CN111723513B - Method for inverting simulation parameters through machine learning neural network - Google Patents
Method for inverting simulation parameters through machine learning neural network Download PDFInfo
- Publication number
- CN111723513B CN111723513B CN202010338720.4A CN202010338720A CN111723513B CN 111723513 B CN111723513 B CN 111723513B CN 202010338720 A CN202010338720 A CN 202010338720A CN 111723513 B CN111723513 B CN 111723513B
- Authority
- CN
- China
- Prior art keywords
- neural network
- weight set
- parameters
- machine learning
- model
- 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
Links
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 82
- 238000000034 method Methods 0.000 title claims abstract description 82
- 238000010801 machine learning Methods 0.000 title claims abstract description 52
- 238000004088 simulation Methods 0.000 title claims abstract description 36
- 238000003062 neural network model Methods 0.000 claims abstract description 20
- 238000002347 injection Methods 0.000 claims abstract description 7
- 239000007924 injection Substances 0.000 claims abstract description 7
- 238000004364 calculation method Methods 0.000 claims abstract description 6
- 230000006870 function Effects 0.000 claims description 9
- 238000001746 injection moulding Methods 0.000 abstract description 21
- 238000012549 training Methods 0.000 abstract description 12
- 238000004422 calculation algorithm Methods 0.000 abstract description 8
- 239000004743 Polypropylene Substances 0.000 description 15
- 229920001155 polypropylene Polymers 0.000 description 15
- 239000004033 plastic Substances 0.000 description 13
- -1 polypropylene Polymers 0.000 description 13
- 238000004519 manufacturing process Methods 0.000 description 6
- 239000000463 material Substances 0.000 description 6
- 230000000694 effects Effects 0.000 description 5
- 238000005259 measurement Methods 0.000 description 5
- 238000012360 testing method Methods 0.000 description 5
- 238000005516 engineering process Methods 0.000 description 4
- 238000005206 flow analysis Methods 0.000 description 4
- 238000002844 melting Methods 0.000 description 4
- 230000008018 melting Effects 0.000 description 4
- 238000004458 analytical method Methods 0.000 description 3
- 238000005094 computer simulation Methods 0.000 description 3
- 238000013459 approach Methods 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 238000002474 experimental method Methods 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 239000002994 raw material Substances 0.000 description 2
- 230000003213 activating effect Effects 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000001473 dynamic force microscopy Methods 0.000 description 1
- 239000012530 fluid Substances 0.000 description 1
- 230000004907 flux Effects 0.000 description 1
- 229910052500 inorganic mineral Inorganic materials 0.000 description 1
- 239000000155 melt Substances 0.000 description 1
- 239000011707 mineral Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 239000002861 polymer material Substances 0.000 description 1
- 239000000843 powder Substances 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 238000011084 recovery Methods 0.000 description 1
- 239000000243 solution Substances 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/10—Geometric CAD
- G06F30/17—Mechanical parametric or variational design
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2113/00—Details relating to the application field
- G06F2113/22—Moulding
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Evolutionary Computation (AREA)
- General Physics & Mathematics (AREA)
- Geometry (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Software Systems (AREA)
- Biophysics (AREA)
- Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Data Mining & Analysis (AREA)
- Computational Linguistics (AREA)
- Mathematical Physics (AREA)
- Biomedical Technology (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Computer Hardware Design (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- Pure & Applied Mathematics (AREA)
- Computational Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Medical Informatics (AREA)
- Injection Moulding Of Plastics Or The Like (AREA)
Abstract
The invention discloses a method for inverting simulation parameters through a machine learning neural network, which comprises the following steps: establishing a machine learning neural network; and taking the injection molding simulation software based on the physical model as a learning object, and adjusting the weight of the neural network by using an inversion algorithm in a machine learning algorithm, so that the corresponding result which accords with the calculation of the injection molding simulation software based on the physical model is calculated after the same input parameters are used. The input parameters that should be used in the injection modeling software are then inverted based on the difference between the measured values in the injection molding system quantified by the sensor and the predicted values of the neural network model learned by the training, with the most significant parameters being, but not limited to, the rheological model coefficients. The input parameters subjected to machine learning inversion adjustment can be replaced by injection molding simulation software, so that the difference between the simulation predicted value and the actual measured value can be reduced, and the digital twin of the simulation injection molding process can be realized for intelligent control.
