CN-121994331-A - Transient flow modeling generation method and transient flow generation and measurement device

CN121994331ACN 121994331 ACN121994331 ACN 121994331ACN-121994331-A

Abstract

The invention relates to the field of fluid flow measurement, and discloses a transient flow modeling generation method and a transient flow generation and measurement device, wherein the device is a computer-controlled electromechanical integrated system and is used for generating transient flow with standard magnitude characteristics, measuring characteristics of a flowmeter under the generated transient flow, repeatedly testing the flowmeter for a plurality of times under different transient flow models by utilizing the transient flow generation and measurement device, collecting the transient flow measured by the flowmeter to form time sequence data, and synchronously collecting the transient flow measured by the transient flow generation and measurement device to form the time sequence data. These time series data are built into a labeled dataset for transient flow pattern recognition model training. Transplanting the trained transient flow pattern recognition model into a flowmeter, and endowing the flowmeter with a transient flow response function.

Inventors

ZHANG NINGNING
ZHAO JIANLIANG
CHEN GUOFU
WU XIAOJIE
FANG YAN
GENG XUEMEI
Yin Zhuofan

Assignees

浙江省质量科学研究院

Dates

Publication Date: 20260508
Application Date: 20260119

Claims (9)

1. A transient flow generation and measurement device for measuring or training transient flow characteristics of a flowmeter comprises an upper computer (11), a main controller (10), a servo driver (9), a servo motor (7), a rotary encoder (8), a speed reducer (6), a ball screw (5), a piston (4) and a piston cylinder body (3), and is characterized in that the inner end surface of the piston (4) extending into the piston cylinder body (3) is a sealing surface, the end surface positioned at the outer side of the piston cylinder body (3) is a non-sealing surface and is coaxially matched with the ball screw (5), the ball screw (5) is connected with an output shaft of the speed reducer (6) and is used for converting rotary motion of the servo motor (7) into linear motion of the piston (4), an input shaft of the speed reducer (6) is connected with a main shaft of the servo motor (7), the tail end of the rotary main shaft is connected with the rotary encoder (8), the rotary encoder (8) measures the rotary speed of the servo motor (7) from the main shaft of the servo motor (7) and transmits the rotary speed to the main controller (10) in a pulse signal mode, the servo driver (9) is connected with the main controller (10) through a communication port and is connected with the main controller (10) to receive the rotary speed command of the main controller (10) and the main controller (10) to control the rotary speed command of the main controller (7), and collecting working state parameters of the servo motor (7) and feeding back the working state parameters to the main controller (10).
2. The transient flow generating and measuring device according to claim 1, wherein the main controller (10) applies a time mark to the acquired data in the process of acquiring a pulse signal output by the rotary encoder (8) or acquiring a flow signal output by the tested flowmeter to form time series data, and the time series data are transmitted to the upper computer (11).
3. The transient flow generating and measuring apparatus of claim 1, wherein said upper computer (11) is provided with a transient flow generating module, a data processing module and a transient flow pattern recognition model training module.
4. The transient flow generation and measurement apparatus of claim 3, wherein said transient flow generation module generates 6 basic transient flow models: the 1 st is a valve-opening flow model, Wherein q (T) is the instantaneous flow of the transient flow, T is a time variable and is not a negative number, q max is the upper limit instantaneous flow and is a non-negative constant parameter, a and b are constant parameters which are larger than 0, valve opening time T o is set, q (0) is approximately equal to 0 when t=0, an approximate value of b is calculated, q (T o )≈q max is calculated when t=T o , and an approximate value of a is calculated; the 2 nd is a valve closing flow model, Wherein q (T) is the instantaneous flow of the transient flow, T is a time variable and is not a negative number, q max is the upper limit instantaneous flow and is a very negative parameter, a and b are constant parameters which are larger than 0, the valve closing time T c is set, q (0) is approximately equal to q max when t=0, the approximate value of b is calculated, q (T c ) is approximately equal to 0 when t=T c , and the approximate value of a is calculated; The 3 rd one is a pulsating flow model, Wherein q (T) is the instantaneous flow of the transient flow, T is a time variable and is not a negative number, q m is the average instantaneous flow and is a very negative constant parameter, c is a constant parameter from 0 to 1, when c=0, the model is degraded into a stable flow, and T is a constant parameter with a pulsation period greater than 0; the 4 th is a linearly increasing flow model, Wherein q (t) is the instantaneous flow of the transient flow, t is a time variable, the value range of t is 0 to tau i ;q min is the lower limit flow of the increment initial time, the non-negative normal parameter, q max is the upper limit flow of the increment end time, the non-negative normal parameter, d is the normal parameter of 0 to 1, and tau i is the normal parameter of which the increment time length is greater than 0; The 5 th is a linearly decreasing flow model, Wherein q (t) is the instantaneous flow of the transient flow, t is a time variable, the value range of t is 0 to tau d ;q max is the upper limit flow at the decreasing initial time, the non-negative normal parameter, q min is the lower limit flow at the decreasing end time, the non-negative normal parameter, d is the normal parameter of 0 to 1, and tau d is the normal parameter with decreasing time length being more than 0; the 6 th is a composite wave flow model, Where q (T) is the instantaneous flow of the transient flow, T is a time variable and is not a negative number, q 0 is the initial instantaneous flow of the transient flow at the time of t=0 and is a very negative parameter, T is the fundamental flow period when k=1 and is a constant parameter, q k is the amplitude of the kth harmonic flow and is a very negative parameter, and n is a positive integer representing the number of composite waves.
5. The transient flow generation and measurement apparatus of claim 4, wherein said 6 basic transient flow models are used alone or in free combination to form a combined transient flow model.
