DEVELOPMENT OF A NEW FUZZY BASED NEURAL NETWORK AND ITS APPLICATION IN WATER FLOW CONTROL IN CANAL

Abstract

A novel method is being proposed in this thesis named Fuzzy Neural Network (FNN) and it used to predict the inlet to outlet key width’s ratio of the PKW. In FNN, the fuzzy logic and Neural Network (NN) is used and combined for the benefits of both. Fuzzy logic is not like traditional binary system where it tells the results in 0 and 1, but the fuzzy logic deals with the degree of membership where it tell the degree of truth, just like the probability, the values of fuzzy logic lies between 0 and 1. NN is inspired by the biological neurons in humans, as biological neurons are interconnected and used to transmit information from one point to another in human brain, similarly the NN is the collection of well-connected neurons which are used to process the information in different ways, it can be used to predict the values in regressive manner, or can be used to classify in different categories and also can be used to recognize patterns and analyse those, in similar ways NN has multiple benefits. The FNN model combines the benefits of both and allows to predict the hydraulic behaviours of PKW with very high accuracy. The dataset used in this thesis is not pre-defined, it is used collected experimentally from the laboratory. Different values were recorded for different ratios of inlet to outlet key width and then that dataset is used to test the FNN model for the prediction. The dataset includes crucial data that is essential for understanding the hydraulic performance of PKWs, including energy dissipation and discharge flow rates. The popular metrics used for the validation of the model are RMSE and MAE. The RMSE of 0.0305 and MAE of 0.0222 showcased the FNN model's exceptional accuracy and reliability. As the values for these metrics fall in the ideal range , it depicts the accurate prediction of the model. These findings tells that the relevance of FNN model can go beyond the predictions for PKW. It can be applied in the multiple fields of problem solving and pattern recognition. It can offer useful insights in different sectors of engineering.