CN109709926B - Method for adjusting melt quality in fiber production process - Google Patents

Method for adjusting melt quality in fiber production process Download PDF

Info

Publication number
CN109709926B
CN109709926B CN201910071615.6A CN201910071615A CN109709926B CN 109709926 B CN109709926 B CN 109709926B CN 201910071615 A CN201910071615 A CN 201910071615A CN 109709926 B CN109709926 B CN 109709926B
Authority
CN
China
Prior art keywords
model
melt
section
data
prediction
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN201910071615.6A
Other languages
Chinese (zh)
Other versions
CN109709926A (en
Inventor
郝矿荣
龚龙浩
王华平
蔡欣
陈磊
任立红
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Donghua University
Original Assignee
Donghua University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Donghua University filed Critical Donghua University
Priority to CN201910071615.6A priority Critical patent/CN109709926B/en
Publication of CN109709926A publication Critical patent/CN109709926A/en
Application granted granted Critical
Publication of CN109709926B publication Critical patent/CN109709926B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
    • Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
    • Y02P90/02—Total factory control, e.g. smart factories, flexible manufacturing systems [FMS] or integrated manufacturing systems [IMS]

Landscapes

  • Spinning Methods And Devices For Manufacturing Artificial Fibers (AREA)

Abstract

The invention relates to a method for adjusting melt quality in the fiber production process, dividing the fiber production melt conveying process into a plurality of sections according to elements through which the melt flows, establishing a prediction model section by section, inputting the process parameter data of the starting point of each section into the prediction model, predicting to obtain the process parameter data of the end point of the section, and adjusting the melt quality according to the process parameter data of the end point and the starting point of each section; in two adjacent sections, the output of the prediction model of the front section, namely the section where the melt arrives first, is the input of the prediction model of the rear section, namely the section where the melt arrives later; when a certain section is provided with a process parameter data acquisition device, the prediction model is a mixed model; otherwise, the prediction model is a mechanism model; the hybrid model is composed of a mechanism model and a data driving model, the data driving model is composed of a GRU (generalized regression Unit) and an X model, and the X model is an SVR (support vector regression), BP (Back propagation) neural network or ELM (element-free matrix). The adjusting method has high prediction accuracy, and can adjust the melt quality in real time and ensure the stability of the melt quality.