Description
Technical Field
The invention relates to the field of intelligent control of physical model simulation, in particular to a method for inverting simulation parameters through a machine learning neural network.
Background
After the plastic part is designed, the plastic part is manufactured through an injection molding process, certain computer simulation operation is firstly carried out on various working conditions, so that a process condition predicted value is obtained by using a die flow analysis and is used for testing a die regulator, sensor measured values installed in a die are obtained after at least one test of the die regulator, differences exist between the measured values and a die flow software predicted value, and the existence of the differences possibly leads to the failure of establishing a reliable intelligent control system.
The simulation software can quantitatively predict various possible results, provide a set of good reference values for manufacturing and processing, predict possible defects, and provide analysis and comparison for various working conditions and plastic part and mold design changes. These simulation analyses may take hours or days, sometimes even weeks, to arrive at a DOE with multiple manufacturing Designs (DFMs), providing the best choice among the various possibilities. However, these prediction results may still be different from reality in the injection molding process, for example, errors are generated due to the direct difference influence factors between the simulation environment and the real environment, or numerical model errors in the simulation software or programming errors are generated, so that the simulation results may deviate, and the problem of insufficient matching degree between simulation and reality exists.
It is currently most important to find the cause of the discrepancy and correct the input parameters, especially the rheological model parameters, accordingly to match the predictions to the measured data. Therefore, there is a need to re-fit parameters of the rheological model and correct other process parameters that may be erroneous with an in-line sensor system in the actual injection molding of plastic parts.
Machine Learning (ML) technology has been widely used in many fields as an important component of Artificial Intelligence (AI), and has been rapidly developed in recent years. Because of the large data volume in manufacturing process systems due to the sensors installed, in order to achieve an efficient Manufacturing Execution System (MES), it is desirable to propose a method for inverting training neural networks to adjust input parameters using machine learning techniques to minimize the difference between predicted and actual measured values.
Disclosure of Invention
In order to solve the defects in the prior art, the invention provides a method for inverting simulation parameters through a machine learning neural network, which utilizes the machine learning technology to invert the training neural network to adjust input parameters so as to minimize the difference between a simulation predicted value and an actual measured value.
In order to achieve the above purpose, the invention adopts the following specific scheme:
a method for inverting simulation parameters through a machine learning neural network, comprising the steps of:
s1, establishing a machine learning neural network;
s2, providing a physical model for simulation prediction;
S3, inputting the same original parameters into the machine learning neural network and the physical model, inverting and adjusting the operation of the machine learning neural network by using forward-backward propagation by taking the physical model as a learning object to match a third weight set, and minimizing the mean square sum L c-n of the difference value between the predicted value of the neural network and the predicted value of the physical model to obtain a trained third weight set of the neural network;
S4, inverting and adjusting the operation of the machine learning neural network to match with the first weight set and/or the second weight set by using forward-backward propagation to obtain input parameters subjected to inversion adjustment of the neural network with the aim of minimizing the mean square sum L e-n of the difference between the trained neural network predicted value and the measured value of the sensor under the same condition;
s5, substituting the input parameters subjected to inversion adjustment in the S4 into the physical model for recalculation.
Preferably, the machine learning neural network and the original parameters input in the physical model at least comprise process parameters and rheological model parameters.
Preferably, the third weight set is disposed in the neural network hidden layer, and the third weight set is used for matching the trained neural network model to perform prediction calculation.
Preferably, the first weight set and the second weight set are arranged in the neural network input data layer, the first weight set at least comprises a process parameter weight set, the second weight set at least comprises a rheological model parameter weight set, the process parameter weight set is used for matching process parameters, and the rheological model parameter weight set is used for matching rheological model parameters.
Preferably, in the step S3, the method further includes adjusting the third weight value to train the machine learning neural network after minimizing the mean square sum L c-n of the difference values.
Preferably, in the step S4, the method further includes adjusting the process parameter weight set and the rheological model parameter weight set to inversely adjust the input parameters according to the measured values with the aim of minimizing the mean square sum L e- n of the difference values.
Preferably, in the step S4, the process parameter weight set and the rheological model parameter weight set are adjusted, specifically, if the process parameter is input correctly, the rheological model parameter weight set is adjusted by inverting the neural network system, and if the process parameter is input incorrectly, the process parameter weight set is adjusted by inverting the process parameter, so that the mean square sum L e-n of the difference is minimized.
