Sensors are widely used in industrial processes, automobiles, robotics, and other systems to monitor and control the system behaviors. Capacitive Sensors, because of their high sensitivity and low power consumption are extensively used in various applications to measure pressure, force, position, speed, acceleration, liquid level, dielectric properties and flow of material. Usually, the drawback of all sensors is that, their response characteristics are influenced by the disturbing environmental parameters, e.g. temperature, humidity and pollution. For example, in the case of Capacitive Pressure Sensor (CPS) its response depends not only on the applied pressure but also on the environmental temperature. This problem becomes severe, especially when the capacitive sensor is operated in a harsh environment where temperature variation is large. Usually, an exact mathematical model of a sensor to show the relationship between the environmental parameters and the sensor output is not analytically available. Furthermore, since most sensors exhibit some degrees of nonlinear response characteristics, and the environmental parameters influence the sensor behavior nonlinearly, the problem of obtaining an accurate sensor model becomes more complex. The main objective of this research is to design an identifier and also a compensator for CPS regarding to variation of the environmental temperature. Firstly, the identifier and compensator are designed based on Support Vector Machine for Regression (SVR) and the simulation and experimental results are depicted with two kernels which are gaussian and polynomial kernels. Results show the satisfactory performance of the proposed identifier and compensator. Then, to improve the accuracy and performance of the proposed method, Particle Swarm Optimization (PSO) algorithm is applied to optimize the parameters of SVR. Additionally, in comparison with the result of identification and compensation based on SVR, the result of the hybrid SVR-PSO show better performance. Finally, another identifier and compensator are designed based on the two most prominent neural networks. The simulation and experimental results are shown. Comparing all the proposed and mentioned methods illustrate that the results of hybrid SVR-PSO with Gaussian kernel have the best performance in identification and compensation of CPS sensor. Key words: Capacitive Pressure Sensor (CPS), Support Vector Regression, Identification, Compensation, Particle Swarm Optimization, Neural Network.
