T–S fuzzy‐based multi‐LAE approach for sensor linearisation

Abstract

Most sensors exhibit a non-linear input–output relationship. Sensor linearity is essential for the true representation of measurand in modern instrumentation systems. Although several linearisation techniques (LTs) have been proposed in recent years, they incur common problems of scalability, reliability, flexibility, complexity and cost. Hence, an effective software lineariser is needed to overcome these difficulties. This study presents a Takagi–Sugeno (T–S) fuzzy-based sensor LT using piecewise linearisation. The non-linear static relationship splits into ‘N’ multiple linear algebraic equations, which are combined using T–S fuzzy logic to ensure a smooth transfer from one region to another. The linearity (mean square error) of 0.1663 is obtained for negative temperature coefficient thermistor in simulation over a temperature range of 307 to 373 K, and of 0.0255 is achieved for infrared sensor in real time over a distance range of 12–30 cm. The performance of the proposed LT is compared with popular curve fitting, a look-up table and soft computing approaches. The results show that the proposed LT outperforms other approaches regarding performance indices. This technique eliminates hardware complexity rendering better accuracy over the span of measurement.