Preferably, in the step S4, the input parameters after inversion adjustment by the neural network are respectively an operation value of the process parameter and the adjusted process parameter weight set, and an operation value of the rheological model parameter and the adjusted rheological model parameter weight value.
Preferably, in the step S3 and the step S4, the mean square sum of differences L c-n and the mean square sum of differences L e-n are calculated by a loss function.
Preferably, the physical model is built by simulation software, and the simulation software is injection molding flow analysis software.
The method disclosed by the invention can quickly realize digital twin effect by integrating the predicted value of simulation software and the measured value of the sensor by using a machine learning neural network to form intelligent control, so that computer simulation generates almost the same result as actual manufacturing, and the difference between the predicted value and the measured value is closed by using the machine learning technology provided by the invention, so that the cause of the difference can be automatically identified and corrected to minimize the difference, and the simulation error is reduced.
Drawings
In order to more clearly illustrate the embodiments of the invention or the technical solutions in the prior art, the drawings that are required in the embodiments or the description of the prior art will be briefly described, it being obvious that the drawings in the following description are only some embodiments of the invention, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a computational block diagram of the present invention;
FIG. 2 is a flowchart of an algorithm of the neural network model of the present invention;
FIG. 3 shows parameters and test results of the injection molding process used in example 1;
FIG. 4 is a Carreau-WLF model parameters for virgin polypropylene and recycled polypropylene;
FIG. 5 shows parameters of the Carreau-WLF model obtained in example 1;
FIG. 6 is a plot showing the convergence of a third set of weights W3 using the Carreau-WLF model in example 1;
FIG. 7 is the parameters of the Cross-WLF model derived by inversion in example 1;
FIG. 8 is a convergence curve of the rheological model parameter weight set W2 in inverting the Cross-WLF model parameters in example 1;
FIG. 9 is the inversion result of the viscosity model for the recycled polypropylene in example 2;
FIG. 10 is the parameters of the injection molding process and test results used in example 3;
FIG. 11 is a die flow analysis of the screw plastic piece used in example 3;
FIG. 12 is the Cross-WLF model parameters according to the inverse of the quadratic viscosity model in example 3.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Example 1
Referring to fig. 1 to 8, a method for inverting simulation parameters by a machine learning neural network includes the following steps:
s1, establishing a machine learning neural network, wherein the machine learning neural network comprises an input layer and an output layer, and the input layer and the output layer are the same as the test conditions of a physical model;
S2, performing simulation prediction by using a simple physical model as a basis, and simulating various possible working conditions of an injection molding process of the designed plastic part; in the embodiment, 18 working conditions of the plastic part are simulated, and input data are shown in fig. 3;
s3, inputting the same original parameters into the machine learning neural network and the physical model according to an injection molding process, inversely adjusting the operation of the machine learning neural network to match with a third weight set W3 by using forward-backward propagation of the physical model as a learning object, and minimizing the mean square sum L c-n of the difference between the calculated value of the neural network and the predicted value of the physical model to obtain a trained third weight set W3 of the neural network; in this embodiment, 18 working conditions of the plastic part are simulated, the original parameters input in the machine learning neural network and the physical model comprise 18 technological parameters, the input data are shown in fig. 3, and the viscosity model parameters are input, as shown in fig. 4, and the original polypropylene plastic part is adopted in this embodiment. The parameters of the obtained Carreau-WLF model are shown in fig. 5, wherein the uppermost curve represents that the melting temperature is 200 ℃, and the melting temperature is 220 ℃ and 240 ℃ downwards in sequence, and if the upper curve is the lower curve, the description is omitted. The convergence curve for the third set of weights W3 was inverted using the Carreau-WLF model, as shown in fig. 6.
S4, the mean square sum L e-n of the difference between the trained neural network predicted value and the measured value of the sensor under the same condition is minimized as a target, and the operation of inverting and adjusting the machine learning neural network is matched with the second weight set W2 and/or the first weight set W1 by using forward-backward propagation so as to obtain the input parameters after inverting and adjusting of the neural network; providing a trained neural network calculated value and a measured value of a sensor under the same condition, and supposing that the input process parameters are correct, inverting and adjusting a second weight set W2 of a rheological model Cross-WLF by using the measured value of the sensor, and minimizing a mean square sum L e-n of a difference value between the measured value of the sensor and a trained neural network predicted value to obtain a Cross-WLF viscosity model parameter after inversion adjustment of the neural network; the calculated Cross-WLF model parameter curve and parameters are inverted, and the convergence curve of the second weight set W2 when inverting the Cross-WLF model parameters is respectively shown in FIG. 7 and FIG. 8.
