چكيده به لاتين
Abstract:
This thesis discusses the design and fabrication of a domestic ultrasonic natural gas flow computer. It is then calibrated based on the Multilayer Perceptron Neural Network (MLANN). The measured flow rate ranges from 0.2 to 4 m3 / h. Due to the transient flow in a considerable part of the range of the above flow rate and the low flow velocity in the laminar flow region, nonlinear effects are created on the flow rate measurement and cause a bias error that varies with fluid flow. Therefore, in order to reduce bias error and increase the accuracy of the flow meter measurement, the use of neural network in the flow meter calibration is considered due to its capabilities in nonlinear mapping. The neural network training is based on the input data (the outputs of the fabricated flow computer) and the target data (which corresponds to the outputs obtained by the Bell Prover). Also, the sing-around method was used to measure the time of flight of ultrasound signals with a higher resolution than the microcontroller resolution. The purpose of this thesis is to achieve an error less than 1.5%. To evaluate the proposed method, two-point calibration methods as well as the bracket calibration method were used, and the results of all three methods were compared. Since calibration equations are, in effect, a mapping of flowmeter readings to prover or master meter reading, the aim is to achieve a method that is capable of predicting or mapping the flow computer output with a lower error than usual. Although the two-point calibration method shows a good statistical agreement between the flow computer outputs and the bell prover, it failed to reach that target. On the other hand, in calibration by the bracketing method, measurement error was slightly better than 1.5%, but, using multilayer perceptron neural network method in calibration, the maximum measurement error was 1.21%, which was the best result obtained. This suggests that the use of artificial intelligence as an efficient calibration tool is worth further work and research.
In transit-time ultrasonic flow meters, one of the most common methods of measuring the time of flight of ultrasonic signals in the time domain is the cross-correlation method, which requires a great deal of time and computation. Hence, time difference approaches are preferred to reduce the computational burden and improve the response time. However, the time-difference method suffers from an additional error (an error in addition to the system error imposed by this method) compared to conventional methods. In this thesis, this additional error is analyzed using numerical simulation, and the effects of some factors such as pipe diameter, flow rate and fluid temperature are investigated, and it is shown that this additional error is a bias error (additional bias error-ABE). Furthermore, an approach to remove or reduce ABE is presented. Since it is mentioned in the literature that this method can only be used for Mach numbers smaller than 0.1, it is shown that by using the second-order approximation of the explicit equation, this method can be used for Mach numbers larger than 0.1, and the ABE of this method was reduced to approximately zero. In this regard, an experiment was carried out on a six-inch ultrasonic flow meter with an eight-path X-configuration to validate the second-order approximation method. The results of the experiment confirm that the proposed method can be successfully used in ultrasonic flow meters with a flow rate greater than Mach 0.1.
