A method to improve the flow measurement and control accuracy of thermal micro mass flowmeter

Abstract

The present invention provides a method for improving the flow measurement and control accuracy of a thermal micro-mass flowmeter, which relates to the field of flow measurement technology. The method includes obtaining gas flow signal data collected by the thermal micro-mass flowmeter, and using a multi-order feedback integer infinite impulse response filter to filter the signal data; performing data calibration fitting on the filtered signal data, and screening out a surface set under the required working condition from the fitting result according to the actual working condition; based on the surface set, calculating the corresponding flow value according to the real-time collected signal data, and sending a valve control instruction to the thermal micro-mass flowmeter based on the flow value to achieve control of the gas flow. The present invention effectively improves the measurement accuracy, response speed and anti-interference ability of the thermal micro-mass flowmeter through innovative filtering algorithms, fitting techniques, model construction and data processing strategies, optimizes data processing and storage, enhances the adaptability to working conditions and reduces the lag effect of valve control instructions.

Description

Technical Field

[0001] The invention relates to the technical field of flow measurement, and in particular to a method for improving the flow measurement and control accuracy of a thermal micro-mass flowmeter. Background Art

[0002] Accurately measuring and controlling gas flow is critical in industrial automation, energy management, environmental monitoring, many precision manufacturing, photovoltaic and semiconductor manufacturing fields. Traditional flow meter technologies, such as turbine flow meters, orifice flow meters and ultrasonic flow meters, play an important role in a wide range of applications, but they have significant limitations when faced with extreme conditions such as high temperature, high pressure, low flow rate or very small flow.

[0003] In a low flow rate environment, the turbine flowmeter has difficulty in accurately measuring small flow rates due to insufficient gas kinetic energy, which causes the turbine speed to be too low. It is also easily affected by changes in gas viscosity and density, and has poor measurement accuracy and stability. When measuring small flow rates, the orifice flowmeter has a large pressure loss and high energy consumption, and its measurement accuracy has extremely strict installation requirements. A small installation deviation will cause a large measurement error. When an ultrasonic flowmeter is used under conditions where the gas medium is complex and the temperature and pressure fluctuate greatly, the ultrasonic propagation characteristics change, and the signal attenuation and interference increase, resulting in reduced measurement reliability.

[0004] Thermal micro-mass flowmeters came into being. They work based on the principle of heat exchange between gas and heating elements and have unique advantages in measuring micro-flows. However, there are still many problems that need to be solved in existing thermal micro-mass flowmeters. In terms of improving measurement accuracy, the sensor is easily affected by factors such as ambient temperature and electromagnetic interference, the measurement data has large noise, the calibration method is rough, and the single calibration model cannot adapt to changes in complex working conditions, resulting in limited measurement accuracy; in terms of response speed, the signal processing and data calculation efficiency is poor, and it is difficult to track rapid changes in flow in a timely manner; in terms of data processing and communication functions, there is a lack of efficient data processing algorithms and stable communication mechanisms, which cannot meet the intelligent and remote control needs of the Industrial 4.0 era, and it is difficult to store and transmit massive measurement data and work in collaboration with the host computer.

[0005] As industrial technology develops towards high precision, intelligence, energy conservation and environmental protection, the expectations of various industries for flow meter performance continue to rise. An innovative measurement and control method is urgently needed to break through the bottleneck of existing thermal micro-mass flow meters and achieve high-precision, fast response, strong anti-interference and intelligent control of gas flow measurement and control to meet the complex and changing production needs of modern industry, improve system operation efficiency and safety, reduce energy consumption and costs, and promote industrial automation technology to a new level. Summary of the invention

[0006] To this end, an embodiment of the present invention provides a method for improving the flow measurement and control accuracy of a thermal micro-mass flowmeter, which is used to solve the problems existing in the prior art when measuring micro-mass flow, such as insufficient accuracy, slow response speed, susceptibility to environmental interference, calibration method susceptibility to gross errors, large amount of data processing calculations, limited storage resources and difficulty in adapting to complex fitting results, poor ability to cope with changes in operating conditions, and valve control command lag affecting measurement and control accuracy.

[0007] In order to solve the above problems, an embodiment of the present invention provides a method for improving the flow measurement and control accuracy of a thermal micro-mass flowmeter, which is applied to a cloud server in a measurement and control system. The method includes:

[0008] Acquire gas flow signal data collected by a thermal micro-mass flowmeter, and use a multi-order feedback integer infinite impulse response filter to filter the signal data;

[0009] Perform data calibration and fitting on the filtered signal data, and select the surface set under the required working conditions from the fitting results according to the actual working conditions;

[0010] Based on the surface set, a corresponding flow value is calculated according to the signal data collected in real time, and a valve control instruction is sent to the thermal micro mass flowmeter based on the flow value to achieve control of the gas flow;

[0011] The method also includes establishing an adaptive multi-operating process system model, pre-training the adaptive multi-operating process system model using transfer learning, and updating the surface set using the trained model; wherein the thermal micro-mass flowmeter interacts with a cloud server, the thermal micro-mass flowmeter stores a certain amount of signal data, and regularly uploads the stored signal data to the cloud server; the cloud server performs model fitting according to a preset algorithm based on the uploaded data set, and sends the fitted model to the thermal micro-mass flowmeter to complete the model update.