S5, substituting the input parameters subjected to inversion adjustment in the S4 into a physical model for recalculation.
Specifically, the viscosity model parameters of the Cross-WLF after inversion adjustment in S4 are plotted and compared with another viscosity model Carreau-WLF and the resulting parameter plots, that is, as shown in FIGS. 5 and 7.
In this embodiment, a method is proposed for applying to inverting viscosity model parameters in a physical model in an injection molding process using a machine learning neural network. In this embodiment, the actual situation is reflected by using the measured value of the sensor, and the physical model in this embodiment calculates the predicted value by using the physical model of the injection mold flow, and in other embodiments, other physical models may be used, and simulation software for analysis of other injection mold flows may be used.
At least the following input parameters may be used to calculate the difference between the injection molding flow analysis software predictions and measurements: namely, the process parameters at least comprise process setting parameters and boundary condition parameters, wherein the process parameters comprise parameters such as a speed curve, a melt temperature, a V/P conversion percentage, a pressure maintaining condition and the like, and the boundary conditions comprise parameters such as a mold surface temperature/heat flux condition, a sliding mold wall, a vent hole size/position, mold deformation in an injection process and the like. And the rheological model parameters include at least systolic fluid pressure and PVT model parameters. Of course, in other embodiments, the corresponding parameters may be adjusted according to possible factors, such as the measuring device and its accuracy parameters, and the solver accuracy parameters programmed into the model flow analysis software.
In this embodiment, a simulation predicted value R c is obtained by inputting a process parameter and a rheological model parameter into injection molding flow analysis software, and simultaneously the same process parameter and rheological model parameter are also input into an established machine learning neural network model, the neural network model uses the injection molding flow analysis software as a learning object to reversely adjust the operation of the machine learning neural network to match with a third weight set W3 so as to minimize the difference between the trained neural network calculated value R n and the physical model predicted value R c, in this process, the weight value is adjusted through an inversion algorithm, the third weight set W3 is arranged in a neural network hidden layer and is used for matching with the trained neural network calculated value R n so as to train the machine learning neural network model to minimize the difference between the neural network calculated value R n and the physical model predicted value Rc, and a final neural network calculated value R n is obtained. The difference value minimization is obtained through loss function calculation, and the specific formula is as follows: min { L c-n＝∑(Rc-Rn)2 }, where L c-n represents the mean square sum of the differences between the neural network calculated value Rn and the physical model predicted value Rc, the difference referred to in this implementation is minimized by making L c-n approach 0, and by continuously adjusting the third weight set W3 in the inversion algorithm process, the mean square sum L c-n of the differences between the neural network calculated value R n and the physical model predicted value R c gradually becomes stable, so that its gradient becomes zero. The inversion algorithm used for the third weight set W3 in this embodiment is a random gradient descent (SGD) algorithm, or any other optimization model that runs rapidly, so as to adjust the neural network model learning rate (iteration step) to achieve rapid convergence.
