Subtractive clustering based fuzzy rule reduction for classification of events on petroleum products pipeline

Abstract

The quest forreductionof Fuzzy Logic (FL) rules dimension to minimize the complexity of fuzzy based systems has attracted pragmatic researches in recent time. In this paper, subtractive clustering technique is used to facilitatereduction in FL rules for classification of Petroleum Products Pipeline (PPP) events arising from significant deviations of pressure, inlet and outlet volume, temperature and flow rate from stipulated threshold of normal operations. FL modelling of PPP data was carried out using fuzzification, inference and defuzzification processes. Triangular membership function was used to map PPP attributes into linguistic terms. Adaptive Neuro-Fuzzy Inference System (ANFIS)based on Takagi Sugeno inference mechanism was used for system training, validation and testing. Subtractive clustering algorithmwith Range of Influence, Squash Factor, Accept and Reject Ratios of 0.5, 1.25, 0.5 and 0.15 respectively, was used to reduce fuzzy rule dimension from 625 to 56 rules representing about 81% reduction. The reduced-rule base was used in ANFIS training.The system was implemented using Matrix Laboratory programming tools and My Structured Query Language database. Data of PPP collected from Pipelines and Products Marketing Company, Port Harcourt, Nigeria was used to assess the functionality of the system. Training, validation and testing Mean Squared Error (MSE) values of 0.0139, 0.0058 and 0.0059 respectively, were observed in theANFIS learning process. Comparison of MSE results showed the superiority of Subtractive Clustering techniqueoverGrid Partition approach.