[0012] Preferably, the process of filtering the signal data using a multi-order feedback integral infinite impulse response filter comprises:

[0013] Passing the original signal In(n) through the multi-order feedback type integer infinite impulse response filter to obtain a filtered signal Out1(n);

[0014] Calculate the difference signal Sub(n) between the original signal and the filtered signal Out1(n): Sub(n)=In(n) - Out1(n);

[0015] The difference signal Sub(n) is further subjected to shaping filtering to obtain a filtered signal Out2(n), and the filtered signal Out2(n) is added to the filtered signal Out1(n) to obtain a filtered signal Out(n): Out(n) = Out2(n) + Out1(n);

[0016] The original signal is subtracted from the filtered signal Out(n) to obtain a difference signal Sub1(n): Sub1(n) = In(n) - Out(n);

[0017] The difference signal Sub1(n) is further subjected to shaping filtering to obtain a filtered signal Out3(n), and the filtered signal Out3(n) is added to the filtered signal Out(n) to obtain a filtered signal Out4(n): Out4(n) = Out3(n) + Out(n);

[0018] Repeat the above steps of difference calculation and addition of filtered signals until a predetermined filtering order is reached.

[0019] Preferably, the process of performing data calibration fitting on the filtered signal data and screening out the surface set under the required working conditions from the fitting results includes:

[0020] Based on the filtered signal data, multiple sample points collected at different temperatures, different pressures, and different flows are selected. Each of the sample points includes the analog-to-digital conversion values of the corresponding temperature, pressure, and flow, forming a multi-dimensional matrix;

[0021] Use the hyper-surface moving variable truncation fitting algorithm to fit the multi-dimensional matrix to obtain a surface fitting function;

[0022] The surface fitting functions are combined to obtain a hyper-surface fitting function, and the surface set under the required working conditions is screened out from the hyper-surface fitting function according to the actual working conditions.

[0023] Among them, the process of using the hyper-surface moving variable truncation fitting algorithm to fit the multi-dimensional matrix to obtain a surface fitting function includes:

[0024] In each support domain, a node combination of k + 1 points is extracted from the sampling data samples of size N, where k + 1 < N, and there are a total of different node combinations;

[0025] Perform least squares fitting on each node combination to obtain the corresponding fitting coefficients , where ;

[0026] Substitute the fitting coefficients of each node combination into the total sample, calculate the residual squares of each point, and sum all the residual squares Arrange in ascending order, that is, ;

[0027] Calculate the difference between adjacent residual squares arranged in ascending order: ,in Indicates the difference;

[0028] Setting a variable cutoff value ,in Indicates rounding. It indicates the parameter value that can make the function reach the maximum value among all the parameters of a function. is the minimum value function;

[0029] Select the first h items in ascending order and sum them to get the target value of local fitting ,in It represents the parameter value that can make the function obtain the minimum value among all the parameters of a function, the target value of the local fitting is used as the fitting regression coefficient of the support domain, and the surface fitting function is reconstructed according to the fitting regression coefficient;

[0030] In each supported domain, As a basis function, Constructing the surface fitting function as a coefficient matrix , where T represents the transpose operation, and weighted least squares is used to solve in the support domain , where the objective function of weighted least squares is:

[0031] ;

[0032] in is the objective function, Indicate point The weight of is the midpoint of the support domain and the support domain follows move, Indicates the upper bound function, that is, greater than or equal to The smallest integer of is the support domain radius, Indicate point The jth basis function evaluated at Indicates at point The observed value at , m represents the number of elements in the linear basis function, n represents the number of sampling points, and the coefficient matrix is ​​obtained by differentiating the objective function.

[0033] Preferably, the process of selecting the surface set under the required working condition from the fitting results according to the actual working condition comprises:

[0034] According to the actual working conditions, the surface set under the required working conditions is selected from the fitting results;

[0035] Project the surface at different temperatures, different pressures and the same flow rate to obtain the correlation coefficient between the temperature and pressure curves;

[0036] The curve segments corresponding to the absolute values ​​of the correlation coefficients less than the set threshold are retained, and the curve segments corresponding to the absolute values ​​of the correlation coefficients greater than or equal to the set threshold are replaced by linear representation.