And calculating the mean square sum L e-n of the difference between the measured value R e of the sensor and the calculated value R n of the neural network, and providing the measured value R e of the sensor under the same condition, wherein the measured value inversion of the sensor adjusts the calculation of the machine learning neural network to match the second weight set W2 and/or the first weight set W1, and the mean square sum L e-n of the difference is minimized by continuously inverting the adjustment weight set. The second weight set W2 and the first weight set W1 are arranged in the neural network input data layer, the weight set arranged in the neural network input data layer at least comprises a process parameter weight set W1 and a rheological model parameter weight set W2, the process parameter weight set W1 is used for matching process parameters, the rheological model parameter weight set W2 is used for matching rheological model parameters, the mean square sum minimization of the difference is calculated through a loss function, and the specific formula is as follows: min { L e-n＝∑(Re-Rn)2 }, where L e-n represents the sum of the mean square differences of the differences between the sensor measurement R e and the neural network calculated R n, and similarly, obtaining the input parameters after inversion adjustment of the neural network by continuously adjusting the weight values in the inversion algorithm process to minimize L e-n to approach 0; specifically, a process parameter weight set W1 and a rheological model parameter weight set W2 are adjusted to train a machine learning neural network model, if the process parameters are input correctly, the rheological model parameter weight set W2 is adjusted by inverting the neural network system, and if the process parameters are input incorrectly, the process parameter weight set W1 is adjusted by inverting the process parameters; specifically, assuming that the process parameters are correctly input, the process parameter weight set W1 multiplied by the process condition parameters is fixed, and the rheological model parameter weight set W2 for adjusting the rheological model parameters is counter-propagated. However, this may not be the case in practical applications of simulation models and equivalent neural network models, because the process parameters and the actual effects of each injection molding machine may be slightly different, and in other cases, human input errors may be present. Thus, if it is found that the adjustment of the process parameter weight set W1 does contribute to minimizing the loss function L e-n, the vector values of the process parameter weight set W1 can be adjusted. Firstly, identification of process parameter input correctness is achieved through back propagation of a rheological model parameter weight set W2, when the rheological model parameter weight set W2 converges to a minimum loss value, a small difference still exists between a simulation model predicted value R c and a measured value R e, and then the back propagation process parameter weight set W1 can be activated through fixing the rheological model parameter weight set W2. Likewise, when the adjustment to the process parameter weight set W1 with back propagation no longer works to minimize the loss function L e-n, activating the adjustment to the parametric model weight set W2 will continue to minimize the loss function L e-n. In other words, the adjustment of the process parameter weight set W1 and the rheological model parameter weight set W2 by back propagation may be performed iteratively. The gradient of the rheological model parameter weight set W2 in this embodiment may be derived by a chained rule in which a neural network differentiates a complex relationship.
After the difference loss function L e-n is minimized, at this time, the input parameters after inversion adjustment by the neural network are respectively the operation value of the process parameter and the adjusted process parameter weight set W1 and the operation value of the rheological model parameter and the adjusted rheological model parameter weight value W2, and the operation method adopted in this embodiment is multiplication, that is, the parameters finally input into the physical model for simulation are the product of the process parameter and the adjusted process parameter weight set W1 and the product of the rheological model parameter and the adjusted rheological model parameter weight value W2.
The finite element method used in the simulation of the physical model is that a three-dimensional space approximately represents an object by using a geometric grid formed by units and nodes, and the discrete form of the physical equation model is realized on the units and the nodes to construct an overall matrix equation formed by a unit matrix, so that the problem to be solved is described by a numerical equation set. Such a system of equations is based on a geometrical connection, wherein each node is contributed by adjacent connection units. The neural network model has similar relations, and each node contributes to other nodes in the form of weights by using the topological relation among the nodes of the neural network.
The neural network model is superior to simulation based on a physical model in calculation speed, can learn different objects, can learn the physical model, can train according to measured data, and has better effect and smaller error value as the data is more. The method disclosed by the invention can quickly realize digital twin effect by integrating the predicted value of the physical model and the measured value of the sensor by using a machine learning neural network, so that computer simulation generates nearly the same result as actual manufacturing, and the difference between the predicted value and the measured value is closed by using the machine learning technology provided by the invention, thereby reducing simulation errors.
Example 2
The present embodiment specifically describes machine learning neural network model training, and differs from the above embodiment in that inversion is performed on rheological model parameters of recycled polypropylene plastic corresponding to the original plastic in embodiment 1, where the rheological model parameters specifically refer to parameters including at least one set of shear viscosity model parameters and one set of additional model parameters for calculating pressure loss in the case of shrinkage flow. This additional model requires that a pressure sensor be placed before and after the contracted flow, respectively, to measure the pressure loss of the contracted flow. Designing and using at least two or more different systolic flows may better provide for measuring additional model parameters.
On the measurement data, two forms can be accepted:
1) A constant value at a certain moment, for example a moment corresponding to the measured peak value of one sensor, and the measured values of other sensors at the same moment.
2) Time series of all sensor measurements.
The Cross-WLF viscosity coefficient of the polymer material tested by the instrument forming system is derived by the method provided by the invention, the used process conditions are listed in figure 3, wherein the process conditions comprise three melt temperatures, namely 200 ℃, 220 ℃, 240 ℃ respectively, corresponding injection speeds, and pressure values measured by two pressure sensors are shown in a table, the material used for the experiment is polypropylene (PP) mixed with 10% mineral filling powder, hostacom CR 1171G1A, and the recovery material corresponding to the original material is tested.
The viscosity model employed in examples 1 and 2 was a modified Carreau-WLF model, the model equation being:
Alpha T is represented in the following manner:
wherein η is viscosity (pa.s); k i is a model coefficient; is the shear rate (1/s); t is the temperature (. Degree. C.)
The wall shear stress is calculated herein by the following formula:
Δp is the difference in pressure (Pa) measured at the two sensor locations; a (=0.003 m) is the height of the rectangular cross section of the measurement channel, and b (=0.02 m) is the width; l (=0.1 m) is the distance between two sensor locations measuring the pressure drop.
The shear rate is calculated according to the Weissenberg-Rabinowitsch equation:
Wherein the apparent shear rate Is calculated using the following formula with form factor F p:
where Q is the injection volumetric flow rate.
Viscosity is the ratio of shear stress to shear rate, and equation (3-5) is a physical model, which is obviously an approximation model, since thermal effects are not considered, i.e., the system is assumed to be isothermal. The coefficient d in equation (4) and Fp in equation (5) are correction coefficients, fp=0.919 for the selected geometry.
The viscosity model coefficients of the original material and the corresponding recycled polypropylene material are shown in fig. 4, and the Carreau-WLF model fitted with the coefficients in example 1 is shown in fig. 5, wherein the uppermost curve represents the melting temperature of 200 ℃, 220 ℃ and 240 ℃ in sequence, and if yes, the description is omitted. The raw material is used as a training model of the neural network model of the present invention by using a Carreau-WLF model to train a third weight set W3, the loss function convergence curve of the third weight set W3 is shown in fig. 6, and the inversion result of the recycled polypropylene plastic corresponding to the raw material in this embodiment is shown in fig. 9. The neural network model with this third weight set W3 is then used as a training model to fit the coefficients of the Cross-WLF model to train the second weight set W2 of rheological model parameters. The Cross-WLF model is as follows:
Wherein the zero shear rate viscosity is:
Wherein the method comprises the steps of And D 1,D2,D3,n,A1,/>Τ * is the model coefficient. It should be noted that the pressure correlation coefficient D 3 is an important parameter that affects the pressure prediction in the injection molding simulation.
The graph of the Cross-WLF model coefficient and the viscosity model after the original polypropylene is fitted is shown in fig. 7, wherein the graph comprises a set of model coefficients of the original PP, the training model for training the neural network model of the present invention trains the rheological model parameter weight set W2, and the convergence graph of the neural network model of the rheological model parameter weight set W2 is shown in fig. 8.
Similarly, the polypropylene recovered in this example was also subjected to machine learning neural network inversion training to verify the present invention in the same manner as the original polypropylene. The Carreau-WLF model and Cross-WLF model coefficients of the recovered polypropylene and the trained third weight set W3 and rheological model parameter weight set W2 are shown in fig. 9.
Example 3
In this embodiment, a spiral mold is provided, the physical model of this embodiment is injection molding flow analysis software, and in other embodiments, any physical model of injection molding flow analysis may be used, and three melt temperature conditions are used, namely 220 ℃, 240 ℃, 260 ℃ with corresponding speed and pressure values at each melt temperature, and specific values are listed in fig. 10. The material used for the experiment was polypropylene STAMYLAN PHC, as shown in fig. 11, which is a model flow analysis of the screw plastic piece used, and the screw mold was subjected to a simulation operation to provide a predictive equation based on a second order viscosity model, as follows:
wherein η is viscosity (pa.s); a i model coefficients; Shear rate (1/s); t temperature (. Degree. C.).
Similarly, in this embodiment, the second-order viscosity model is used as a training model for the neural network model of the present invention to train the third weight set W3, then the neural network model with this third weight set W3 is used as a training model to fit to obtain a Cross-WLF model, and then the rheological model parameter weight set W2 is trained with the Cross-WLF model coefficients. Fig. 12 shows graphs of the second order viscosity model coefficients provided, fitted Cross-WLF model coefficients based on the invention, and convergence graphs of neural network models of the third set of weights W3 and the set of rheological model parameters W2 after training, wherein the uppermost graph in the graphs of the second order viscosity model coefficients and Cross-WLF model coefficients represents a melting temperature of 220 ℃ and 240 ℃ and 260 ℃ in order.
In the present specification, each embodiment is described in a progressive manner, and each embodiment is mainly described in a different point from other embodiments, and identical and similar parts between the embodiments are all enough to refer to each other.
The previous description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the invention. Thus, the present invention is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope possible consistent with the principles and novel features disclosed herein.
Claims (8)
1. A method for inverting simulation parameters through a machine learning neural network, comprising the steps of:
S1, establishing a machine learning neural network;
S2, providing a physical model for simulation prediction;
S3, inputting the same original parameters into the machine learning neural network and the physical model, inverting and adjusting the operation of the machine learning neural network by using forward-backward propagation by taking the physical model as a learning object to match a third weight set, and minimizing the mean square sum L c-n of the difference value between the predicted value of the neural network and the predicted value of the physical model to obtain a trained third weight set of the neural network; the third weight set is arranged in the neural network hidden layer and is used for matching the trained neural network model to perform prediction calculation;
S4, inverting and adjusting the operation of the machine learning neural network to match with the first weight set and/or the second weight set by using forward-backward propagation to obtain input parameters subjected to inversion adjustment of the neural network with the aim of minimizing the mean square sum L e-n of the difference between the trained neural network predicted value and the measured value of the sensor under the same condition; the first weight set and the second weight set are arranged in the neural network input data layer, the first weight set at least comprises a technological parameter weight set, the second weight set at least comprises a rheological model parameter weight set, the technological parameter weight set is used for matching technological parameters, and the rheological model parameter weight set is used for matching rheological model parameters;
s5, substituting the input parameters subjected to inversion adjustment in the S4 into the physical model for recalculation.
2. A method of inverting simulation parameters via a machine learning neural network as claimed in claim 1, wherein: the original parameters input in the machine learning neural network and the physical model at least comprise process parameters and rheological model parameters.
3. A method of inverting simulation parameters via a machine learning neural network as claimed in claim 1, wherein: in the step S3, further comprising adjusting the third weight value to train the machine learning neural network after minimizing the mean square sum of the differences L c-n.
4. A method of inverting simulation parameters via a machine learning neural network as claimed in claim 1, wherein: in the step S4, the method further includes adjusting the process parameter weight set and the rheological model parameter weight set with the aim of minimizing the mean square sum L e-n of the difference value, so as to inversely adjust the input parameters according to the measured values.
5. A method of inverting simulation parameters via a machine learning neural network as claimed in claim 4, wherein: in the step S4, the process parameter weight set and the rheological model parameter weight set are adjusted, specifically, if the process parameter is input correctly, the rheological model parameter weight set is adjusted by inverting the neural network system, and if the process parameter is input incorrectly, the process parameter weight set is adjusted by inverting the process parameter, so that the mean square sum L e-n of the difference is minimized.
6. A method of inverting simulation parameters via a machine learning neural network as claimed in claim 4, wherein: in the step S4, the input parameters after inversion adjustment by the neural network are respectively the operation values of the technological parameters and the adjusted technological parameter weight set, and the operation values of the rheological model parameters and the adjusted rheological model parameter weight values.
7. A method of inverting simulation parameters via a machine learning neural network as claimed in claim 1, wherein: in the step S3 and the step S4, the mean square sum of differences L c-n and the mean square sum of differences L e-n are calculated by a loss function.