Description

Method for adjusting melt quality in fiber production process
Technical Field
The invention belongs to the technical field of polyester fiber production control, relates to a method for adjusting melt quality in a fiber production process, and particularly relates to a method for dynamically and intelligently predicting melt quality in the polyester fiber production process and adjusting the melt quality based on mixed model segmentation integration.
Background
Polyester fiber is commonly called Polyester (PET) fiber, which is synthetic fiber obtained by spinning polyester formed by polycondensation of organic dibasic acid and dihydric alcohol. Because the polyester fiber has the characteristics of higher strength, good elastic recovery capability, good wrinkle resistance and good shape retention, the polyester fiber is widely applied to the fields of textile, industry and the like at present.
With the development of society, people have higher and higher requirements on polyester fibers. The improvement of the performance of the polyester fiber is an urgent need of the industry. The main direction of research is to improve the performance of polyester fibers, and along with the continuous and deep research in recent years, the difficulty in improving the performance of polyester fibers from the aspect of improving raw materials is increasing, so that the attention of the industry is paid to improving the performance of polyester fibers from other aspects.
The melt conveying is a process of fiber production, and each parameter of the melt conveying process can influence the quality of the polyester melt, and the quality of the polyester melt can greatly influence the quality of the final finished yarn, so that the research on the relation between each parameter of the melt conveying process and the quality, namely the performance parameter, of the polyester melt so as to improve the quality of the polyester melt is of practical significance. At present, systematic research is not carried out in the industry according to the relation between each parameter of the melt conveying process and the quality of the polyester melt. Generally, a sensor is used for collecting partial performance parameters of a melt in real time so as to monitor a process, and although the sensor can warn abnormal conditions, the sensor cannot prevent the abnormal conditions or change the abnormal conditions. Some researchers segment the conveying process of the polyester melt and establish a mechanism model to study the relation between each parameter of the melt conveying process and the performance parameter of the polyester melt, for example, chemical fiber companies segment the large-capacity fiber production process and establish the mechanism model, Jiangsu textile industry design research institute takes a 24-head/position melt direct spinning polyester filament device as an example, introduces the technical key points and design methods of the melt conveying system design in the direct spinning device, and middle petroleum engineering construction companies discuss the influence factors of long-distance polyester melt conveying and calculate the performance indexes of the melt in each segment according to the mechanism model, but because the mechanism models are adopted, the accuracy of the mechanism model depends on the definition of the flow mechanism, and part of the mechanism of the fiber production process is unknown at present, the accuracy of the mechanism model is poor, it is difficult to guide the actual production.
Therefore, the development of a method capable of accurately predicting the melt quality in the polyester fiber production process and adjusting the melt quality is of great practical significance.
Disclosure of Invention
The invention aims to provide a method capable of accurately predicting the melt quality in the polyester fiber production process and adjusting the melt quality, aiming at the defect that the melt quality in the polyester fiber production process is difficult to accurately predict in the prior art. The invention divides the whole melt conveying process into sub-sections according to the distribution of components, selects different models for prediction aiming at different sub-sections, wherein the prediction model of the pipeline section is a mechanism model, the prediction models of other sections are mixed models, and the sections are in seamless connection, thereby facilitating troubleshooting and optimizing the melt quality.
In order to achieve the purpose, the invention adopts the following technical scheme:
the method for adjusting the melt quality in the fiber production process comprises the steps of dividing the conveying process of the fiber production melt into a plurality of sections according to elements through which the melt flows, establishing a prediction model section by section, inputting the process parameter data of the starting point of each section, namely the point where the melt arrives first, into the prediction model, predicting to obtain the process parameter data of the end point of the section, namely the point where the melt arrives last, and adjusting the melt quality according to the process parameter data of the end point and the starting point of each section;
in the fiber production process, if the quality of the melt in the melt conveying link has problems, the subsequent spinning link is influenced, the quality of the melt is predicted in advance according to the process parameter data of the end point and the starting point of each section, once the quality of the melt is predicted to have problems, the adjustment can be carried out in advance to prevent the problems from continuously worsening, the adjustment comprises the steps of adjusting the opening degree of each valve and adding some raw materials on line to ensure that the quality of the melt is carried out towards an ideal state, and therefore the subsequent production is normally carried out, and the specific adjustment method comprises the following steps: when the percentage of the variation of the end point and start point process parameter data to the variation of the ideal process parameter data (which is a set value, for example, the variation of the ideal temperature data is 5 ℃ and the variation of the ideal pressure data is 5MPa, which is only an example and is not limited thereto, and can be adjusted according to actual requirements) is greater than 15% (depending on specific production varieties) or the percentage of the end point process parameter data to the variation of the ideal process parameter data (which is a set value and can be adjusted according to actual requirements) is lower than the ideal value, the melt quality is considered to be poor, if the temperature variation exceeds the standard, the related setting of the heat medium pump is adjusted in a reciprocating manner, if the temperature is higher, the valve opening of the heat medium pump is reduced, namely the valve opening of 1% is reduced, if the temperature variation is not recovered to be normal, the valve opening of 1% is reduced again until the temperature; if the pressure change exceeds the standard, the related setting of the booster pump part is adjusted in a reciprocating mode, if the pressure is higher than the standard, the valve opening degree of the booster pump is reduced, namely the valve opening degree is reduced by 1%, if the pressure change does not return to the normal state, the valve opening degree is reduced by 1% again until the pressure change returns to the normal state, otherwise, the valve opening degree of the booster pump is adjusted, and the amplitude of the valve opening degree of the regulating valve every time is not limited to the value, and can be determined according to specific conditions;
