CN101901283A - Prediction method of numerical control bending forming quality of conduit and device - Google Patents
Prediction method of numerical control bending forming quality of conduit and device Download PDFInfo
- Publication number
- CN101901283A CN101901283A CN 201010205834 CN201010205834A CN101901283A CN 101901283 A CN101901283 A CN 101901283A CN 201010205834 CN201010205834 CN 201010205834 CN 201010205834 A CN201010205834 A CN 201010205834A CN 101901283 A CN101901283 A CN 101901283A
- Authority
- CN
- China
- Prior art keywords
- conduit
- numerical control
- bending forming
- parameter
- control bending
- 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.)
- Pending
Links
Images
Landscapes
- Media Introduction/Drainage Providing Device (AREA)
Abstract
The invention provides a prediction method of the numerical control bending forming quality of a conduit and a device. The method comprises the following steps: building a finite element model for simulating the technical process of the numerical control bending forming of the conduit; organizing the numerical simulation experiment of the finite element of the numerical control bending forming of the conduit by the finite element model, obtaining a first experiment result, and determining the technological parameter and the conduit design parameter which have obvious impact on numerical control bending forming quality of the conduit according to the first experiment result; taking the technological parameter and the conduit design parameter which have obvious impact on the numerical control bending forming quality of the conduit as an input parameter, and taking the preset quality index as an output parameter to build an artificial neural network model; training the artificial neural network model; and utilizing the trained artificial neural network model to predict the numerical control bending forming quality of the conduit. When the above technical scheme is adopted, a user can quickly predict the bending forming quality index of the conduit after the conduit parameter is determined; and the invention has favorable timeliness and high prediction precision.
Description
Technical field
The present invention relates to Plastic Forming prediction of quality field, particularly relate to a kind of Forecasting Methodology and device of numerical control bending forming quality of conduit.
Background technology
Along with the expansion of conduit range of application, the conduit plastic working has become an important directions of advanced Technology of Plastic Processing research and development.Because the compacting mechanism complexity that hollow, thin-wall pipes are gone for a stroll, to lateral wall attenuate in the BENDING PROCESS, break, madial wall thickens, wrinkling, xsect elliptical distortion and catheter center's line elongation, and prediction, optimization and the evolutionary process analysis of the resilience of unloading back be to comprise that engineering circle of tube bending forming fails the technical barrier that effectively solves always, also is difficult point and the focus that current plastic working subject is both at home and abroad studied.Along with the application of heavy caliber, thin-walled, small-bend radius and difficult-to-deformation material conduit, the problems referred to above become increasingly conspicuous.
Be to solve the difficulty in the conduit actual production, U.S. Eton Leonard company at first development and production a kind of computer control numerical control (CNC) vector bending tube equipment, be commonly each aeromotor company of west in recent years and adopt.The method that adopts this kind equipment to carry out bend pipe is referred to as numerical-controlled bending, gains the name because of clamping die and bending die clamping tubing rotatablely move around the bending die center.Fig. 1 is the principle of work synoptic diagram of the numerical controlled bending of pipe of an example of prior art.
The factor that influences the forming tubular product quality is many and complicated, also has coupled relation between the factor, and product variety is various in addition, and therefore theoretical parsing is very difficult, relies on deviser's experience and trial-production repeatedly for a long time.And experimental study cost height, cycle are long, and the test data error is bigger.In order to reach the specific (special) requirements that tubing is gone for a stroll and processed, adopt finite element method for simulating tube bending forming process.Non-linear demonstration Dynamic Finite Element Analysis method is applicable to the simulation to METHOD IN METAL FORMING PROCESSES, its essence is a kind of passing through the labyrinth discretize, and then the method for numerical simulation of simplification METHOD FOR LARGE DEFORMATION ELASTOPLASTIC physical process, make computing machine to find the solution challenge by extensive matrix, this method precision height, abundant information, but because method complexity, the calculated amount of finite element analogy are huge, when causing utilizing finite element analogy to carry out the guiding-tube bend prediction of quality, computing time is long, and inconvenient user uses.
