Condition Monitoring of Plastic Extrusion Machine Using Artificial Neural Network

Proactive programs as condition monitoring justify the most extreme demands of plastic industry as safety, reliability and cost-competitiveness with other ones. It is a less expensive and precautionary way, rather than the reactive one. Condition monitoring program is utilizing different emerging technologies for better results. The aim of this practical research work, carried out on a plastic extrusion machine, is to develop cheaper and better condition monitoring expert system to detect incipient faults in it. This expert system consists of vibration data acquisition hardware, developed using MMA7260QT Micro machined accelerometer and interfaced with computer through line in port of sound card. Vibrations are acquired from the three different places of the test machine, saved and analyzed for features extraction in both time and frequency domain. These features are fed to Probabilistic neural network (PNN) i.e. ANN constructed choosing probabilistic network model. PNN is used for classification and interpretation of faults. This developed fault diagnostic expert system is non-invasive, cheaper, reliable, and NN based. The experimental results using actual data show high accuracy. The developed system can be used to detect faults of other small or medium machines. Keywordsartificial neural networks (ANNs), condition monitoring, features extraction, Plastic extrusion machine, vibration acquisition