in two adjacent sections, the output of the prediction model of the front section, namely the section where the melt arrives first, is the input of the prediction model of the rear section, namely the section where the melt arrives later;
when a certain section is provided with a process parameter data acquisition device, the prediction model is a mixed model; otherwise, the prediction model is a mechanism model;
the mixed model is composed of the mechanism model and the data driving model, the input of the mixed model is the input of the mechanism model, the input of the mixed model and the residual form the input of the data driving model together, the residual is the difference obtained by subtracting the output of the mechanism model from the actual value of the process parameter, and the output of the mixed model is the sum of the output of the mechanism model and the output of the data driving model; the mechanism model is a formula which is established according to the internal construction mechanism of the object and reflects the change of various properties of the object, and the data driving model is a 'black box' model which is brought by a large amount of actual data for learning and training, so that the 'black box' model learns various relationships hidden in the large amount of data until the corresponding relationship between the data can be perfectly fitted;
the training process of the data-driven model is a process of continuously adjusting the parameters of the data-driven model by taking SR and SC as input and theoretical output respectively, wherein SR is SR1And SR2SC is the actual value of the collected end point process parameter data of each section in the historical fiber production process, SR1For each section collected in the historical fiber production processActual value of point process parameter data, SR2Is obtained by subjecting SR1After the input of the input signal into the mechanism model, the output of the mechanism model is subtracted by SC, and the training termination condition is that the maximum iteration times are reached;
the specific training process of the data-driven model is as follows:
(1) selecting 30000 groups of input data I and output data O from actual production data (the number of the data can be selected according to actual conditions, and the data is taken as an example here), substituting the input data I into a mechanism model to obtain a calculation result O1, comparing O1 with O to obtain a residual R, and respectively substituting the input data I and the residual R into two data driving models (GRU and X models) for training;
(2) selecting the weight and the bias of a network structure by using a normal random method, and initializing the network;
(3) substituting the input (input data I) and the label (residual error R) of a batch (the size is 500, the batch size can be selected according to the actual situation, and only taking the example as the example here) into the network in each iteration, calculating the error of the output of the network and the label R, solving the partial derivatives of the error to the weight and the offset, and then updating the weight and the offset;
(4) repeating the step (3) until the maximum iteration number (5000 times) is reached, and taking the weight and the bias at the moment as final parameters of the network;
the data driving model is composed of a GRU (gated cyclic unit) and an X model, wherein the X model is SVR (support vector regression), BP neural network or ELM, the input of the data driving model is the input of the GRU and the X model, the output of the data driving model is the combination of the outputs of the GRU and the X model, and the combination formula is as follows:
S＝W1·S1+W2·S2；
where S is the output of the data-driven model, S1And S2Outputs of GRU and X models, W, respectively1And W2Weights, W, for GRU and X models, respectively1And W2The calculation formula of (a) is as follows:
Figure BDA0001957463470000041
Figure BDA0001957463470000042
W2＝1-W1；
wherein i is 1 or 2, and when i is 1, DiWhen i is 2, D is the relative error between the predicted value and the measured value of GRUiThe relative error between the predicted value and the measured value of the X model is shown.
Taking the data-driven model of GRU + SVR as an example, the specific functions of GRU and SVR are as follows: the GRU predicts a time sequence by combining the reservation quantity of past time data with current data through a gating mechanism; after the SVR transforms the input space to a high dimensional space by a non-linear transformation defined by an inner product function, the optimal hyperplane is solved in this space.
As a preferred technical scheme:
the method for adjusting the melt quality in the fiber production process as described above, wherein the fiber production melt conveying process is the whole process from the outflow of the polyester melt from the final polymerization kettle to the extrusion from the spinneret plate of the spinning assembly.
In the method for adjusting the melt quality in the fiber production process, the multiple sections are 12 sections, namely a pipeline I section, a booster pump section, a pipeline II section, a heat exchanger section, a pipeline III section, a static mixer section, a pipeline IV section, a three-way valve section, a pipeline V section, a metering pump section, a pipeline VI section and a spinning assembly section in sequence. The invention takes each section of pipeline or each component as a section to be segmented according to the industrial layout, compared with the segmentation of melt conveying in the prior art, the invention divides the melt conveying process of fiber production more carefully, so that the whole melt conveying link is clearer and the melt indexes of all places can be predicted. The specific segmentation mode of the present invention is not limited to this, and those skilled in the art can segment the fiber production melt conveying process into several segments according to actual needs.
According to the method for adjusting the melt quality in the fiber production process, the prediction models of all pipeline sections are mechanism models, and the prediction models of other sections are mixed models. The prediction model can be set according to the actual application scene, but the hybrid model needs the data parameters of the segment to drive. In an actual application scene, a pipeline section is generally not provided with a sensor, namely data cannot be collected, so that a mechanism model is used as a prediction model, and a data driving model can be established for the section provided with the sensor and the data which can be collected, namely a mixed model is used as the prediction model.