Summary of the invention
Purpose of the present invention provides a kind of Forecasting Methodology and device of numerical control bending forming quality of conduit, and is consuming time many when carrying out prediction of quality with the method for passing through finite element analogy that solves prior art, the technical matters of poor in timeliness.
To achieve these goals, the invention provides a kind of Forecasting Methodology of numerical control bending forming quality of conduit, wherein, comprise the steps:
Step 1 is set up the finite element model in order to simulate catheter numerical control bending forming technological process;
Step 2, utilize described finite element model, the finite element numerical simulation test of tissue tract numerical control bending forming obtains first test findings, and determining according to described first test findings influences significant technological parameter and catheter design parameter to numerical control bending forming quality of conduit;
Step 3, with described determine numerical control bending forming quality of conduit is influenced significant technological parameter and catheter design parameter as input parameter, as output parameter, set up artificial nerve network model with predetermined quality index;
Step 4 is trained described artificial nerve network model, and utilizes housebroken artificial nerve network model to carry out the prediction of numerical control bending forming quality of conduit.
Preferably, described method wherein, before the described step 4, also comprises:
With described determine numerical control bending forming quality of conduit is influenced significant technological parameter and catheter design parameter as design factor, the experiment of tissue tract numerical control bending forming or the numerical experiments of finite element think that described artificial nerve network model provides the initial training sample.
Preferably, described method wherein, is organized the numerical experiments of described conduit numerical control bending forming finite element based on orthogonal design method.
Preferably, described method, wherein, in the described step 3, described artificial nerve network model is: radially basic artificial nerve network model.
Preferably, described method, wherein, in the described step 4,
Before described artificial nerve network model is trained, also comprise: described initial training sample data is carried out normalized;
The described step of utilizing described housebroken artificial nerve network model to carry out the prediction of numerical control bending forming quality of conduit comprises: before the input parameter value is imported described artificial nerve network model, carry out normalized earlier; And, after obtaining the output valve of described artificial nerve network model, described output valve is carried out obtaining after the anti-normalized quality index value of the numerical control bending forming of described conduit.
Preferably, described method, wherein, in the described step 3, the described step of setting up artificial nerve network model comprises:
A plurality of artificial nerve network models are set up in different segmentations according to the geometric parameter of the difference of tube material and/or conduit.
Preferably, described method, wherein, the geometric parameter of described conduit comprises: the conduit external diameter.
Preferably, described method wherein, describedly influences significant technological parameter to numerical control bending forming quality of conduit and comprises: pressing mold pressure, boosting power, mandrel overhang and mandrel gap; Describedly numerical control bending forming quality of conduit is influenced significant catheter design parameter comprise: conduit external diameter, wall thickness, bending radius and angle of bend.
On the other hand, provide a kind of prediction unit of numerical control bending forming quality of conduit, wherein, comprising:
Finite element model is set up module, is used to set up the finite element model in order to simulate catheter numerical control bending forming technological process;
The parameter determination module, be used to utilize described finite element model, the finite element numerical simulation test of tissue tract numerical control bending forming obtains first test findings, and determining according to described first test findings influences significant technological parameter and catheter design parameter to numerical control bending forming quality of conduit;
Neural network module, be used for described determine numerical control bending forming quality of conduit is influenced significant technological parameter and catheter design parameter as input parameter, as output parameter, set up artificial nerve network model with predetermined quality index;
Prediction module is used for described artificial nerve network model is trained, and utilizes housebroken artificial nerve network model to carry out the prediction of numerical control bending forming quality of conduit.
Preferably, described prediction unit, wherein, in the described prediction module, the initial training sample that is used for described artificial nerve network model is trained is:
By with described numerical control bending forming quality of conduit influenced significant technological parameter and catheter design parameter obtaining of determining as the experiment of the conduit numerical control bending forming of design factor, tissue or the numerical experiments of finite element.
Preferably, described prediction unit, wherein, based on the orthogonal design method tissue during numerical experiments of described conduit numerical control bending forming finite element.
