RADIAL BASIS AND LVQ NEURAL NETWORK ALGORITHM FOR REAL TIME FAULT DIAGNOSIS OF BOTTLE FILLING PLANT

Abstract

In this study, an Artificial Neural Network (ANN) is developed to find faults rapidly on a pneumatic system. The data were saved and evaluated considering system is working perfectly and faults are predetermined. These faults include having no bottle, a nonworking cap closing cylinder B, a nonworking bottle cap closing cylinder C, insufficient air pressure, water not filling and low air pressure faults. The signals of six sensors were collected during the entire sequence and the 18 most descriptive features of the data were encoded to present to the ANNs. Two different ANNs were applied for interpretation of the encoded signals. The ANNs tested in the study were learning vector quantization (LVQ) and radial basis network (RBN). The performance of LVQ and RBN was found to be fine with the presented procedures for a system having very repetitive sequential data.