In the method for adjusting the melt quality in the fiber production process, the mechanism models of all the pipeline sections and the three-way valve section are the same, and the expression is as follows:
Figure BDA0001957463470000051
Figure BDA0001957463470000052
Figure BDA0001957463470000053
Figure BDA0001957463470000054
Figure BDA0001957463470000055
wherein V is the flow rate in m/min, Δ P is the pressure drop in MPa, T is the residence time in min, T is the melt temperature in deg.C, LyThe length of the pipeline is m, and the subscript value corresponds to the serial number of the pipeline; dyIs the inner diameter of the pipeline in mm, the subscript value corresponding to the serial number of the pipeline, G is the melt flow rate in t/d, rho is the melt density in kg/m3Mu is melt flow viscosityDegree in Pa · s, IV in the unit of melt intrinsic viscosity in dl/g; delta T is the temperature difference in degrees CpIs the specific heat capacity of the melt, the unit is kJ/kg DEG C, 192 is the molecular weight of a PET chain link, k is a degradation rate constant, IV0Initial PET intrinsic viscosity in dl/g;
the heat exchanger section and the static mixer section have the same mechanism model, and the expression is as follows:
Figure BDA0001957463470000056
Figure BDA0001957463470000057
wherein Δ T is a temperature difference in units of ℃μ1Is the viscosity of the fluid at the average temperature and has a unit of Pa.s, mumIs the viscosity of the tube wall at the average temperature and has a unit of Pa.s, lambda2The coefficient of thermal conductivity of the fluid is W/(m.K), G is the flow rate of the fluid and is kg/h, CpThe specific heat capacity of the fluid is expressed in J/(kg. K), A is the heat transfer area and is expressed in m2，ΔtaveIs the logarithmic mean temperature difference, in units of K, D is the inner diameter of a hollow circular tube of the heat exchanger or static mixer, in units of m, and L is the length of the tube of the heat exchanger or static mixer, in units of m;
the booster pump section and the metering pump section have the same mechanism model, and the expression is as follows:
Figure BDA0001957463470000061
ΔT＝-3.973+0.296Rev；
ΔP＝0.473Rev；
Figure BDA0001957463470000062
wherein, VBPIs the volume of the booster pump, and has the unit of cc/R, and the volume of the booster pump in a semi-open configuration is 1500cc/RevIs the rotation speed of the booster pump, and the unit is rpm and VvzIs the inner volume of the booster pump piping, and has a unit of m3G is the melt flow rate in t/d, rho is the melt density in kg/m3，VgIs the volume in m in the booster pump piping3；
The expression of the mechanism model of the spinning component section is as follows:
Figure BDA0001957463470000063
Figure BDA0001957463470000064
wherein q is the melt flow rate of a single hole of the distribution plate and is m3/s，l*The length is corrected for the orifice entrance in m.
In the method for adjusting the melt quality in the fiber production process, the process parameters are temperature, pressure and intrinsic viscosity, and the units are respectively temperature, MPa and Pa & s.
In the method for adjusting the melt quality in the fiber production process, the process parameter data acquisition device is a sensor. The specific form of the process parameter data acquisition device is not limited thereto, and the invention is only taken as an example.
The method for adjusting the melt quality in the fiber production process as described above, wherein the prediction model is connected with the processor, and the processor is used for calculating the difference between the input and the output of the prediction model. The processor can be equipment with calculation capability such as a CPU, and the performance parameters of the final melt are obtained through calculation of the processor, so that the quality of the melt is judged.
According to the method for adjusting the melt quality in the fiber production process, the prediction model and the processor are connected with the visual display module, and the visual display module is used for displaying the data predicted by the prediction model and the data processed by the processor in real time. The visual display module can be a fixed display, or a mobile display device such as a mobile phone, as long as the data predicted by the prediction model and the data processed by the processor can be displayed.
The invention mechanism is as follows:
the invention adopts different prediction models aiming at different sections, wherein the prediction model of the pipeline section is a mechanism model, the prediction models of other sections are all mixed models, because the pipeline section is generally not provided with a sensor, and the other sections are generally provided with sensors for collecting data, the invention takes the mixed models as the prediction models aiming at other sections, the mechanism model is corrected by a data driving model through the collected data, the precision of the model is obviously improved, the pipeline section takes the mechanism model as the prediction model, the sensor is not required to be arranged, and the prediction precision can be ensured, the selected mixed model is composed of the mechanism model and the data driving model, the invention carries out supplementary correction on the mechanism model through the data driving model, the precision of the model can be obviously improved, wherein the data driving model is composed of GRU and an X model (SVR, BP neural network or ELM), the SVR, BP neural network or ELM selected by the invention has stronger universality, and simultaneously, because industrial data has certain time sequence, the GRU selected by the invention overcomes the defect that the existing data driving model is difficult to solve the data time sequence.
The GRU model is mainly applied to the fields related to the time sequence of data, such as financial stocks and text analysis, at present, while the X model has been developed for many years and has been applied to a plurality of application fields, such as classification, clustering, pattern recognition, function approximation, data compression and the like, but the GRU and the X model are not applied to the fields related to melt conveying. The GRU model predicts the output of the next moment according to historical information (data of the previous moment), the model only reflects the time relation of information and is irrelevant to the data change principle, in a sense, the GRU model in the prior art is trained on disordered random data and reflects the front-back relation of the disordered random data, the change of temperature and pressure data in the melt conveying process is not the same as stock data, and the specific change of the temperature and the pressure is closely related to other process parameters in the melt conveying process. The result error of the whole model caused by the error of the result of one algorithm is avoided.