[0037] Preferably, the method further comprises establishing a valve control command delay compensation model, wherein the valve control command delay compensation model is based on an operating condition matrix of three parameters, namely, pressure, opening degree and flow rate, and obtains a linear expression of the delay time and each parameter by fitting using the least squares method.

[0038] Preferably, the valve control command delay compensation model is expressed as:

[0039] ;

[0040] in, is the predicted delay time, For pressure, is the opening degree, For flow, and Parameters related to the valve control unit.

[0041] An embodiment of the present invention further provides a gas flow measurement and control system based on a thermal micro-mass flowmeter, the system comprising a cloud server, and the cloud server is used to implement the above-mentioned method for improving the flow measurement and control accuracy of the thermal micro-mass flowmeter.

[0042] Preferably, the system also includes a plurality of thermal micro-mass flow meters, which are used to collect gas flow signal data, send the signal data to the cloud server, and receive the valve control command delay compensation model trained by the cloud server.

[0043] Preferably, the thermal micro mass flowmeter comprises:

[0044] A data sampling unit, used for collecting gas flow signal data;

[0045] A valve control unit, used to control the gas flow;

[0046] A communication unit, which is connected to the cloud server for communication, and is used to transmit the gas flow signal data collected by the data sampling unit and the working state of the valve control unit, and receive the surface set required by the flow meter and the valve control command delay compensation model sent by the cloud server;

[0047] A control unit is connected to the data sampling unit, the valve control unit and the communication unit, and is used to send the gas flow signal data collected by the data sampling unit and the working status of the valve control unit to the cloud server via the communication unit, and adaptively issue the valve control command to the proportional valve according to the valve control command delay compensation model trained by the cloud server.

[0048] Preferably, the data sampling unit uses a capillary of set specifications as a sensing element, the outer wall of the capillary is wound to serve as a heater and a winding of a detection component, the winding is divided into two or more groups, and the sensing element collects gas flow signal data by detecting changes in the wall temperature when the gas flows through the capillary.

[0049] It can be seen from the above technical solutions that the present invention has the following beneficial effects:

[0050] (1) The present invention adopts a multi-order feedback integer infinite impulse response filter and combines it with a hypersurface moving variable truncation fitting algorithm to effectively remove noise, reduce the influence of gross errors, and significantly improve measurement accuracy.

[0051] (2) The present invention establishes an adaptive multi-operating condition process system model, uses transfer learning pre-training, and combines clustering and LSTM prediction to update the surface set, which can quickly adapt to complex and changing operating conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] In order to more clearly illustrate the implementation cases of the present invention or the technical solutions in the prior art, the following is a brief description of the drawings required for use in the embodiments. By referring to the drawings, the features and advantages of the present invention will be more clearly understood. The drawings are schematic and should not be understood as limiting the present invention in any way. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. Among them:

[0053] Figure 1 A flow chart of a method for improving the flow measurement and control accuracy of a thermal micro-mass flowmeter provided in an embodiment;

[0054] Figure 2 A flow chart of a surface fitting function obtained by fitting a multidimensional matrix using a hypersurface moving variable truncation fitting algorithm in an embodiment;

[0055] Figure 3 A block diagram of a measurement and control system provided in an embodiment;

[0056] Figure 4 It is a schematic structural diagram of a thermal micro mass flowmeter provided in an embodiment. DETAILED DESCRIPTION

[0057] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0058] Embodiment 1

[0059] In order to solve the problems in the existing technology of measuring small mass flow, such as insufficient accuracy, slow response speed, susceptibility to environmental interference, calibration method susceptible to gross errors, large amount of data processing calculation, limited storage resources and difficulty in adapting to complex fitting results, poor ability to cope with changes in working conditions, and valve control command hysteresis affecting measurement and control accuracy. Figure 1 As shown, an embodiment of the present invention proposes a method for improving the flow measurement and control accuracy of a thermal micro-mass flowmeter, which is applied to a cloud server in a measurement and control system. The method includes:

[0060] The gas flow signal data collected by the thermal micro-mass flowmeter is obtained, and the signal data is filtered using a multi-order feedback integer infinite impulse response filter;

[0061] Perform data calibration and fitting on the filtered signal data, and select the surface set under the required working conditions from the fitting results according to the actual working conditions;