Preferably, described prediction unit, wherein, the described artificial nerve network model that described neural network module is set up is: radially basic artificial nerve network model.
Preferably, described prediction unit, wherein, described neural network module is further used for setting up a plurality of artificial nerve network models according to the different segmentations of the geometric parameter of the difference of tube material and/or conduit.
Preferably, described prediction unit, wherein, described parameter determination module is determined describedly influences significant technological parameter to numerical control bending forming quality of conduit and comprises: pressing mold pressure, boosting power, mandrel overhang and mandrel gap; Describedly numerical control bending forming quality of conduit is influenced significant catheter design parameter comprise: conduit external diameter, wall thickness, bending radius and angle of bend.
A technical scheme in the technique scheme has following technique effect:
By setting up simulate catheter numerical control bending forming technological process finite element model, utilize finite element model to carry out numerical experiments, having determined more exactly influences significant technological parameter and catheter design parameter to numerical control bending forming quality of conduit, and definite numerical control bending forming quality of conduit is influenced significant parameter as input with above-mentioned, set up the artificial nerve network model that is used to carry out conduit numerical control bending forming technology as output with predetermined quality index, considered the bigger factor of bending forming influence, ignored the less parameter of bending forming influence, under the prerequisite that guarantees precision of prediction, simplified model, make that the user can be after the conduit parameter be determined, dope the quality index that guiding-tube bend is shaped faster, ageing good.
Another technical scheme in the technique scheme has following technique effect:
By with described determine to numerical control bending forming quality of conduit influence significant technological parameter and catheter design parameter as design factor, organize the numerical experiments of finite element, and utilize test for data to provide the initial training sample as artificial nerve network model, can realize predicting by the quality index of already known processes parameter and catheter design parameter and correspondence thereof the quality index of unknown technological parameter and conduit parameter correspondence, instant processibility analysis and process optimization during for the conduit process planning provide support.
Description of drawings
Fig. 1 is the principle of work synoptic diagram of the numerical controlled bending of pipe of an example of prior art;
Fig. 2 is the schematic flow sheet of Forecasting Methodology of the numerical control bending forming quality of conduit of the embodiment of the invention;
Fig. 3 is the schematic flow sheet of finite element analogy.
Embodiment
For making the purpose, technical solutions and advantages of the present invention clearer, describe the present invention below in conjunction with the accompanying drawings and the specific embodiments.
Fig. 2 is the schematic flow sheet of Forecasting Methodology of the numerical control bending forming quality of conduit of the embodiment of the invention.As Fig. 2, the Forecasting Methodology of the embodiment of the invention comprises the steps:
Preferably, above-mentionedly numerical control bending forming quality of conduit is influenced significant technological parameter comprise: pressing mold pressure, boosting power, mandrel overhang and mandrel gap; Describedly numerical control bending forming quality of conduit is influenced significant catheter design parameter comprise: conduit external diameter, wall thickness, bending radius and angle of bend;
Preferably, before the above-mentioned steps 204, also comprise: with described determine numerical control bending forming quality of conduit is influenced significant technological parameter and catheter design parameter as design factor, the experiment of tissue tract numerical control bending forming or the numerical experiments of finite element think that described artificial nerve network model provides the initial training sample.Particularly, utilize repeatedly the test figure in the numerical experiments of finite element, comprise that quality index value known in each time test, that numerical control bending forming quality of conduit is influenced the value correspondence of the value of the value of the value of significant technological parameter and catheter design parameter and this technological parameter and catheter design parameter provides the initial training sample to come the artificial nerve network model of setting up is trained as artificial nerve network model.Can improve the reliability of neural network prediction like this, can realize predicting by the quality index of already known processes parameter and catheter design parameter and correspondence thereof the quality index of unknown technological parameter and conduit parameter correspondence, instant processibility analysis and process optimization during for the conduit process planning provide support.Exemplarily, when obtaining the initial training sample by experiment, can utilize the computer control numerical control vector bending tube equipment of prior art to experimentize.
