Search

CN-122016016-A - Mass flow calibration method and system

CN122016016ACN 122016016 ACN122016016 ACN 122016016ACN-122016016-A

Abstract

The invention discloses a mass flow calibration method and a mass flow calibration system, which comprise the steps of installing and connecting a Coriolis mass flowmeter to a calibration rack pipeline, installing an intelligent sensor comprising an ultrasonic Doppler array on the upstream side along the axial direction, starting a liquid supply pump to pump calibration fluid, controlling the ultrasonic Doppler array to perform section scanning on the fluid to obtain dynamic flow velocity section time sequence data, inputting the dynamic flow velocity section time sequence data to a pre-trained fluid relaxation state prediction machine learning model, outputting a steady state arrival time point by the fluid relaxation state prediction machine learning model, triggering a calibration sampling instruction at the steady state arrival time point, obtaining a sampling mass flow value and comparing the sampling mass flow value with a reference standard flow value to finish calibration. The invention solves the problem that systematic errors exist in calibration sampling caused by unsteady flow pattern change in the recovery period of high-viscosity fluid in shear thinning.

Inventors

  • FAN MINGJUN
  • Fan Huiwei
  • GAO HE

Assignees

  • 北京均友欣业科技有限公司

Dates

Publication Date
20260512
Application Date
20260414

Claims (10)

