Identification of a significant systematic error in pressure sensor readings based on an autoregressive model

Abstract

A new method of identifying the validity of pressure sensor readings has been developed. The method uses duplication of measurements which allows an estimation of the magnitude of an error, although it does not make it possible to establish which sensor is responsible for the error. The method helps to evaluate the systematic error magnitude and to test whether the error exists within a permission range accounting for a correlation structure of measurement series. An autoregressive model with a drift coefficient is applied to investigate a time series of the differences in readings. To test the significance of this coefficient, a modified Student?s test is used. Unlike the standard Student?s test, this new method tests an interval hypothesis. The null hypothesis assumes the systematic error is within the range and the alternative is out of the range. Error probabilities of type I and type II are calculated. An example of 2nd order autoregressive model was considered and the sensitivity of the proposed method was investigated.