Preferably, in the method for the embodiment of the invention, based on the numerical experiments of orthogonal design method tissue tract numerical control bending forming finite element.
Preferably, in the method for the embodiment of the invention, in the described step 203, described artificial nerve network model is: radially basic artificial nerve network model
Preferably, in the method for the embodiment of the invention, the step of setting up artificial nerve network model comprises: a plurality of artificial nerve network models are set up in the different segmentations according to the geometric parameter of the difference of tube material and/or conduit.Exemplarily, the geometric parameter of above-mentioned conduit comprises: the conduit external diameter.
To the finite element model of setting up in the Forecasting Methodology of the conduit numerical control bending quality of the embodiment of the invention, simulate catheter numerical control bending forming technological process is specifically described below.
Show that the dynamics finite element analogy is the physical simulation to real world, different with how much simple motion simulations with the static(al) simulation, most physical attributes of simulated object have been considered in the finite element analogy of demonstration dynamics, comprise material physical property, original state, stand under load situation, inertial force, the dynamic change of Contact Boundary and the large deformation of object etc. of object in the simulated object, thereby be applicable to simulation processes such as metal forming, collision, blasts.Finite element analysis is a kind of board design load-up condition, and determines the method for all kinds of responses under load-up condition.Finite Element Method is based upon on the solid flow variational principle basis, after analyzed object is separated into many junior units, given boundary condition, load and material behavior are found the solution linearity or Nonlinear System of Equations, just can obtain the result such as displacement, strain, stress, internal force of analytic target.By The present computer technology, these steps can comparatively fast be finished, and can use the explicit result of calculation of graph technology.Adopt the support programs of LS-DYNA in the embodiments of the invention as finite element analogy.
The process of finite element analogy mainly is divided into: pre-treatment promptly sets up models treated, find the solution and three big steps of analog result are promptly checked in aftertreatment, and each big step can be subdivided into some little steps again.Detailed demonstration dynamics finite element analogy flow process as shown in Figure 3.As Fig. 3, the flow process of finite element analogy comprises the steps:
In this step,, find out its central factor, take out the physical model that finite element can be simulated by suitable simplification according to the technological problems of Tube Bending.
Wherein, which kind of cell type is unit selection promptly select for use object is carried out discretize, comprises bar unit, shell unit, spring damping unit, cable unit etc., and suitable unit with the thin walled tube bending forming is a shell unit.
The definition of material model comprises two aspects: select the concrete numerical value of parameter in material constitutive model and the definite constitutive model, as elastic modulus, yield strength etc.The former need in a large amount of constitutive models that program provides, select one can reflect material deformation in the process be reflected to the influential Main physical rule of result, and ignore unessential characteristic; The latter need obtain by the precise material Experiments of Machanics.
At first set up its geometric model according to the guiding-tube bend physical model that the first step takes out, according to cell type and material model the geometry entity model carry out finite element gridization again, obtain unit and nodal information.
Starting condition such as the initial velocity of each parts (part), initial stress in the definition model, and boundary constraint, as: the symmetrical border of 1/2 model, the constraint of each parts, degree of freedom etc.
Contact type between the object that definition may come in contact and parameter are as dynamic and static friction factor, Contact Algorithm, penalty factor etc.
Above-mentioned steps 301 to step 305 belongs to pre-treatment.
Step 306 is found the solution setting and is found the solution;
The output file form of result of calculation is set, and the computing information that comprises in the destination file etc., as computing time step-length, animation output frame number etc.The file that calls at last after solver is finished pre-treatment is found the solution.
Step 307 is dynamically checked forming process information;
Call the poster processing soft, continuously, dynamically check the wall thickness change, elliptical distortion of conduit in the BENDING PROCESS, macroscopic information such as wrinkling, and ess-strain distribution, pressure distribution etc., analyze the rationality of its forming process.
The information that comprises in the result of calculation file is exported, after over-fitting, comparative analysis, calculating, obtained the forming quality index of conduit.
