Development of Prediction Model For System Performance

A system by definition is an assemblage or combination of things or parts forming a complex or unitary whole. When a system fails to operate its function, it interrupts the system's performance. The aim of this study is to develop a neural network based prediction model to predict a system's performance. A system in this scope of study is portrayed as equipments centrifugal pumps, centrifugal compressors and expanders. Failure modes of these equipments are listed down and the causes of failure will be monitored and used as inputs for the prediction. The prediction model is simulated using Neural Network tool from MATLAB software. Initial modelling of the model has been done using data from Jahirul et al research paper to test the functionality of the model. Actual data of centrifugal compressor and expander were then used in the model. The result for both compressor and expander models were very accurate with an average percentage errors of 0.13% and 0.176% respectively. These models are considered reliable and can be used to predict future target data of these equipment. The objective of this research is achieved as these models are able to predict the performance of a system in this case equipments centrifugal compressor and expander.