Sensorless Speed Estimation of Induction Motor using ANFIS

Abstract

In modern industrial applications, particularly in the era of Industry 5.0, accurate estimation of induction motor speed is essential for ensuring optimal performance, energy efficiency, and system reliability. Traditional methods using mechanical sensors for speed measurement are often limited by increased system complexity, high maintenance costs, and vulnerability to environmental conditions. This study presents a sensor less approach for predicting the acceleration for the induction motor with ANFIS, an Adaptive Neuro-Fuzzy Inference Technology. ANFIS integrates the learning capabilities of neural networks with the inferential benefits of fuzzy logic, making it suitable for nonlinear systems and uncertain environments. The study entails the creation and training of an ANFIS model utilizing motor data. The model employs many membership functions, such as Trimf, Trapmf, Gbellmf, and Gaussmf, and the assessment utilized RMSE to evaluate their efficacy. The findings reveal that the Trimf membership function yields the highest predictive accuracy, with an RMSE of 0.0187, whilst the Trapmf function shows the lowest accuracy, with an RMSE of 0.0213. The ANFIS system successfully estimates motor speed with minimal deviation from the actual output, as observed in simulations. This sensor less approach not only reduces costs and complexity but also supports the development of intelligent, energy-efficient systems in line with Industry 5.0 objectives. Overall, the findings highlight the potential of ANFIS for advanced motor control applications, contributing to smarter automation and improved sustainability.