The method of the embodiment of the invention is used ANSYS/LS-DYNA and has been set up the finite element model that is identical with the actual process process on the basis of analysing in depth conduit forming technology process and processing mechanical mechanism thereof.Modeling process adopts the population parameter modeling method based on ANSYS Parametric Design Language (APDL), and as driving parameters, die parameters etc. are undertaken related with driving parameters by the opening relationships formula with catheter design parameter, technological parameter.Conduit and mould all adopt the shell163 unit, and counting along the thickness product branch is 5, employing can reasonable dismissal the full integration Belytschko-Tsay shell unit algorithm of distortion.Through data analysis and experimental verification, show that the finite element model of being set up can accurately reflect the Main physical rule in the conduit plastic forming process, can be according to the variation and the generation of defects of the variation prediction forming quality of processing technology environment, can reflect stress, strain variation and relation thereof in the forming process quantitatively, qualitative analysis is accurate, the believable standard of quantitative test through reaching behind the deviation compensation.
In the method for the embodiment of the invention, the catheter design parameter typically refers to: after pipe-line layout or tubing structural design are finished, be handed down to the parameter of bend pipe department, comprise: conduit outside diameter d, wall thickness t0, radius of curvature R CL, angle of bend θ, and the material parameter of conduit.In addition, relative bending radius Dr=RCL/d and the wall thickness factor t r=d/t0 that derives by the absolute dimension of conduit.The selection of the definite and mould of conduit process parameters range is a foundation with the design parameter of conduit all with design.
Different with numerical control cutting processing, numerical controlled bending of pipe processing technology factor is many and complicated, has reciprocal effect between the factor.Present embodiment by the in-depth analysis to conduit compacting mechanism and rule, is divided into conduit technological parameter and mold structure parameter with the technological factor parameter in the numerical controlled bending of pipe in conjunction with finite element analogy and experiment, puts the relation between the technological factor in order.
In the method for the embodiment of the invention, conduit technological parameter in the bent tube technique factor is meant: determine fully and finish according to operation instruction under the prerequisite of initialization Installation and Debugging at catheter blank, lathe, mould etc., by changing variablees such as lathe setting or mould pose, reach the purpose of effective control forming quality, these variablees are the technological parameter of conduit Plastic Forming, comprising: pressure molding pressure, boosting power, mandrel overhang, mandrel gap, rate of bending, andfrictional conditions, clamping die are nipped, wrinkle resistant mould inclination angle.Mold structure parameter is meant: influential to the conduit forming quality, the mould key dimension parameter that can change according to the design parameter of conduit when mould designs mainly comprises the mandrel structural parameters, pressure mould length and clamping die length; Wherein, the mandrel structural parameters comprise again: mandrel gap, mandrel ball number, mandrel ball thickness.
Numerical control bending forming quality of conduit is estimated by evaluation index.Quality evaluation index is an index of weighing the conduit shaping back mass defect order of severity, be called for short quality index, comprise: spring back angle, elongation, elliptical distortion rate, wall thickness reduction, wall thickness thicken rate and wrinkling value, the above two determine the geometric accuracy and the cutting length of conduit, and the back determines the usability and the outward appearance of conduit.
Spring back angle Δ θ
In the guiding-tube bend process, the distortion of tubing includes elastic component.After process finishing, the suffered constraint of tubing is removed, and elastic deformation partly certainly leads to recovery, makes the angle of shaping angle less than the bending die rotation, and formative radius is greater than bending radius.Resilience has directly changed the final shaping form of conduit, to being equipped with very big influence.Usually weigh the degree of conduit resilience with spring back angle and resilience radius, wherein spring back angle is the difference of mould rotation angle and the actual forming angle of conduit, and the resilience radius is the formative radius of unloading rear tube.
Elongation δ
In the guiding-tube bend deformation process, neutral line is translation to the inside, and after the process finishing, the length of catheter center's line increases.The elongation of correct prediction conduit can improve the precision of blanking, to the very big meaning that is processed with of precious metal material.