[0062] Based on the surface set, the corresponding flow value is calculated according to the real-time collected signal data, and the valve control instruction is sent to the thermal micro-mass flowmeter based on the flow value to realize the control of the gas flow. It can be seen from the above technical scheme that the present invention proposes a method for improving the flow measurement and control accuracy of the thermal micro-mass flowmeter. First, after obtaining the signal data collected by the thermal micro-mass flowmeter, a multi-order feedback integer infinite impulse response filter is used for filtering. The filter can effectively remove noise, reduce environmental interference, and has a small amount of calculation, which solves the problems of insufficient accuracy, susceptibility to environmental interference and large amount of data processing calculation. Then, the filtered data is calibrated and fitted, and the hypersurface moving variable truncation fitting algorithm is used to screen the surface set of the required working condition, which can avoid the influence of gross errors, adapt to complex fitting results, and solve the problem that the calibration method is susceptible to gross errors and limited storage resources. Finally, the flow value is calculated based on the surface set and the real-time collected data and the valve control instruction is sent. At the same time, an adaptive multi-condition process system model and a valve control command delay compensation model are established, which enhances the ability to cope with working condition changes, reduces the influence of valve control command hysteresis, and realizes high-precision gas flow measurement and control.

[0063] Since the working environment of thermal micro-mass flowmeter is not ideal, there are various interference factors in practice, which cause the signal output by the thermal micro-mass flowmeter to be mixed with random noise. These random noises will make the signal obtained by the thermal micro-mass flowmeter blurred. For example, the originally clear flow signal may not accurately reflect the actual flow situation due to noise interference. This has a serious impact on the reliability and accuracy of the flowmeter, making the measurement results unreliable and inaccurate, which may lead to the wrong direction for subsequent control and analysis based on flow data.

[0064] In order to solve this problem, the present invention uses a multi-order feedback type infinite impulse response (IIR) filter to filter the gas flow signal data collected by the thermal micro-mass flowmeter. It processes the original signal mixed with noise through specific design steps, removes noise interference, makes the signal clear and accurate, and thus improves the reliability and accuracy of the thermal micro-mass flowmeter measurement. Its specific design steps include:

[0065] The original signal In(n) is passed through the multi-order feedback type integer infinite impulse response filter to obtain a filtered signal Out1(n);

[0066] Calculate the difference signal Sub(n) between the original signal and the filtered signal Out1(n): Sub(n) = In(n) - Out1(n);

[0067] The difference signal Sub(n) is further shaped and filtered to obtain a filter signal Out2(n), and the filter signal Out2(n) is added to the filter signal Out1(n) to obtain a filter signal Out(n): Out(n)=Out2(n)+Out1(n);

[0068] Subtract the filtered signal Out(n) from the original signal to obtain the difference signal Sub1(n): Sub1(n)=In(n)-Out(n);

[0069] The difference signal Sub1(n) is further shaped and filtered to obtain a filtered signal Out3(n), and the filtered signal Out3(n) is added to the filtered signal Out(n) to obtain a filtered signal Out4(n): Out4(n)= Out3(n)+Out(n);

[0070] Repeat the above steps of difference calculation and filtered signal addition until a predetermined filtering order is reached.

[0071] Compared with traditional IIR digital filters which have a large amount of computation, the integer IIR filter has coefficients of its transfer function all being integers and does not require floating-point operations, greatly reducing the amount of computation. It meets the condition that the edge-side computing resources of the thermal micro mass flowmeter are limited while also satisfying the requirement for real-time monitoring of gas flow.

[0072] Due to the complex working conditions in industrial sites, a wide variety of gas types, and scarce working condition data, special methods are needed to calibrate and fit the filtered signal data, and select the surface set under the required working conditions from the fitting results according to the actual working conditions.

[0073] Specifically, first, based on the filtered signal data, multiple sample points collected at different temperatures, different pressures, and different flows are selected. Each sample point contains the analog-to-digital conversion values of the corresponding temperature, pressure, and flow, forming a multi-dimensional matrix. Then, the metasurface moving variable trimmed square (MMVTS) fitting algorithm is used to fit the multi-dimensional matrix to obtain a surface fitting function. Finally, the surface fitting functions are combined to obtain a metasurface fitting function, and the surface set under the required working conditions is selected from the metasurface fitting function according to the actual working conditions.

[0074] Furthermore, as Figure 2 shown, the process of using the metasurface moving variable trimmed square (MMVTS) algorithm to fit the multi-dimensional matrix to obtain a surface fitting function includes:

[0075] In each support domain, a node combination of k + 1 points is extracted from the sampling data samples of size N, where k + 1 < N, and there are different node combinations.

[0076] Perform least squares (LS) fitting on each node combination to obtain the corresponding fitting coefficients , where .

[0077] Substitute the fitting coefficients of each node combination into the total sample, calculate the residual square of each point , and arrange all the residual squares in ascending order, that is .

[0078] Calculate the difference between adjacent residual squares after ascending order arrangement: , where represents the difference.

[0079] Set the variable trimming value , where represents rounding, It indicates the parameter value that can make the function reach the maximum value among all the parameters of a function. is the minimum value function. Compared with the traditional intercept constant ,Using variable intercept values ​​can maximize the use of sample points, while eliminating gross errors and improving the utilization rate.

