Research on in-use inspection method of ultrasonic gas flow meter based on supervised learning
Abstract
Ultrasonic flowmeters are the most commonly used instruments in natural gas trade, the in-use inspection methods can be used to verify the performance of it. The supervised learning algorithm with the digital measurement technology is utilized to investigate the method in-use of the ultrasonic flowmeter. Random Forest and BP-Artificial Neural Network are used to construct the soft-measurement models to estimate flow rate deviation. With the close loop gas flow standard facility of NIM in China, there were 6 flow points selected to conduct test experiments on the DN200 ultrasonic gas flowmeter with cross 4-paths. Together with the test data, 15 indicators are determined as the input of the model, and the data is denoised by means of Fast Fourier transform with Gaussian window function. MCM is used to assess the uncertainty. The results show that the two models can estimate well the flow rate deviation of the flowmeter, in which Random Forest model has the better result, with the advantages of high accuracy and good stability, it can effectively monitor and diagnose the performance of flowmeter during the operation. Thus, a new reference is provided for the inspection process of ultrasonic flowmeter in use.
Event details
- Event
- FLOMEKO 2022
- Technical Committee
- TC9
- lich@nim.ac.cn
- Place
- Chongqing, CHINA
- Time
- 1 November 2022 - 4 November 2022
- Website
- http://flomeko2022.msmk.tech/c/index.html