Has the advantages that:
(1) according to the method for adjusting the melt quality in the fiber production process, different prediction models are adopted for different sections, wherein the prediction model of the pipeline section is a mechanism model, and the prediction models of other sections are mixed models, so that the prediction accuracy is obviously improved;
(2) according to the method for adjusting the melt quality in the fiber production process, the mechanism model and the data driving model are organically combined and applied to the conveying process of the melt in the fiber production process, so that the prediction accuracy is improved, and the fault tolerance rate is increased;
(3) according to the adjusting method for the melt quality in the fiber production process, the whole fiber production melt conveying process is reasonably segmented and segmented, so that the melt performance index of each segment can be predicted, and the melt performance index of the whole fiber production melt conveying process can be predicted;
(4) a method for adjusting the quality of a melt in the fiber production process can predict the performance of a product in advance, prevent the product performance from deteriorating in the production process in time, and introduce an automatic adjusting module to adjust the quality of the melt in real time so as to ensure the stability of the quality of the melt.
Drawings
FIG. 1 is a flow chart of melt quality prediction in a fiber production process of the present invention;
FIG. 2 is a schematic diagram of the structure of the hybrid model of the present invention;
FIG. 3 is a schematic diagram of an integration method of the hybrid model of the present invention;
FIG. 4 is a schematic illustration of melt delivery in a fiber production process of the present invention;
FIG. 5 is a schematic interface diagram of a visualization display module of the present invention;
FIG. 6 is a comparison of the predicted results of the prediction method, the mechanism model prediction method, the GRU prediction method, and the SVR prediction method of the present invention;
wherein, 1-a final polymerization kettle, 2-a junction, 3-a pipeline I, 4-a booster pump inlet, 5-a booster pump, 6-a booster pump outlet, 7-a pipeline II, 8-a heat exchanger inlet, 9-a heat exchanger, 10-a heat exchanger outlet, 11-a pipeline III, 12-a static mixer, 13-a pipeline IV, 14-a three-way valve, 15-a pipeline V, 16-a metering pump inlet, 17-a metering pump, 18-a metering pump outlet, 19-a pipeline VI, 20-a spinning component inlet, 21-a spinning component and 22-a spinning component outlet.
Detailed Description
The invention will be further illustrated with reference to specific embodiments. It should be understood that these examples are for illustrative purposes only and are not intended to limit the scope of the present invention. Further, it should be understood that various changes or modifications of the present invention may be made by those skilled in the art after reading the teaching of the present invention, and such equivalents may fall within the scope of the present invention as defined in the appended claims.
A method for adjusting the melt quality in the fiber production process comprises the following specific steps:
(1) the whole process of conveying the fiber production melt, namely the whole process from the outflow of the polyester melt from the final polymerization kettle 1 to the extrusion from the spinneret plate of the spinning assembly, as shown in fig. 4 is divided into a pipeline I3 section, a booster pump 5 section, a pipeline II 7 section, a heat exchanger 9 section, a pipeline III 11 section, a static mixer 12 section, a pipeline IV 13 section, a three-way valve 14 section, a pipeline V15 section, a metering pump 17 section, a pipeline VI 19 section and a spinning assembly 21 section from an interface 2 in sequence according to elements through which the melt flows, wherein sensors are arranged on a booster pump inlet 4, a booster pump outlet 6, a heat exchanger inlet 8, a heat exchanger outlet 10, a static mixer 12, a three-way valve 14, a metering pump inlet 16, a metering pump outlet 18, a spinning assembly inlet 20 and a spinning;
(2) establishing prediction models section by section, wherein the prediction models of all pipeline sections are mechanism models, and the prediction models of other sections are mixed models;
the structure of the hybrid model is shown in fig. 2, the integration method of the hybrid model is shown in fig. 3, the hybrid model is composed of a mechanism model and a data driving model, the input of the hybrid model is the input of the mechanism model, the input of the hybrid model and a residual error jointly form the input of the data driving model, the residual error is a difference value obtained by subtracting the output of the mechanism model from the actual value of the process parameter, and the output of the hybrid model is the sum of the output of the mechanism model and the output of the data driving model;
the data driving model is composed of a GRU model and an X model, the X model is an SVR (support vector regression) or BP (back propagation) neural network or ELM (element-free model), the input of the data driving model is the input of the GRU model and the X model, the output of the data driving model is the combination of the outputs of the GRU model and the X model, and the combination formula is as follows:
S＝W1·S1+W2·S2；
where S is the output of the data-driven model, S1And S2Outputs of GRU and X models, W, respectively1And W2Weights, W, for GRU and X models, respectively1And W2The calculation formula of (a) is as follows:
Figure BDA0001957463470000091
Figure BDA0001957463470000092
W2＝1-W1；
wherein i is 1 or 2, and when i is 1, DiWhen i is 2, D is the relative error between the predicted value and the measured value of GRUiRelative error between the predicted value and the measured value of the X model is obtained;
the training process of the data-driven model is as follows: are respectively provided withSR and SC are input and theoretical output, and SR is SR and the process of continuously adjusting the parameters of the data-driven model1And SR2SC is the actual value of the collected end point process parameter data of each section in the historical fiber production process, SR1For actual values, SR, of the process parameter data of the starting points of the segments collected during the production of the historical fibers2Is obtained by subjecting SR1After the input of the input signal into the mechanism model, the output of the mechanism model is subtracted by SC, and the training termination condition is that the maximum iteration times are reached;
the mechanism model of each segment is as follows:
wherein all pipeline sections are the same as the mechanism model of the three-way valve section, and the expression is as follows:
Figure BDA0001957463470000101
Figure BDA0001957463470000102
Figure BDA0001957463470000103
Figure BDA0001957463470000104
Figure BDA0001957463470000105