[0080] Select the first The target value of the local fitting is obtained by summing the items ,in It indicates the parameter value that can make the function achieve the minimum value among all the parameters of a function. The target value of the local fitting is used as the fitting regression coefficient of the support domain, and the surface fitting function is reconstructed according to the fitting regression coefficient.

[0081] Furthermore, the analog-to-digital conversion (AD) values ​​collected at different temperatures and flow rates are selected, and in each support domain, As a basis function, Constructing the surface fitting function as a coefficient matrix , where T represents the transpose operation, and weighted least squares is used to solve in the support domain , where the objective function of weighted least squares is:

[0082] ;

[0083] in is the objective function, Indicate point The weight of is the midpoint of the support domain and the support domain follows move, Indicates the upper bound function, that is, greater than or equal to The smallest integer of is the support domain radius, Indicate point The jth basis function evaluated at Indicates at point The observation value at , m represents the number of basis functions, and n represents the number of sample points participating in the local fitting in the support domain. Solve the coefficient matrix This is transformed into the minimum value problem of the objective function J, which can be solved by taking the derivative of J:

[0084] ;

[0085] Solve to get the coefficient matrix:

[0086] ;

[0087] in

[0088] ;

[0089] ;

[0090] ;

[0091] ;

[0092] ;

[0093] in, represents the weighted covariance matrix, represents the weighted eigenvector matrix, Represents the column matrix composed of actual gas flow calibration points, represents the characteristic matrix composed of linear basis, represents the weight diagonal matrix, represents the weight of the nth sampling point, represents the nth element of the mth term in the linear basis, represents the number of elements in the linear basis function, Indicates the number of sampling points, Indicates the collected AD value, Indicates the actual gas flow rate.

[0094] Furthermore, during the fitting process, the weight function The selection and parameter setting of is also crucial. The present invention adopts an exponential weight function, which is used to give the closest calibration point a higher weight. In this way, when fitting, the calibration data closer to the target point has a greater impact on the fitting result, which helps to improve the fitting accuracy. The selection of the value is usually based on experience, while the present invention is based on the range of thermal micro-mass flowmeter required for industrial production. The calculation formula is:

[0095] ;

[0096] in , They are the maximum and minimum values ​​of the range respectively. The value size affects the weight decay speed. The larger it is, the faster the weight decays, and more attention is paid to the most recent signal data; The smaller the value, the opposite is true. When selecting calibration points, the larger the range, the more sparse the calibration points are, so the exponential weight function needs to pay more attention to the closer calibration points when the range is large. This calculation method not only meets the inverse proportionality between the range and the value, but also increases nonlinearity, making the fitting surface approximation effect better and adaptable to a variety of range requirements.

[0097] Furthermore, another surface fitting function can be obtained by selecting AD values ​​collected under different pressures and different flow rates according to the above method. Combining the two surface fitting functions, a hypersurface fitting function can be obtained. The hypersurface fitting function takes into account multiple variable factors and is more suitable for flow measurement under different working conditions. However, hypersurface fitting also brings about the problem of exponential increase in computing and storage capacity, which is not suitable for edge devices such as flow meters with limited computing and storage capacity. Therefore, at the end of the MMLTS algorithm, the surface set under the required working conditions will be screened out from the fitting results according to the possible working conditions faced in reality, and the surface projection will be performed under different temperatures, different pressures, and the same flow rate to obtain the correlation coefficient between the temperature and pressure curves. The curve segments corresponding to the absolute value of the correlation coefficient less than the set threshold (set to 0.5) are retained, and the curve segments corresponding to the absolute value of the correlation coefficient greater than or equal to the set threshold (set to 0.5) are replaced by linear representation, achieving a certain degree of dimensionality reduction and compressing the volume of the surface set.

[0098] In order to improve the accuracy of thermal micro-mass flowmeters and better cope with different working conditions, the present invention establishes an adaptive multi-condition process system model with continuous learning capabilities. In the early stage of model operation, it faces the problem of scarce working condition data, and it is difficult to meet the amount of data required for model training. To solve this problem, the present invention uses a transfer learning method to pre-train the model by using a similar valve control data set. This method is to transfer the knowledge of other related fields (similar to the valve control data set) to the current model training, and use it as the starting model of this model, so that the model has certain basic capabilities in the initial stage. Although the amount of data is insufficient at this time, the model training can be started with the experience of existing relevant data.

[0099] The thermal micro-mass flowmeter of the present invention exchanges data with a cloud server. The thermal micro-mass flowmeter stores a certain amount of signal data and regularly uploads the stored signal data to the cloud server. The cloud server performs model fitting according to the uploaded data set and a preset algorithm, and sends the fitted model to the thermal micro-mass flowmeter to complete the model update.

