ePrints

A fault-tolerant flow measuring method based on PSO-SVM with transit-time multipath ultrasonic gas flowmeters

Lookup NU author(s): Dr Xiang XieORCiD

Downloads

Full text for this publication is not currently held within this repository. Alternative links are provided below where available.

Abstract

Transit-time multipath ultrasonic flow meters (TM-UFMs) have been widely employed for measuring flowrate of gas and liquid. However, its applicability is still dragged by the difficulties to describe the physical model by precise mathematical expressions and to reduce the deviations led by the simplified principle models. Besides, fault-tolerance, including fault diagnosis and measure maintaining method, namely, rectification, is of great value in practical engineering, but there are few effective methods provided ever since to provide a solution for faulty UFMs. Therefore, data integration for TM-UFMs is one of the complications to obtain the accurate measurements, except for precise transit-time detection and improved transducers and circuits. This paper proposes a novel data integration method for TM-UFM calibration, as well as fault diagnosis and rectification based on particle swarm optimized support vector machines (PSO-SVM). Besides, extensive experiments have been conducted on a platform of TM-UFMs, and the results have illustrated the effectiveness of this method. The PSO-SVM-based models trained by the proposed method lead to a decrease of deviations to ±1% (full scale) in normal state for data integration, compared with ±2% deviations when a traditional method is adopted. When there exist one or two dysfunctional acoustic paths, the method diagnoses the faulty paths and maintains the measurement with deviations below ±2%.

Publication metadata

Author(s): Tang XY, Xie X, Fan B, Sun YX

Publication type: Article

Publication status: Published

Journal: IEEE Transactions on Instrumentation and Measurement

Year: 2018

Volume: 67

Issue: 5

Pages: 992-1005

Online publication date: 05/02/2018

Acceptance date: 06/12/2017

ISSN (print): 0018-9456

ISSN (electronic): 1557-9662

Publisher: IEEE

URL: https://doi.org/10.1109/TIM.2018.2795298

DOI: 10.1109/TIM.2018.2795298

Altmetrics

Share