DEVELOPMENT OF ULTRASONIC FLOWMETER FOR ASYMMETRIC GAS FLOW VELOCITYMEASUREMENT USING ADAPTIVE WEIGHTING METHOD

• Main Author: Amri, Khairul
• Format: Dissertations
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/38313
• Institution: Institut Teknologi Bandung

A transit time ultrasonic flowmeter (USM) is a high accuracy for symmetrical and fully developed flow. AGA-9 makes the installation guidance for USM’s practitioners to ensure the symmetrical flow. A flow conditioner is used to convert asymmetric into symmetric flow with high repeatability. The USM without flow conditioning need a straight run of 50D (D=diameter) upstream the meter. Thus, it is often difficult to follow due to lack of space. Therefore, an advance USM is needed to maintain it accuracy in spite of the presence of asymmetric profiles. The USM is composed of single or multi path. The mean flow velocity for multipath USM is calculated by integrating the local velocity along all measurement paths. The path weights can be obtain by conventional or adaptive methods. The weight of each path is fixed for the measurement range and obtained by mathematical calculations in the conventional one. Meanwhile, the set of adaptive weighing in advance USM is commonly determine using supervised learning machines. A set of weighting can be view as mapping function for specific 2D flow profile. Futhermore, the mapping function is then used to calculate the mean flow velocity in the measurement phase. In order to ensure that USM is able to measure asymmetric profile flow, the conventional weighing requires the path configurations of USM are robust against the asymmetric profiles. Therefore, numerical simulation are performed to evaluate eight configurations namely, Diam, Orth, Three, Dorth, Mid, Lin, Dtri, and Tomography. The result shows that USM Tomography (USM-Tomo) 12 transducers has the best performance for asymmetric flow measurement. In addition, it is able to reconstruct flow profiles. Flow profiles can be used to analyze a disturbance as well as for further assessment and treatment of USM. However, it has a low temporal resolution due to large amount of data to be collected and the length of computation time. To reduce the amount of data and speed up computing time, this study combines FMU-Tomo with adaptive weighting methods. The set of pair ultrasound transducers is setting in multi-paths of USM-Tomo. In the training phase, all paths of USM-Tomo is used for calculate 2D flow profile and also mean flow velocity. In the same time, parallel multi-path configuration is also applied in order to determine flow velocity along the each path. For different 2D flow profile, the re-mapping is perfomed. Numerical simulations are carried out to obtain the best adaptive weighting architecture. The results show that by activating 6 parallel paths of USM-Tomo and applying Artificial Neural Network (ANN) or Support Vector Regression (SVR) are numerically able to maintain UMS errors below 1%. Besides that, both methods are immune to 35 dB of Gaussian noise. Experimental validation of the proposed method is conducted by making two experimental setups. Both of them are designed to produce asymmetric flow profiles with various forms. In experiment 1, a butterfly damper with adjusting blade angle was employed to produce asymmetric profiles in the straight pipe and the measurements take place at 20D downstream of damper. In experiment 2, 4 different shapes of obstructionswere placed in at 1D after outlet elbow and the measurements takes place at some point from the obstructiondownstream; 3D, 5D and 13D. In addition, variable air speed is generated by a 3 phase centrifugal blower coltrolled by an inverter. The result of USM-Tomo is compared with hot-wire anemometer because both instruments are able to reconstruct profiles and calculate mean flow velocity. The result of experiments show that the reconstruction profiles of USM-Tomo successfully demonstrate several phenomena; 1. Vortex and turbulence flow with asymmetricprofiles when gas flows through a 900pipe bend, 2. A natural barrier in outlet blower, 3. Successfully identified the obstructions from profile made. USM’s measurement error is from 0 to 14% with average 6% compared to hot-wire anemometer for gas speeds in range 0.8-3.9 m/s. In addition, the RMSE of ANN and SVR are far lower than AGA-9 which means the presence of small error predictions. Dual-transducers method is introduced for data collection strategy. This technique is able to record the two transit time in each path at the same time. By activating 6 path of USM-Tomo and using dual-transducers, the sampling period is about ± 7 ms and computation time are around 10-30 ms for SVR and 50-90 ms for ANN. Thus, the proposed method is possible an alternative solution to obtain USM system which has high temporal resolution, good accuracy for symmetric and asymmetric flow, and easily adjusted in different circumstances.