[0100] Specifically, as the thermal micro-mass flowmeter continues to upload data, a data set for the hysteresis compensation task of the valve control instruction of the thermal micro-mass flowmeter control unit and the task of predicting working condition changes will gradually be formed, and this data set will continue to expand. In order to ensure that the model is always based on the latest and valid data for learning and prediction, the model uses a forgetting factor to discard old data. The reason for this is that as time goes by, some early data may no longer reflect the current working conditions and system status. By discarding old data, the model can continuously update knowledge and adapt to new situations, thereby ensuring the timeliness of the model, enabling it to continuously and accurately analyze and predict the working conditions of the flowmeter.

[0101] The data uploaded by the thermal micro-mass flowmeter includes pressure, temperature, flow rate and valve control unit working status. Considering that the industrial process is relatively fixed, the change law of the working state of the valve control unit can reflect the working mode of the thermal micro-mass flowmeter. Based on this, the thermal micro-mass flowmeters under the same process are clustered according to the change law of the working state of the valve control unit. For example, in chemical production, the working state of the valve control unit of the thermal micro-mass flowmeter in a specific reaction stage is similar, and these thermal micro-mass flowmeters are classified into one category because their working condition changes will also be similar. The clustering results are used as labels to divide the data set, so that different categories of data represent different working conditions. The long short-term memory network (LSTM) training model is used to predict working condition changes. LSTM is suitable for processing time series data, can capture long-term data dependencies, and learn data change trends and characteristics under different working conditions. By predicting working condition changes through LSTM, the surface set required for the thermal micro-mass flowmeter can be updated in time. The surface set is related to the flow measurement calculation, and timely update can ensure accurate flow measurement under different working conditions. This improves the utilization rate of the limited storage space of the thermal micro-mass flowmeter, avoids storing a large amount of useless data, and only retains and updates the surface set data related to the current working conditions. At the same time, it enhances the adaptability of the flowmeter to changing working conditions, so that the thermal micro-mass flowmeter can quickly adjust the measurement calculation basis according to the changes in working conditions, ensure the accuracy of measurement and control, and meet the complex and changeable flow measurement and control needs in industrial production.

[0102] In this embodiment, based on the surface set obtained above, the corresponding flow value is calculated according to the signal data collected in real time, and a valve control instruction is sent to the thermal micro mass flowmeter based on the flow value to realize the control of the gas flow.

[0103] In addition, because there is a delay between the valve control command of the thermal micro-mass flowmeter control unit and the actual action of the proportional valve, this delay has a serious impact on the case of small flow. However, if the proportional valve adjustment speed is too large to reduce the impact of the delay, it will affect the life of the proportional valve. Therefore, the present invention uses clustering to form a data set for the same process. Taking into account the different shutdown delay times under different working conditions, the operating condition matrix is ​​established with the three parameters of pressure, opening degree, and flow as the model input. The time interval between the issuance of the proportional valve control command and the flow reaching the set value is used as a label. Based on the least squares method, the delay time and pressure ( ) is a linear expression , the delay time and the opening degree are 0.8 ( ) is a linear expression , delay time and flow rate ( ) is a linear expression :

[0104] ;

[0105] Through the input of the working condition matrix and the least squares fitting, it can be known that there is a good linear relationship between a single variable and the delay time, but the weight relationship between different variables cannot be known. Therefore, a valve control command delay compensation model is established. The delay time and pressure ( ), 0.8 power of the opening degree ( ),flow( ) show a good linear relationship, so the composite variable is defined for:

[0106] ;

[0107] The delay time is expressed as:

[0108] ;

[0109] in, is the predicted delay time, For pressure, is the opening degree, For flow, and are the parameters related to the valve control unit, is a composite variable, Considering that the valve control unit parameters of each thermal micro-mass flowmeter are different, the model can be independently trained based on the historical data uploaded, so that the valve control command delay compensation model can adapt to the specific working conditions and proportional valve parameters, and then the proportional valve can be issued an adjustment command in advance based on the model prediction to improve the actual control accuracy of the thermal micro-mass flowmeter.

[0110] Embodiment 2

[0111] like Figure 3 As shown, the present invention provides a measurement and control system, which includes a cloud server, which is used to implement the method for improving the flow measurement and control accuracy of a thermal micro-mass flowmeter in the above-mentioned embodiment 1. In addition, the system also includes a plurality of thermal micro-mass flowmeters, which are used to collect gas flow signal data, and send the signal data to the cloud server and receive the valve control command delay compensation model trained by the cloud server.