Elliptical distortion rate Ω
For the guiding-tube bend distortion, the tubing tension power effect of the neutral line outside, the inboard of making a concerted effort to point to of pulling force.The effect of being stressed of the tubing of neutral line inboard, the directed outside of making a concerted effort.Conduit is under the effect of these power, and shape of cross section changes, and forms the shape of sub-elliptical, is called the cross section elliptical distortion, weighs the size of cross section elliptical distortion degree usually with the elliptical distortion rate.
Wall thickness reduction η h with thicken rate η k
In BENDING PROCESS, the tension power effect of the conduit neutral line outside, outer wall attenuate and interior wall thickening can take place in the inboard effect of being stressed inevitably.Because catheter interior often need be passed through fluid, the tube wall attenuation will directly have influence on load performance, wall thickness thickens and reflects wrinkling possibility size to a certain extent, therefore wall thickness change is a very important defective of catheter fabrication, usually tube wall reduction and wall thickness is thickened rate as the index of estimating this defective.
Wrinkling value Δ h
In the guiding-tube bend process, when inside pressure surpassed the material ability to bear, inner-wall material flocked together, and takes place wrinkling.Wrinkling meeting causes the eddy current of tube fluid, causes the vibration of pipe fitting, reduces the serviceable life of conduit.Usually with the index of wrinkling value as the wrinkling degree of evaluation.
In the realization of the embodiment of the invention, utilize LS-PREPOST to carry out aftertreatment as the poster processing soft of finite element analogy, this software is except that wall thickness reduction, macroscopical this software of forming quality evaluation index for conduit can not directly extract, need the needed interdependent node data of output earlier, and data are analyzed the output quality desired value.Particularly, in the embodiments of the invention, because the result that LS-DYNA analyzes is stored in the d3plot binary file, can utilize this document to carry out the output of interdependent node.For this document, information is with the storage of the form of node, therefore will obtain qualitative datas such as resilience that conduit is shaped, wall thickness change, must be at first the data of respective nodes be derived.Powerful macroefficiency is provided in LS-PREPOST, all corresponding macros of each step operation of LS-PREPOST, carry out these macros and have identical effect with direct control LS-PREPOST, the derivation of node data can rely on these macros of execution to finish.In SCRIPTO, the function of special execution macros is arranged: void ExecuteCommand (Char*cmd), wherein cmd is the character string of storage macros.The information that derives is the coordinate of each node, by carrying out macros, can also derive the data such as stress, strain and thickness of shell elements information of each node in addition.These data are handled accordingly, just can be obtained qualitative datas such as resilience that conduit is shaped, wall thickness change, cross section ellipticity.Certainly, when obtaining quality index, also can use the existing software of ready-made energy output quality index and realize.
For the guiding-tube bend forming technology process of finite element analogy, can verify by experiment.Exemplarily, can adopt the VB300HP CNC tube bending machine of U.S. Eaton to carry out the checking that effective guiding-tube bend is tested.
To the research of conduit numerical control bending forming mechanism, promptly be the research that the factor that influences the conduit forming quality is influenced mechanism and rule to quality index.Principal element and secondary cause are separated consideration, find out the contradictory relation between the quality index, help the essence of seeing clearly that conduit is gone for a stroll and is shaped.
On the basis of finite element analogy, use orthogonal design method and organize a large amount of finite element numerical simulation experiments, determine the technological factor that forming quality is had a significant effect according to test findings, and its shaping law is studied.And then, to influence significant technological parameter and conduit parameter, think that artificial neural network provides the initial training sample as design factor tissue simulation orthogonal test.In the concrete test, exemplarily, can design single factor experiment earlier, analyzing each technological factor influences rule and size thereof to forming quality; And then, can be at one group of conduit geometric parameter of determining, design simulation orthogonal test.In the embodiments of the invention, determine pressing mold pressure Fp, boosting power Fb, mandrel overhang e and mandrel clearance G ap, for numerical control bending forming quality of conduit being influenced significant technological parameter, and influence the rule complexity, it is the contradictory factor of control forming quality, and the clamping die distance of nipping waits other parameter very little or dull relevant to the influence of forming quality, and at most only a quality index is had a significant effect, and can relatively easily determine; Conduit external diameter, wall thickness, bending radius and angle of bend are for to influence significant catheter design parameter to numerical control bending forming quality of conduit.