6. The transient flow generating and measuring device according to any one of claims 3 to 5, wherein the transient flow generating module digitizes a transient flow model, converts the transient flow model into a motion control parameter of a servo motor, the motion control parameter is issued to the main controller (10) by the upper computer (11), the servo driver (9) is controlled by the main controller (10) to drive the servo motor (7) to rotate, the piston (4) is pushed to move, the fluid in the piston cylinder (3) is discharged to generate transient flow, and the rotary encoder (8) measures a rotation signal of the servo motor (7) and feeds the rotation signal back to the main controller (10).
7. The transient flow generation and measurement apparatus of any of claims 3-5, wherein said data processing module performs the following: Calculating the actual instantaneous flow of the transient flow according to the frequency of the pulse signal output by the rotary encoder with the time mark and collected by the main controller (10), fitting the actual instantaneous flow curve of the transient flow according to the generated transient flow model, and adopting a goodness-of-fit index The actual transient flow curve fitting effect is evaluated, wherein q i is a transient flow sample, To fit the instantaneous flow estimate of the curve, Is the instantaneous flow sample mean value; Calculating the generated transient flow model according to time integration to obtain a theoretical accumulated flow, calculating the actual accumulated flow of the transient flow according to the accumulated pulse number of the pulse signals output by the rotary encoder with the time mark and acquired by the main controller (10), dividing the difference between the actual accumulated flow and the theoretical accumulated flow by the theoretical accumulated flow to obtain the comprehensive error caused by the digitization and the motion control of the transient flow model; fitting an actual instantaneous flow curve of the tested flowmeter according to the generated instantaneous flow model according to the flow signal output by the tested flowmeter with the time mark and acquired by the main controller (10), and adopting a goodness-of-fit index Evaluating the fitting effect of the instantaneous flow curve of the flowmeter, wherein q mi is the instantaneous flow sample of the flowmeter, Fitting a curve of the instantaneous flow estimate to the flow meter, Is the average value of the instantaneous flow sample of the flowmeter, And calculating according to the flow signal which is acquired by the main controller (10) and is output by the tested flowmeter with the time mark to obtain the flowmeter accumulated flow, subtracting the actual accumulated flow of the transient flow from the flowmeter accumulated flow, and dividing the actual accumulated flow of the transient flow by the actual accumulated flow of the transient flow to obtain the transient flow measurement error of the tested flowmeter.
8. The transient flow generating and measuring device according to claim 3, wherein the transient flow pattern recognition model training module is constructed by adopting a Convolutional Neural Network (CNN) based model, a cyclic neural network (RNN) model, a long-short-term memory neural network model (LSTM) based model, or a Bi-directional cyclic neural network model (Bi-RNN) based model, and performs transient flow pattern recognition model training of the flowmeter by utilizing a labeled data set based on a time sequence, and the transient flow pattern recognition model obtained through training is migrated into the flowmeter.
9. A modeling generation method of transient flows is characterized by defining 6 basic transient flow models, comprising the following steps: A valve flow opening model is adopted to realize the flow opening model, Wherein q (T) is the instantaneous flow of the transient flow, T is a time variable and is not a negative number, q max is the upper limit instantaneous flow and is a non-negative constant parameter, a and b are constant parameters which are larger than 0, valve opening time T o is set, q (0) is approximately equal to 0 when t=0, an approximate value of b is calculated, q (T o )≈q max is calculated when t=T o , and an approximate value of a is calculated; the valve flow model is closed and the valve flow model is opened, Wherein q (T) is the instantaneous flow of the transient flow, T is a time variable and is not a negative number, q max is the upper limit instantaneous flow and is a very negative parameter, a and b are constant parameters which are larger than 0, the valve closing time T c is set, q (0) is approximately equal to q max when t=0, the approximate value of b is calculated, q (T c ) is approximately equal to 0 when t=T c , and the approximate value of a is calculated; A pulsating flow model is provided, wherein, Wherein q (T) is the instantaneous flow of the transient flow, T is a time variable and is not a negative number, q m is the average instantaneous flow and is a very negative constant parameter, c is a constant parameter from 0 to 1, when c=0, the model is degraded into a stable flow, and T is a constant parameter with a pulsation period greater than 0; a linearly increasing flow model is used, Wherein q (t) is the instantaneous flow of the transient flow, t is a time variable, the value range of t is 0 to tau i ;q min is the lower limit flow of the increment initial time, the non-negative normal parameter, q max is the upper limit flow of the increment end time, the non-negative normal parameter, d is the normal parameter of 0 to 1, and tau i is the normal parameter of which the increment time length is greater than 0; a linearly decreasing flow model is used, Wherein q (t) is the instantaneous flow of the transient flow, t is a time variable, the value range of t is 0 to tau d ;q max is the upper limit flow at the decreasing initial time, the non-negative normal parameter, q min is the lower limit flow at the decreasing end time, the non-negative normal parameter, d is the normal parameter of 0 to 1, and tau d is the normal parameter with decreasing time length being more than 0; a composite wave-stream model is provided, Wherein q (T) is the instantaneous flow of the transient flow, T is a time variable and is not a negative number, q 0 is the initial instantaneous flow of the transient flow at the moment of t=0 and is a very negative parameter, T is a fundamental wave flow period of k=1 and is a constant parameter larger than 0, q k is the amplitude of the kth harmonic flow and is a very negative parameter, and n is a positive integer and represents the number of synthesized waves; the transient flows are generated by 6 basic transient flow models alone or in free combination.