wherein V is the flow rate in m/min, Δ P is the pressure drop in MPa, T is the residence time in min, T is the melt temperature in deg.C, LyThe length of the pipeline is m, and the subscript value corresponds to the serial number of the pipeline; dyIs the inner diameter of the pipeline in mm, the subscript value corresponding to the serial number of the pipeline, G is the melt flow rate in t/d, rho is the melt density in kg/m3Mu is the melt flow viscosity,the unit is Pa.s, IV is the inherent viscosity of the melt, and the unit is dl/g; delta T is the temperature difference in degrees CpIs the specific heat capacity of the melt, the unit is kJ/kg DEG C, k is a degradation rate constant,
IV0initial PET intrinsic viscosity in dl/g;
the heat exchanger section and the static mixer section have the same mechanism model, and the expression is as follows:
Figure BDA0001957463470000111
Figure BDA0001957463470000112
wherein Δ T is a temperature difference in units of ℃μ1Is the viscosity of the melt at the average temperature and has the unit of Pa.s, mumIs the viscosity of the tube wall at the average temperature and has a unit of Pa.s, lambda2Is the melt thermal conductivity in W/(m.K), G is the melt flow in kg/h, CpIs the specific heat capacity of the melt, and has the unit of J/(kg. K), A is the heat transfer area, and has the unit of m2，ΔtaveIs the logarithmic mean temperature difference, in units of K, D is the inner diameter of a hollow circular tube of the heat exchanger or static mixer, in units of m, and L is the length of the tube of the heat exchanger or static mixer, in units of m;
the booster pump section and the metering pump section have the same mechanism model, and the expression is as follows:
Figure BDA0001957463470000113
ΔT＝-3.973+0.296Rev；
ΔP＝0.473Rev；
Figure BDA0001957463470000114
wherein, VBPFor increasing pressurePump volume in cc/R, RevIs the rotation speed of the booster pump, and the unit is rpm,
Vvzis the inner volume of the booster pump piping, and has a unit of m3G is the melt flow rate in t/d, rho is the melt density in kg/m3，VgIs the volume in m in the booster pump piping3；
The mechanical model expression of the spinning component section is as follows:
Figure BDA0001957463470000115
Figure BDA0001957463470000116
wherein q is the melt flow rate of a single hole of the distribution plate and is m3/s，l*Correcting the length for the hole entrance in m;
(3) inputting the process parameter data, namely temperature, pressure and intrinsic viscosity, of the starting point of each section, namely the point where the melt arrives first, into a prediction model, predicting to obtain the process parameter data of the end point of the section, namely the point where the melt arrives last, of the two adjacent sections, wherein the output of the prediction model of the section where the melt arrives first, namely the front section, is the input of the prediction model of the section where the melt arrives later, namely the rear section, and the prediction flow chart of each section is shown in figure 1, and the units of the temperature, the pressure and the intrinsic viscosity are respectively, namely, MPa and Pa & s;
(4) a processor connected with the prediction model calculates to obtain a difference value between input and output of the prediction model, and a visualization display module displays data predicted by the prediction model and a display interface of the data visualization display module processed by the processor in real time as shown in fig. 5;
(5) the processor adjusts the melt quality according to the process parameter data of the end point and the starting point of each section, namely, the quality of the melt is predicted in advance according to the process parameter data of the end point and the starting point of each section, once the quality of the melt is predicted to have a problem, the adjustment can be carried out in advance, the problem is prevented from being continuously worsened, the adjustment comprises the steps of adjusting the opening degree of each valve and adding some raw materials on line to ensure that the quality of the melt is carried out towards an ideal state, and therefore the subsequent production is normally carried out, and the specific adjustment method comprises the following steps: when the variation of the end point and start point process parameter data accounts for more than 15 percent of the variation of the ideal process parameter data (according to specific production varieties) or the end point process parameter data is lower than an ideal value, the melt quality is considered to be poor, if the temperature variation exceeds the standard, the related setting of the heat medium pump is adjusted, if the temperature is higher than the standard, the temperature of the heat medium is reduced, so the temperature of the melt is reduced, otherwise, the temperature is increased, if the pressure variation exceeds the standard, the related setting of the booster pump part is adjusted, if the pressure is higher than the standard, the valve opening is adjusted to be smaller, otherwise, the valve opening is adjusted to be larger (the specific valve opening is determined according to specific conditions).
Example 1
A method for adjusting the quality of melt in the production process of fiber, wherein an X model adopts SVR.
Taking 400 groups of real input and output temperature data in the melt conveying process as shown in the figure 4, and respectively applying the prediction method, the mechanism model prediction method, the GRU prediction method and the SVR prediction method of the embodiment to predict the data, wherein the prediction result diagram is shown in the figure 6, True is a True value, HYB is a prediction result curve of the method, GRU is a prediction result curve of a GRU algorithm, SVR is a prediction result curve of an SVR algorithm, and ORG is a prediction result curve of a mechanism model;
the MSE of each prediction method is as follows in sequence: the MSE of the process of the invention is 5.84X 10-4The MSE of GRU is 6.25 × 10-4MSE of SVR is 1.64X 10-3The MSE of the mechanism model was 1.26X 10-3. Comparing and observing fig. 6, it can be found that the accuracy of the prediction method of the present invention is much higher than that of other prediction methods because the present invention adopts complementation between models to improve the prediction accuracy, adds a data-driven model to compensate the unknown part of the mechanism model based on the mechanism model, and selects GRU and SVR satisfying the time sequence and nonlinearity of the industrial data as the data-driven model, thereby overcoming the defect of the time sequence of the industrial data and remarkably improving the prediction accuracy.
The verification proves that the method for adjusting the melt quality in the fiber production process can accurately predict the melt parameter of the conveying end point according to the initial melt parameter, and can adjust the melt quality in real time according to the predicted final melt parameter so as to ensure the stability of the melt quality of the conveying end point.

