USE OF ADAPTIVE NEURO FUZZY INFERENCE SYSTEM FOR FAULT DIAGNOSIS OF POWER PLANT
USE OF ADAPTIVE NEURO FUZZY INFERENCE SYSTEM FOR FAULT DIAGNOSIS OF POWER PLANT
Madu, C.,Fadayini, O., Folami, N.A., Ipaye, T. A, and Ovowarie A. P.
Department of Chemical Engineering, Lagos State Polytechnic, Ikorodu, Lagos Nigeria.
Department of Civil Engineering, Lagos State Polytechnic, Ikorodu, Lagos Nigeria.
ABSTRACT
This work used the ANFIS (Adaptive Neuro-Fuzzy Inference System) as a framework to diagnose faults in the boiler section of a power plant. A backpropagation algorithm was used in modeling the ANFIS network. Industrial boiler data were obtained from the power plant, compiled using Excel, andthe ANFIS network was then simulated using Artificial Neural Network Tool in MATLAB2015a. A GUI (Graphical User Interface) was generated to easily interpret the fault results obtained. After the simulation, the ANFIS network was tested using the industrial data and using the Graphical User Interface, it was able to identify the size, root causes, and location of the faults and gave an explanation as to the corrective measures required for all the five faults that occurred in the power plant to be remedied. When the boiler was operating at its set point/standard value no fault was observed. Boiler overheating was experienced when the temperature of overheated steam and super-heated steam pressure increased from their points. Boiler feed pump failure occurred when there was a deviation from the setpoint value of the feedwater flowrate. The boiler plant used was that ofthe Egbin power plant in Lagos.
Keywords: ANFIS, boiler plant, setpoint, explanation ability, graphical user interface, root cause.
INTRODUCTION
A fault is generally defined as the departure of an observed variable or calculated parameter from an accepted range (Barak, 2016). The monitoring of industrial processes for performance and fault detection is an essential part of the drive to improve process quality. The requirements of improved productivity, efficiency, safety, and reduced levels of manning have led to the increased investigation into fault detection and diagnosis. With the increased use of