  1. 1. A method of mass flow calibration comprising the steps of: attaching a coriolis mass flowmeter to a calibration rig line and mounting a smart sensor comprising an ultrasonic doppler array along an axial direction of the calibration rig line on an upstream side of the coriolis mass flowmeter; Starting a liquid supply pump to continuously pump calibration fluid into the calibration rack pipeline, and controlling the ultrasonic Doppler array in the intelligent sensor to perform cross-section scanning on the calibration fluid to obtain dynamic flow velocity cross-section time sequence data; Inputting the dynamic flow velocity profile time sequence data into a pre-trained fluid relaxation state prediction machine learning model, and outputting a steady state arrival time point by the fluid relaxation state prediction machine learning model; triggering a calibration sampling instruction at the steady state arrival time point, acquiring a sampling mass flow value from the coriolis mass flowmeter, and comparing the sampling mass flow value with a corresponding reference standard flow value to complete the calibration of the coriolis mass flowmeter.
  2. 2. The mass flow calibration method of claim 1, wherein the fluid relaxation state prediction machine learning model includes a spatiotemporal feature extraction network and a state regression output layer, the spatiotemporal feature extraction network comprising a three-dimensional convolutional neural network and a long-term short-term memory network, the step of inputting the dynamic flow velocity profile time series data into a pre-trained fluid relaxation state prediction machine learning model, outputting steady-state arrival time points from the fluid relaxation state prediction machine learning model comprising: Inputting the dynamic flow velocity profile time sequence data into the three-dimensional convolutional neural network to execute convolutional operation and pooling operation, and outputting a spatial feature sequence; Inputting the space feature sequence into the long-term and short-term memory network to extract time dependency relationship and outputting a space-time fusion feature vector; and controlling the state regression output layer to execute full-connection mapping operation on the space-time fusion feature vector, and outputting the steady state arrival time point.
  3. 3. The mass flow calibration method of claim 1, wherein the training process of the pre-trained fluid relaxation state prediction machine learning model comprises: acquiring multiple groups of historical dynamic flow velocity profile time sequence data under different output pressure conditions of a liquid supply pump; Recording the time when ten continuous sampling periods of the flow velocity fluctuation variance are smaller than a preset variance threshold as an actual steady-state time point by continuously monitoring the flow velocity fluctuation variance of the historical dynamic flow velocity profile time sequence data; Taking the historical dynamic flow velocity profile time sequence data as sample input data, taking the actual steady state time point as sample tag data, and constructing a training data set; Inputting the sample input data into an initial machine learning model to obtain a prediction time point, and calculating a mean square error value between the prediction time point and the sample label data; And reversely transmitting the mean square error value as a loss function by adopting a reverse propagation algorithm, and updating the weight parameter of the initial machine learning model until the mean square error value reaches a preset error lower limit threshold value to obtain the pre-trained fluid relaxation state prediction machine learning model.
  4. 4. The mass flow calibration method of claim 1, wherein the step of comparing the sampled mass flow value to a corresponding reference standard flow value to complete the calibration of the coriolis mass flowmeter comprises: reading a standard mass flow value generated by the static weighing scale as the standard flow value at the steady state arrival time point; dividing the reference standard flow value by the sampling mass flow value, and obtaining an instrument calibration coefficient through a division operation result; The meter calibration coefficients are written to a control memory unit of the coriolis mass flowmeter.
  5. 5. The mass flow calibration method of claim 1, wherein a static stress decoupling step is performed prior to the starting of the feed pump, the static stress decoupling step comprising: maintaining the calibration fluid inside the coriolis mass flowmeter in a stationary state; controlling an external mechanical vibration exciter to apply a single mechanical pulse to the calibration rack pipeline; Controlling an electromagnetic sensor inside the Coriolis mass flowmeter to collect a free vibration amplitude sequence of a measuring tube, and extracting an envelope line logarithmic decrement of the free vibration amplitude sequence as a free vibration decrement; Multiplying the free vibration attenuation rate by a preset mechanical damping mapping matrix to output a structural damping increment; Mapping and converting the structural damping increment into a mechanical stress pseudo-phase difference through a preset structural dynamics lookup table; subtracting the mechanical stress pseudo-phase difference from an initial zero parameter of the coriolis mass flowmeter, and outputting an updated zero reference.
  6. 6. The mass flow calibration method of claim 5, wherein said step of mapping said structural damping delta to a mechanical stress pseudo-phase difference via a predetermined structural dynamics look-up table comprises: performing Fourier transformation on the free vibration amplitude sequence to obtain a frequency domain signal sequence, and extracting the frequency corresponding to the maximum amplitude peak value from the frequency domain signal sequence as the natural resonance frequency of the measuring tube; Searching phase shift angle values which correspond to the natural resonant frequency and the structural damping increment in the structural dynamics lookup table; and setting the found phase shift angle value as the mechanical stress pseudo phase difference.
  7. 7. The mass flow calibration method of claim 5, wherein the step of obtaining a sample mass flow value from the coriolis mass flowmeter comprises: Inputting the updated zero point reference into a phase difference operation module of the coriolis mass flowmeter at the steady state arrival time point; and controlling the phase difference operation module to apply the updated zero reference to execute internal subtraction correction operation and output the sampling mass flow value.
  8. 8. The mass flow calibration method of claim 5, wherein the fluid relaxation state prediction machine learning model comprises a spatiotemporal feature extraction network, a cross-attention mechanism layer, and a state regression output layer; when the state that the delivery temperature or the delivery pressure of the calibration fluid is continuously changed is obtained, the dynamic flow velocity profile time sequence data is input into the space-time feature extraction network to output a space-time fusion feature vector, and the free vibration attenuation rate is used as a structure reference feature vector to be input into the cross-attention mechanism layer; Controlling the cross attention mechanism layer to execute attention weight distribution operation between the structural reference feature vector and the space-time fusion feature vector, and isolating structural damping drift components caused by continuous change of the conveying temperature or the conveying pressure; And controlling the state regression output layer to execute full-connection mapping operation after eliminating the structural damping drift component, and outputting the steady state arrival time point.
  9. 9. The mass flow calibration method of claim 8, wherein the step of controlling the cross-attention mechanism layer to perform an attention weight distribution operation between the structural reference feature vector and the spatiotemporal fusion feature vector comprises: Multiplying the structural reference feature vector by a first weight matrix to output a query matrix; Multiplying the space-time fusion feature vector by a second weight matrix and a third weight matrix respectively to output a key matrix and a value matrix; calculating dot products of the query matrix and the key matrix, and outputting an attention score matrix; Multiplying the attention score matrix with the value matrix, and outputting an adjusted space-time fusion feature vector; and executing the full-connection mapping operation by the state regression output layer by utilizing the adjusted space-time fusion feature vector.
  10. 10. A mass flow calibration system for implementing the method of claim 1, comprising: A mounting module for mounting and connecting a coriolis mass flowmeter to a calibration rig pipeline and mounting an intelligent sensor comprising an ultrasonic doppler array along an axial direction of the calibration rig pipeline on an upstream side of the coriolis mass flowmeter; The data acquisition module is used for starting a liquid supply pump to continuously pump calibration fluid into the calibration rack pipeline, controlling the ultrasonic Doppler array in the intelligent sensor to perform cross-section scanning on the calibration fluid, and acquiring dynamic flow velocity cross-section time sequence data; The model processing module is used for inputting the dynamic flow velocity profile time sequence data into a pre-trained fluid relaxation state prediction machine learning model, and outputting a steady state arrival time point by the fluid relaxation state prediction machine learning model; And the execution module is used for triggering a calibration sampling instruction at the steady-state arrival time point, acquiring a sampling mass flow value from the coriolis mass flowmeter, and comparing the sampling mass flow value with a corresponding reference standard flow value to finish the calibration of the coriolis mass flowmeter.