Claims (9)

1. The method for adjusting the melt quality in the fiber production process is characterized by comprising the following steps: dividing the conveying process of the fiber production melt into a plurality of sections according to elements through which the melt flows, establishing a prediction model section by section, inputting the process parameter data of the starting point of each section, namely the point where the melt arrives first, into the prediction model, predicting to obtain the process parameter data of the end point of the section, namely the point where the melt arrives last, and adjusting the melt quality according to the process parameter data of the end point and the starting point of each section;
in two adjacent sections, the output of the prediction model of the front section, namely the section where the melt arrives first, is the input of the prediction model of the rear section, namely the section where the melt arrives later;
when a certain section is provided with a process parameter data acquisition device, the prediction model is a mixed model; otherwise, the prediction model is a mechanism model;
the mixed model is composed of the mechanism model and the data driving model, the input of the mixed model is the input of the mechanism model, the input of the mixed model and the residual form the input of the data driving model together, the residual is the difference obtained by subtracting the output of the mechanism model from the actual value of the process parameter, and the output of the mixed model is the sum of the output of the mechanism model and the output of the data driving model;
the training process of the data-driven model is a process of continuously adjusting the parameters of the data-driven model by taking SR and SC as input and theoretical output respectively, wherein SR is SR1And SR2SC is the actual value of the collected end point process parameter data of each section in the historical fiber production process, SR1For actual values, SR, of the process parameter data of the starting points of the segments collected during the production of the historical fibers2Is obtained by subjecting SR1After the input of the input signal into the mechanism model, the output of the mechanism model is subtracted by SC, and the training termination condition is that the maximum iteration times are reached;
the data driving model is composed of a GRU model and an X model, the X model is an SVR (support vector regression) or BP (back propagation) neural network or ELM (element-free model), the input of the data driving model is the input of the GRU model and the X model, the output of the data driving model is the combination of the outputs of the GRU model and the X model, and the combination formula is as follows:
S＝W1·S1+W2·S2；
where S is the output of the data-driven model, S1And S2Outputs of GRU and X models, W, respectively1And W2Weights, W, for GRU and X models, respectively1And W2The calculation formula of (a) is as follows:
Figure FDA0002641481120000011
Figure FDA0002641481120000012
W2＝1-W1；
wherein i is 1 or 2, and when i is 1, DiWhen i is 2, D is the relative error between the predicted value and the measured value of GRUiThe relative error between the predicted value and the measured value of the X model is shown.
2. The method of claim 1, wherein the conveying of the melt for fiber production is the entire process from the point of the polyester melt flowing out of the final polymerizer to the point of the polyester melt extruding out of the spinneret of the spinning pack.
3. The method of claim 2, wherein the plurality of stages comprises 12 stages, which are a pipe I stage, a booster pump stage, a pipe II stage, a heat exchanger stage, a pipe III stage, a static mixer stage, a pipe IV stage, a three-way valve stage, a pipe V stage, a metering pump stage, a pipe VI stage, and a spinning pack stage.
4. The method of claim 3, wherein the predictive models for all the sections of the pipe are mechanistic models and the predictive models for the other sections are hybrid models.
5. The method of claim 4, wherein the mechanical model of all the pipe sections and the three-way valve section is the same, and the expression is as follows:
Figure FDA0002641481120000021
Figure FDA0002641481120000022
Figure FDA0002641481120000023
Figure FDA0002641481120000024
Figure FDA0002641481120000025
wherein V is the flow rate in m/min, Δ P is the pressure drop in MPa, T is the residence time in min, T is the melt temperature in deg.C, LyThe length of the pipeline is m, and the subscript value corresponds to the serial number of the pipeline; dyIs the inner diameter of the pipeline in mm, the subscript value corresponding to the serial number of the pipeline, G is the melt flow rate in t/d, rho is the melt density in kg/m3Mu is melt flow viscosity with Pa.s, IV is melt intrinsic viscosity with dl/g; delta T is the temperature difference in degrees CpThe specific heat capacity of the melt is expressed in kJ/(kg DEG C)) K is the degradation rate constant, IV0Initial PET intrinsic viscosity in dl/g;
the heat exchanger section and the static mixer section have the same mechanism model, and the expression is as follows:
Figure FDA0002641481120000031
Figure FDA0002641481120000032
wherein Δ T is a temperature difference in units of ℃μ1Is the viscosity of the melt at the average temperature and has the unit of Pa.s, mumIs the viscosity of the tube wall at the average temperature and has a unit of Pa.s, lambda2Is the melt thermal conductivity in W/(m.K), G is the melt flow in kg/h, CpIs the specific heat capacity of the melt, and has the unit of J/(kg. K), A is the heat transfer area, and has the unit of m2，ΔtaveIs the logarithmic mean temperature difference, in units of K, D is the inner diameter of a hollow circular tube of the heat exchanger or static mixer, in units of m, and L is the length of the tube of the heat exchanger or static mixer, in units of m;
the booster pump section and the metering pump section have the same mechanism model, and the expression is as follows:
Figure FDA0002641481120000033
ΔT＝-3.973+0.296Rev；
ΔP＝0.473Rev；
Figure FDA0002641481120000034
wherein, VBPIs the volume of the booster pump and has the unit of cc/R, RevIs the rotational speed of the booster pump in rpm, G is the melt flow rate in t/dRho is the melt density in kg/m3，VgIs the inner volume of the booster pump piping, and has a unit of m3；
The expression of the mechanism model of the spinning component section is as follows:
Figure FDA0002641481120000035
Figure FDA0002641481120000036
wherein q is the melt flow rate of a single hole of the distribution plate and is m3/s，l*The length is corrected for the orifice entrance in m.
6. The method of claim 1, wherein the process parameters are temperature, pressure and intrinsic viscosity in units of MPa and Pa-s, respectively.
7. The method of claim 1, wherein the process parameter data acquisition device is a sensor.
8. The method of claim 1, wherein the predictive model is coupled to a processor, and the processor is configured to calculate a difference between an input and an output of the predictive model.
9. The method of claim 8, wherein the prediction model and the processor are connected to a visualization display module, and the visualization display module is used for displaying the data predicted by the prediction model and the data processed by the processor in real time.
CN201910071615.6A 2019-01-25 2019-01-25 Method for adjusting melt quality in fiber production process Active CN109709926B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201910071615.6A CN109709926B (en) 2019-01-25 2019-01-25 Method for adjusting melt quality in fiber production process

