DistiSens - Distributed flow characterisation
| Reference number | |
| Coordinator | RISE Research Institutes of Sweden AB - Primär, Borås |
| Funding from Vinnova | SEK 500 000 |
| Project duration | December 2019 - November 2020 |
| Status | Completed |
Important results from the project
Within the process industry, multiple benefits are associated with a simple, robust and uninterupted method for flow characterization. A technique with adequate measurement uncertainty and advance signal processing could yield large opportunities for improved services. Such a technology would yield a excavation free alternative for digitalization of the water utilities as well. The aim has been to develop a technique for characterization of flow with distributed sensors and machine learning. In addition, a strategy for making the technology available has been developed.
Expected long term effects
The algorithm consists of a neural network predicting the flow based on two input features - the time lag of accelerometer signals and their frequency content. A flow rate is predicted with a RMSD of 13% of range. In many applications, this is not good enough, but the value of the project lies within the discoveries and experiments we have gathered, important ingredients in further development. Currently, the temperature algorithm estimates if the liquid temperature is 20 C or 40 C with an accuracy of 82 %. The impact assessment is possible when the methodology has reached a higher TRL.
Approach and implementation
The project consisted of five WPs, but the implementation can be divided into three parts. Firstly, in the theoretical part, earlier similar work was examined. The experiments were designed and suitable methods for data analysis were evaluated. Next, the experiments were performed. Part three consisted of analysis and dissemination of the results. The analysis would have been more sucessful with more data. Since additional measurements would have entailed increased people interaction, it was decided to avoid that due to the current pandemic.