No abstract

No abstract

No abstract

The use of the method of biharmonic spline interpolation in the approximation of the conversion function of sensors is considered. A division of the method into two phases, which makes it possible to transfer most of the computations to the stage of calibration, is proposed. The advantages of the method by comparison with polynomial interpolation and its applicability in the case of the conversion function of several variables are demonstrated.With the development of computer engineering, smart sensors with built-in microcontroller or functional devices (systems) on a single integrated circuit on a single chip have become common. Such systems are implemented both as a digitized representation of analog information that arrives from a primary transducer as well as its subsequent transformation, processing, and transmission through a digital interface [1]. Because of the computational power of modern microcontrollers and the volume of memory elements, computations may be performed directly in the sensor in volumes that could not been accomplished previously. Today a microcontroller executes such functions as self-diagnostics, self-recovery, and remote access to the parameters of the sensor. The microcode comprises algorithms for initialization, identification, and operation in a network.Because of improvements in production processes involved in the creation of sensing elements, it has become possible to begin the production of sensors possessing several primary transducers of physical quantities on a single chip or substrate with the use of thin-film nano-and micro-electromechanical systems functioning as the sensing elements. Such sensors may simultaneously measure a basic physical quantity and the factors influencing this quantity and (or) several physical quantities at a single sampling point. Measurements of temperature to compensate the temperature error of a pressure sensor and (or) pressure and temperature at a single sampling point may serve as classical examples.The computational resources of a microcontroller may be enabled as a way of eliminating the error caused by the nonlinearity of the conversion function (calibration characteristic). In order to linearize the calibration characteristic, the conversion function is often approximated by polynomial expressions of low degrees. Such techniques as specification of a functional dependence of each of the coefficients by a one-dimensional curve of the second parameter [2] or the use of polynomial approximation of a surface [3] are used in the case of a function of two variables.The substantial error of the approximation is a drawback of such methods. The results of an approximation of a function of two variables by polynomial expressions of low degrees in the case of a large number of check points are often unsatisfactory. Increasing the degree of the polynomial leads to the appearance of oscillations in the interpolation curve, though this will not correspond to the actual dependence. In the general case, the value of a measured quantity computed by means

No abstract