Signal drift modeling and analysis of an intelligent instrument using a multivariate adaptive regression splines technique

Abstract

Signal drift is an important failure mode of intelligent instruments under the combined effect of complex loading conditions and usually exhibits strongly nonlinear characteristics with a high degree of scatter. To solve this problem, this paper applies a multivariate adaptive regression splines (MARS) model to flexibly fit the nonlinear and highly inconsistent signal drifts of intelligent instruments. To investigate the pressure signal drift characteristics of different test transmitters, a unified parameter extraction method is proposed that uses a four-loop recursive algorithm to obtain consistent MARS model parameters. Through an application case study on five intelligent pressure transmitters tested at two temperatures, it is verified that the proposed unified parameter extraction method works perfectly for the fitting and analysis of pressure signal drift of these transmitters. Furthermore, it has advantages in providing quantitative interpretation of signal drift scatter, with potential practical application to investigation and comparison of drift fault curves of instruments that exhibit significant inconsistency.

Keywords:

• Curve fitting
• intelligent instrument
• multivariate adaptive regression spline (MARS)
• parameter extraction
• signal drift

Acknowledgments

The authors would like to acknowledge the support of the National Key Research and Development Program of China (No. 2020YFB2009401). We are also greatly indebted to Mr. Yong Zhao from Shanghai Automation Instrumentation for his valuable suggestions and help with instrumental experiments.

Disclosure statement

No potential conflict of interest was reported by the author(s).