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201910071615.6A CN109709926B (en) 2019-01-25 2019-01-25 Method for adjusting melt quality in fiber production process

Publications (2)

Publication Number Publication Date
CN109709926A CN109709926A (en) 2019-05-03
CN109709926B true CN109709926B (en) 2020-11-24

Family

ID=66263026

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201910071615.6A Active CN109709926B (en) 2019-01-25 2019-01-25 Method for adjusting melt quality in fiber production process

Country Status (1)

Country Link
CN (1) CN109709926B (en)

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112379594B (en) * 2020-11-09 2022-04-29 武汉理工大学 Data-driven carbon fiber stock solution temperature control parameter optimization method
CN112831849B (en) * 2020-12-29 2022-04-29 浙江恒逸石化有限公司 Antimony-free polyester melt conveying process optimization method
CN112801398B (en) * 2021-02-07 2024-04-30 盛景智能科技(嘉兴)有限公司 Cooling device failure prediction method and device, electronic equipment and storage medium
CN116736817B (en) * 2023-08-11 2023-11-03 西安启银致远企业管理有限公司 Intelligent control method for hydrophilic silver fiber dressing production and processing

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102906650A (en) * 2010-05-21 2013-01-30 费希尔-罗斯蒙特系统公司 Multi-stage process modeling method
CN108345219A (en) * 2018-03-01 2018-07-31 东华大学 Fypro production technology based on class brain memory GRU

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102906650A (en) * 2010-05-21 2013-01-30 费希尔-罗斯蒙特系统公司 Multi-stage process modeling method
CN108345219A (en) * 2018-03-01 2018-07-31 东华大学 Fypro production technology based on class brain memory GRU

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
细旦涤纶长丝熔体直纺工程;蒋汉雄;《中国优秀硕士学位论文全文数据库工程科技I辑》;20110815;第10-24页 *

Also Published As

Publication number Publication date
CN109709926A (en) 2019-05-03

Similar Documents

Publication Publication Date Title
CN109709926B (en) Method for adjusting melt quality in fiber production process
CN106569056B (en) The method for diagnosing faults and diagnostic device of power transformer
CN109766989B (en) Intelligent configuration method for technological parameters in polyester fiber production process
CN104809306A (en) Injection molding technology parameter optimizing method based on multiple quality indexes
CN103440368A (en) Multi-model dynamic soft measuring modeling method
CN111738482B (en) Method for adjusting technological parameters in polyester fiber polymerization process
CN103226728B (en) High density polyethylene polymerization cascade course of reaction Intelligent Measurement and yield optimization method
CN115936060B (en) Substation capacitance temperature early warning method based on depth deterministic strategy gradient
Nasiri et al. Faults and failures prediction in injection molding process
CN117008557A (en) Production control method and system for blending type interpenetrating network thermoplastic elastomer
CN115320064B (en) Method for monitoring wear condition of extruder in plastic box blow molding process
CN115828466A (en) Fan main shaft component fault prediction method based on wide kernel convolution
Yuan et al. Online adaptive modeling framework for deep belief network-based quality prediction in industrial processes
CN109242093A (en) A kind of method for evaluating reliability of the motor in electric automobile based on fuzzy neural network
CN117828387A (en) Data enhancement method for polyester polymerization process
CN106651461A (en) Film personalized recommendation method based on gray theory
CN111723857B (en) Intelligent monitoring method and system for running state of process production equipment
CN113033079A (en) Chemical fault diagnosis method based on unbalanced correction convolutional neural network
CN115071090B (en) Injection molding quantity dynamic compensation system and method based on injection foaming forming mold
CN115564021A (en) Fault root cause sequencing method in polyester fiber polymerization process
CN113988488B (en) Method for predicting ETC passing probability of vehicle by multiple factors
Malkin et al. Bicomponent flow of molten polymers in annular dies of extruders
CN106202918B (en) A kind of blast furnace molten iron silicon content On-line Estimation method and system
Dutta Role of quench air profiles in multifilament melt spinning of PET fibers
CN114841461A (en) Air quality integration prediction method based on time sequence missing perception and multi-source factor fusion

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant