gearboxes: OH-58A and 5-61, are used. A sal ient feature of SBCN is its reliance on the knowledge of the gearbox s tructure and the type of features obtained from processed vibration si gnals as a substitute to training. To formulate this knowledge, approx imate vibration transfer models are developed for the two gearboxes an d utilized to derive the connection weights representing the influence of component faults on vibration features. The validity of the struct ural influences is evaluated by comparing them with those obtained fro m experimental RMS values. These influences are also evaluated by comp aring them with the weights of a connectionist network trained through supervised learning. The results indicate general agreement between t he modeled and experimentally obtained influences. The vibration data from the two gearboxes are also used to evaluate the performance of SB CN in fault diagnosis. The diagnostic results indicate that the SBCN i s effective in detecting the presence of faults and isolating them wit hin gearbox subsystems based on structural influences, but its perform ance is not as good in isolating faulty components, mainly due to lack of appropriate vibration features.