[0112] Furthermore, if Figure 4 As shown, the thermal micro-mass flowmeter includes a data sampling unit, a valve control unit, a communication unit and a control unit. The data sampling unit is used to collect gas flow signal data; the valve control unit is used to control the gas flow; the communication unit is connected to the cloud server for communication, and is used to transmit the gas flow signal data collected by the data sampling unit and the working state of the valve control unit, and receive the surface set required for the flowmeter and the valve control command delay compensation model sent by the cloud server; the control unit is connected to the data sampling unit, the valve control unit and the communication unit, and is used to send the gas flow signal data collected by the data sampling unit and the working state of the valve control unit to the cloud server via the communication unit, and according to the valve control command delay compensation model trained by the cloud server, the valve control command is adaptively issued to the proportional valve.

[0113] Furthermore, the data sampling unit uses a capillary tube with set specifications (outer diameter 0.5mm, inner diameter 0.4mm) as a sensing element, which improves the sensitivity and accuracy of the measurement. The outer wall of the capillary tube is wound as a heater and a winding of the detection component. The winding is divided into two or more groups. The sensing element collects gas flow signal data by detecting the change in the wall temperature when the gas flows through the capillary tube. The capillary tube of this thermal micro-mass flowmeter is suitable for low-velocity gas flow. Compared with traditional flow meters, it has a wider range ratio, better adaptability and higher accuracy.

[0114] A measurement and control system of the present embodiment is used to implement the aforementioned method for improving the flow measurement and control accuracy of a thermal micro-mass flowmeter. Therefore, the specific implementation method of the measurement and control system can be seen in the embodiment part of the method for improving the flow measurement and control accuracy of a thermal micro-mass flowmeter in the previous text. Therefore, its specific implementation method can refer to the description of the corresponding embodiments of each part. In order to avoid redundancy, it will not be repeated here.

[0115] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0116] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0117] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to produce a computer-implemented process, so that the instructions executed on the computer or other programmable device provide for implementing the process in the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0118] Obviously, the above embodiments are merely examples for clear explanation and are not intended to limit the implementation methods. For those skilled in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the implementation methods here. The obvious changes or modifications derived from them are still within the protection scope of the invention.

Claims

1. A method for improving the flow measurement and control accuracy of a thermal micro-mass flowmeter, characterized in that: Applied to a cloud server in a measurement and control system, the method comprises: Acquire gas flow signal data collected by a thermal micro-mass flowmeter, and use a multi-order feedback integer infinite impulse response filter to filter the signal data; Perform data calibration and fitting on the filtered signal data, and select the surface set under the required working conditions from the fitting results according to the actual working conditions, including: Based on the filtered signal data, multiple sample points collected at different temperatures, pressures, and flow rates are selected, wherein each of the sample points contains analog-to-digital conversion values ​​of corresponding temperature, pressure, and flow rate to form a multidimensional matrix; Fitting the multidimensional matrix using a hypersurface moving variable truncation fitting algorithm to obtain a surface fitting function; The surface fitting functions are combined to obtain a hypersurface fitting function, and a surface set under a required working condition is screened out from the hypersurface fitting function according to the actual working condition; Based on the surface set, a corresponding flow value is calculated according to the signal data collected in real time, and a valve control instruction is sent to the thermal micro mass flowmeter based on the flow value to achieve control of the gas flow; The method also includes establishing an adaptive multi-operating process system model, pre-training the adaptive multi-operating process system model using transfer learning, and updating the surface set using the trained model; wherein the thermal micro-mass flowmeter interacts with a cloud server, the thermal micro-mass flowmeter stores a certain amount of signal data, and regularly uploads the stored signal data to the cloud server; the cloud server performs model fitting according to a preset algorithm based on the uploaded data set, and sends the fitted model to the thermal micro-mass flowmeter to complete the model update.

2. The method for improving the flow measurement and control accuracy of a thermal micro-mass flowmeter according to claim 1, characterized in that: The process of using a multi-order feedback type integral infinite impulse response filter to filter the signal data includes: The original signal In(n) is passed through the multi-order feedback type integer infinite impulse response filter to obtain a filtered signal Out1(n); Calculate the difference signal Sub(n) between the original signal and the filtered signal Out1(n): Sub(n)=In(n)-Out1(n); The difference signal Sub(n) is further shaped and filtered to obtain a filter signal Out2(n), and the filter signal Out2(n) is added to the filter signal Out1(n) to obtain a filter signal Out(n): Out(n)=Out2(n)+Out1(n); Subtract the filtered signal Out(n) from the original signal to obtain the difference signal Sub1(n): Sub1(n)=In(n)-Out(n); The difference signal Sub1(n) is further shaped and filtered to obtain a filter signal Out3(n), and the filter signal Out3(n) is added to the filter signal Out(n) to obtain a filter signal Out4(n): Out4(n)=Out3(n)+Out(n); Repeat the above steps of difference calculation and filtered signal addition until a predetermined filtering order is reached.