Description

Transient flow modeling generation method and transient flow generation and measurement device Technical Field The invention relates to the field of fluid flow measurement, provides a fluid transient flow generation method and provides a transient flow generation and measurement device for measuring and evaluating the metering characteristics of a flowmeter based on the fluid transient flow generation method. Background Traditionally, the magnitude of the flow meter is defined above the steady state flow. Steady state flow, also called steady state flow, is characterized by a flow parameter of the fluid that does not change over time at any spatial location. Flow measurement standard devices commonly established in laboratories around the world are based on testing and evaluating the metering characteristics of a flowmeter in a stable fluid flow state. The flow meter works under steady state flow created in a laboratory, the magnitude of a flow measurement result is represented by a mean value, the flow meter has the characteristics of easy measurement and reproducibility, measurement noise can be easily identified by means of a mean value and variance statistics method on a measurement data processing method, and the measurement noise is removed by a filtering method, so that a very high measurement accuracy index is obtained. Laboratory establishment of steady-state flow-based measurement standard devices is essential for flow magnitude definition, assignment and meter characterization. In the real world, there is very little steady state flow comparable to a laboratory, and flowmeters are instead commonly operated under transient flow conditions. Transient flows, also known as unsteady flows, are characterized by a change in flow parameters of the fluid at any spatial location over time. For example, the flow rate of the fluid after the valve is opened is in a transient change process from small to large, the flow rate of the fluid after the valve is closed is in a transient change process from large to small, the fluid flow at the outlet of the pump has strong pulsatility, the common fluid filling or consuming process flow of a factory is usually in an intermittent transient change process, and the filling or filling of the fuel such as compressed hydrogen, compressed natural gas, liquefied natural gas and the like is in a transient change process from small to large and then from large to small. In transient flow scenarios where dynamics have non-stationary characteristics, one problem to be solved is if the metering characteristics of the flow meter operating at transient flow can be equated with steady state flow. The flow meter works under transient state to carry out discretization measurement, the obtained measurement result is a group of time series data, the statistical characteristic of the time series data changes along with the time, the front and back data points always show autocorrelation, the measurement result is not represented by a mean value, and random errors can not be distinguished only by a mean value and variance statistical method, so that a filtering algorithm based on steady state flow can not adapt to transient state flow conditions. The more suitable method is to perform curve fitting on time series data, and the fitting result is used as the estimation of actual flow, so that the measurement error of the flowmeter can be greatly reduced, and the metering performance of the flowmeter under transient flow is improved. Therefore, the flowmeter only gives metering characteristics to the flow through the steady-state flow, and when the transient response algorithm is not provided, the characteristic performance of the flow under the transient state is obviously different from that of the steady-state flow, and larger measurement errors can be generated. In fact, it is a common practice for the maximum allowable measurement error of a flowmeter engineering field measurement application to be amplified by a factor of 2 over the maximum allowable measurement error under laboratory steady state flow conditions. Thus, when a flow meter needs to accommodate transient flow measurement scenarios, another problem that needs to be solved is how to give the flow meter a suitable transient response algorithm. The time series transient flow sampling data also faces the complexity problem of real world transient flows in the curve fitting process, namely the selection problem of a fitting curve model. A feasible scheme is to decompose complex transient flows into characterizable basic transient flows and combinations thereof, perform pattern recognition on transient flow sampling data by using an artificial intelligence technology, find a matched fitting curve model, and further obtain a good curve fitting effect. One problem that needs to be addressed during the application of artificial intelligence techniques is the training of transient flow pattern recognition models. The basic transient