8. A method of inverting simulation parameters via a machine learning neural network as claimed in claim 1, wherein: the physical model is established through simulation software, and the simulation software is injection mold flow analysis software.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010338720.4A CN111723513B (en) | 2020-04-26 | 2020-04-26 | Method for inverting simulation parameters through machine learning neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010338720.4A CN111723513B (en) | 2020-04-26 | 2020-04-26 | Method for inverting simulation parameters through machine learning neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111723513A CN111723513A (en) | 2020-09-29 |
| CN111723513B true CN111723513B (en) | 2024-04-26 |
Family
ID=72563678
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010338720.4A Active CN111723513B (en) | 2020-04-26 | 2020-04-26 | Method for inverting simulation parameters through machine learning neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111723513B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112487715A (en) * | 2020-11-27 | 2021-03-12 | 江苏科技大学 | Method for optimizing reliability of process parameters of key hole system of marine diesel engine body |
| CN113537354B (en) * | 2021-07-19 | 2022-07-12 | 吉林大学 | Aquifer structure staged stochastic inversion identification method based on deep learning |
| CN114596919B (en) * | 2022-05-10 | 2022-09-02 | 安徽瑞邦数科科技服务有限公司 | Index prediction method and system and application thereof in phosphoric acid production |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR20070025111A (en) * | 2005-08-31 | 2007-03-08 | 동명대학교산학협력단 | A neural network-based warpage estimation method of an injection-molded thin plastic part |
| CN103914581A (en) * | 2013-12-27 | 2014-07-09 | 华中科技大学 | Optimization method for plastic injection molding technological parameter |
| CN110377948A (en) * | 2019-06-12 | 2019-10-25 | 江苏师范大学 | A kind of injection parameters Multipurpose Optimal Method |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| DE10119853A1 (en) * | 2001-04-24 | 2003-01-09 | Bayer Ag | Hybrid model and method for determining mechanical properties and processing properties of an injection molded part |
| US8589134B2 (en) * | 2011-02-11 | 2013-11-19 | Gentex Optics, Inc. | Method and system employing flow simulation for improving material delivery in lens manufacturing |
-
2020
- 2020-04-26 CN CN202010338720.4A patent/CN111723513B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR20070025111A (en) * | 2005-08-31 | 2007-03-08 | 동명대학교산학협력단 | A neural network-based warpage estimation method of an injection-molded thin plastic part |
| CN103914581A (en) * | 2013-12-27 | 2014-07-09 | 华中科技大学 | Optimization method for plastic injection molding technological parameter |
| CN110377948A (en) * | 2019-06-12 | 2019-10-25 | 江苏师范大学 | A kind of injection parameters Multipurpose Optimal Method |
Non-Patent Citations (1)
| Title |
|---|
| 考虑压力对黏度影响的注塑填充过程数值模拟;麻向军;模具技术(2010年第1期);第11-13页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111723513A (en) | 2020-09-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111723513B (en) | Method for inverting simulation parameters through machine learning neural network | |
| CN109255160B (en) | Neural network-based unit delay prediction method and unit delay sensitivity calculation method | |
| Fratczak et al. | Virtual commissioning for the control of the continuous industrial processes—Case study | |
| CN105678015B (en) | A kind of Multidisciplinary systems pneumatic structure coupling optimum design method of hypersonic three-dimensional wing | |
| Taghizadeh et al. | Warpage prediction in plastic injection molded part using artificial neural network | |
| EP3520987A1 (en) | Method for monitoring and control of the injection process of plastic materials | |
| CN114818549B (en) | Method, system, equipment and medium for calculating fluid mechanics parameter of object | |
| Zhu et al. | A data-driven approach for turbulence modeling | |
| Ke et al. | Enhancement of multilayer perceptron model training accuracy through the optimization of hyperparameters: a case study of the quality prediction of injection-molded parts | |
| CN114912364A (en) | Natural gas well flow prediction method, device, equipment and computer readable medium | |
| Holland | Integrated field inversion and machine learning with embedded neural network training for turbulence modeling | |
| Toukhtarian et al. | Parameter identification of transport PDE/nonlinear ODE cascade model for polymer extrusion with varying die gap | |
| Ma et al. | Intelligent recommendation system of the injection molding process parameters based on CAE simulation, process window, and machine learning | |
| Yu et al. | Offline prediction of process windows for robust injection molding | |
| CN111625995B (en) | Online time-space modeling method integrating forgetting mechanism and double ultralimit learning machines | |
| Srinivasarengan et al. | Nonlinear joint state-parameter observer for VAV damper position estimation | |
| KR102091815B1 (en) | Intelligent super precision plastic mold design apparatus | |
| CN110705187B (en) | Instant on-line instrument checksum diagnosis method through least square algorithm | |
| Huang et al. | Hyperparameter optimization strategy of multilayer perceptron model for injection molding quality prediction | |
| Faes et al. | Product reliability optimization under plate sheet forming process variability | |
| CN113887026A (en) | Oil pipeline hydraulic prediction method based on ensemble learning | |
| CN114297940B (en) | Method and device for determining unsteady state reservoir parameters | |
| CN114861912B (en) | Data verification method and device based on big data | |
| CN117973236B (en) | Correction method, system, equipment and storage medium for simulation result | |
| Sardiña | Multi-Fidelity Data Fusion With Uncertainty Quantification |
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 |