Description

Mass flow calibration method and system Technical Field The invention relates to the field of metering sensing and data processing, in particular to a mass flow calibration method and a mass flow calibration system. Background The coriolis mass flowmeter is a high-precision metering device capable of directly measuring the mass flow of fluid, and is widely applied to industrial production flow control and bulk trade handover settlement. The core measurement principle is that coriolis force generated when fluid flows in a vibrating measuring tube is utilized, the force can force the measuring tube to generate tiny torsion, and the real mass of the fluid can be converted by detecting the vibration phase difference at two ends of the tube body. In order to ensure absolute accuracy of the measurement, the meter must be mounted on a dedicated calibration stand for continuous fluid flow through the meter and for comparison of its measurement readings with a high accuracy standard at the time of shipment or periodic inspection. However, with the increasing demands of industry for metering accuracy of high viscosity fluids (such as crude oil, high molecular polymers, special calibration oils, etc.), the existing real-flow calibration techniques gradually expose multiple deep physical defects. The flow conditions of high viscosity fluids in the calibration line are extremely complex. To overcome the line resistance, calibration systems typically require the use of high-powered fluid pumps to forcibly pump fluid into the tubing. After the high-viscosity fluid is subjected to strong physical shearing by the water pump impeller, the internal microscopic macromolecular chain structure of the high-viscosity fluid is briefly broken or arranged in parallel, so that the viscosity of the fluid is instantaneously reduced, namely the shear thinning phenomenon occurs. When these fluids leave the water pump and enter the straight line of the calibration rig and flow through the calibrated flowmeter, the microstructure of the fluid is in an unstable transition where it is re-entangled, striving to recover the original viscosity. During this relaxation recovery period, the fluid viscosity recovery velocity at the center of the tube cross section and near the tube wall is not consistent, resulting in extremely complex and unpredictable spatiotemporal dynamic distortions of the flow velocity profile of the different liquid layers within the tube. Since coriolis forces are extremely sensitive to the mass distribution and flow rate of the fluid within the tube, if calibration sampling is performed in an unsteady flow field where such internal flow patterns are not yet stable, but are dynamically distorted, the calibrated instrument coefficients will have severe transient deviations. When a user installs such a flowmeter with a deviation back into service under steady-state conditions on site, significant systematic metering errors must occur. The existing calibration method generally can only set a fixed waiting time by experience, or can only blindly judge by a single average flow rate, and can not accurately capture the critical point that the microstructure of the fluid really reaches a physical steady state. In addition, the mechanical clamping process of the flowmeter to the calibration stand itself may introduce concealment errors. In actual installation, technicians usually fix the flowmeter on the test pipeline through the flange and the bolts, and in different installation processes, the tightening moment of the bolts, the stress distribution among the bolts and the contact state of the flange are often difficult to be completely consistent. For coriolis mass flowmeters, the measuring tube is in vibration mode and is sensitive to external boundary constraint changes, so that this clamping difference is not simply an installation deviation, but changes the stress state, boundary stiffness and structural damping of the measuring tube. It can be like a tuning fork, and when the fixing degree of the tuning fork bottom is different, the vibration frequency, the vibration damping characteristic and the vibration symmetry are changed, and similarly, the vibration characteristic of the measuring tube is changed when the flowmeter is clamped to different degrees. In particular, in the no-flow state, the vibration response of the two sides of the measuring tube should be kept symmetrical in theory, the theoretical value of the coriolis phase difference caused by the mass flow of the fluid should be zero, but in the case of uneven stress of the two sides of the flange, etc., the dynamic response of the two sides of the measuring tube may not be kept ideal symmetry any more, so that the flow meter still has a tiny residual phase difference, namely a pseudo phase difference, in the no-flow state. The traditional full-pipe still water zeroing method is usually only used for zeroing the current output result, and cannot identify whether the