3. The method for improving the flow measurement and control accuracy of a thermal micro-mass flowmeter according to claim 1, characterized in that: The process of fitting the multidimensional matrix using the hypersurface moving variable truncation fitting algorithm to obtain the surface fitting function comprises: In each support domain, a node combination of k + 1 points is extracted from a sampling data sample of size N, where k + 1 < N, and there are different node combinations; Perform least squares fitting on each node combination to obtain the corresponding fitting coefficient X i ,in Bring the fitting coefficients of each node combination into the total sample and calculate the residual square of each point And square all the residuals Arrange in ascending order, that is, Calculate the difference between adjacent residual squares arranged in ascending order: Where D i Indicates the difference; Setting a variable cutoff value Among them, int[] represents rounding, argmax() represents the parameter value that can make the function reach the maximum value among all the parameters of a function, and min() is the minimum value function; Select the first h items in ascending order and sum them to get the target value of local fitting Wherein arg min() represents the parameter value that can make the function obtain the minimum value among all the parameters of a function, the target value of the local fitting is used as the fitting regression coefficient of the support domain, and the surface fitting function is reconstructed according to the fitting regression coefficient; In each support region, p(x) = [1, x, y] T As a basis function, a(x) = [1, x, y, x 2 ,xy,y 2 ] as the coefficient matrix to construct the surface fitting function Where T represents the transposition operation, and weighted least squares is used to solve a(x) in the support domain, where the objective function of weighted least squares is: Where J is the objective function, Represents point x i The weight of x is the midpoint of the support region and the support region moves with x. Indicates taking the upper bound function, that is, greater than or equal to xx i The smallest integer, r is the radius of the support domain, p j (x i ) represents point x i The jth basis function evaluated at i Indicates that at point x i The observed value at , m represents the number of elements in the linear basis function, n represents the number of sampling points, and the coefficient matrix is ​​obtained by differentiating the objective function.

4. The method for improving the flow measurement and control accuracy of a thermal micro-mass flowmeter according to claim 1, characterized in that: The process of selecting the surface set under the required working condition from the fitting results according to the actual working condition includes: According to the actual working conditions, the surface set under the required working conditions is selected from the fitting results; Project the surface at different temperatures, different pressures and the same flow rate to obtain the correlation coefficient between the temperature and pressure curves; The curve segments corresponding to the absolute values ​​of the correlation coefficients less than the set threshold are retained, and the curve segments corresponding to the absolute values ​​of the correlation coefficients greater than or equal to the set threshold are replaced by linear representation.

5. The method for improving the flow measurement and control accuracy of a thermal micro-mass flowmeter according to claim 1, characterized in that: The method also includes establishing a valve control command delay compensation model, which is based on an operating condition matrix of three parameters: pressure, opening degree and flow rate, and obtains a linear expression of the delay time and each parameter through least squares fitting.

6. The method for improving the flow measurement and control accuracy of a thermal micro-mass flowmeter according to claim 5, characterized in that: The valve control command delay compensation model is expressed as: t p =m·p·v·k 0.8 +n; Among them, t p is the predicted delay time, p is the pressure, k is the opening degree, v is the flow rate, and m and n are the relevant parameters of the valve control unit.

7. A measurement and control system, characterized in that: The system includes a cloud server, and the cloud server is used to implement the method for improving the flow measurement and control accuracy of a thermal micro-mass flowmeter as described in any one of claims 1 to 6.

8. The measurement and control system according to claim 7, characterized in that: The system also includes a plurality of thermal micro-mass flow meters, which are used to collect gas flow signal data, send the signal data to the cloud server, and receive the valve control command delay compensation model trained by the cloud server.

9. The measurement and control system according to claim 8, characterized in that: The thermal micro mass flowmeter comprises: A data sampling unit, used for collecting gas flow signal data; A valve control unit, used to control the gas flow; A communication unit, which is connected to the cloud server for communication, and is used to transmit the gas flow signal data collected by the data sampling unit and the working state of the valve control unit, and receive the surface set required by the flow meter and the valve control command delay compensation model sent by the cloud server; A control unit is connected to the data sampling unit, the valve control unit and the communication unit, and is used to send the gas flow signal data collected by the data sampling unit and the working status of the valve control unit to the cloud server via the communication unit, and adaptively issue the valve control command to the proportional valve according to the valve control command delay compensation model trained by the cloud server.

10. The measurement and control system according to claim 9, characterized in that: The data sampling unit uses a capillary of set specifications as a sensing element. The outer wall of the capillary is wound with a winding that serves as a heater and a detection component. The winding is divided into two or more groups. The sensing element collects gas flow signal data by detecting changes in the wall temperature when the gas flows through the capillary.

Images