With described determine numerical control bending forming quality of conduit is influenced significant technological parameter and catheter design parameter as input parameter, predetermined quality index is set up artificial nerve network model as radially basic (RBF) artificial nerve network model as output parameter, and numerical control bending forming quality of conduit is influenced significant technological parameter and catheter design parameter as design factor with above-mentioned, be organized in the bending forming experiment on the CNC tube bending machine or utilize the finite element model of setting up to carry out numerical experiments, obtain quality index, with experiment or test figure initial training sample, in order to artificial nerve network model is trained as the artificial nerve network model of setting up.
Exemplarily, utilize the neural network algorithm among the MATLAB to set up the RBF artificial nerve network model.Because characteristic is approached in the singularity of numerical controlled bending of pipe and the part of RBF network, in order to improve precision, the embodiments of the invention segmentation is set up a plurality of neural network models to adapt to the wide variation of tube material and conduit geometric parameter such as conduit external diameter.Exemplarily, can set up different neural network models to the conduit of different materials; As, the neural network model difference of the conduit correspondence of different materials such as stainless steel, aldary, aluminium alloy, titanium alloy.Further,, have and further to have divided into mandrel and no mandrel, have mandrel different with the neural network model of centreless spindle guide pipe correspondence for the conduit of same material.For same material the mandrel conduit arranged, set up different neural network models according to the difference of conduit external diameter.Exemplarily, can be divided into different sections according to the size of outer diameter D, as: D20～30 section, D30～40 section, D40～50, D50～60, D60～70 etc., the initial training sample that the corresponding neural network model of different sections adopts when training is different, be respectively that conduit according to the correspondingly-sized scope experimentizes or tests acquisition, the difference of initial training sample causes the neural network model difference that trains, be used to carry out prediction of quality.
With the RBF neural network in stainless-steel tube D20～30 section is example, as shown in table 1 below with an example of the training sample of orthogonal method tissue:
Table 1
In the table 1, the mm representation unit is a millimeter; The KN representation unit is thousand Ns.
After obtaining the training sample data, sample data is carried out normalized, make all DATA DISTRIBUTION between 0 to 1, so that the identification of neural network.Data after the normalization can be trained neural network, and parameters such as the hidden layer number by adjusting the RBF neural network, training speed are to improve precision of prediction.
Artificial neural network is to be interconnected by a large amount of simple primary element neurons, and the mode of the cerebral nerve process information by the anthropomorphic dummy is carried out the complex networks system of information parallel processing and non-linear conversion.Because neural network has powerful learning functionality, can realize the Nonlinear Mapping process by low weight loose ground, and have the ability of large-scale calculations.Radially base (RBF) network is to be a class feedforward network of base configuration with the approximation of function theory.Because the characteristic of local acknowledgement, it is optimum to approaching of function, can approach arbitrary function with arbitrary accuracy, and training process is short, precision of prediction and initial weight are irrelevant, thereby in fields such as hypersurface match, free form surface reconstruct and main equipment fault diagnosises many application are arranged.The inventor finds in realizing process of the present invention: numerical controlled bending of pipe technological parameter and conduit parameter influence rule and can describe by hypersurface forming quality, most shaping laws of conduit are linearity or simple nonlinear, the second derivative of law curve also is the partial derivative reversion not of hypersurface, extreme point and flex point are few, so just having guaranteed that the part is approached almost is equal to the overall situation and approaches, thereby the RBF network is particularly suitable for solving the forecasting problem of numerical controlled bending of pipe forming quality.SPREAD is the distribution density of radial basis function, and the SPREAD value is big more, and function is level and smooth more, and the SPREAD value is more little, and is just accurate more to approaching of function, but the process of approaching is just unsmooth more.Therefore when planned network, should adjust the value of several spread, up to reaching reasonable precision more.The SPREAD value is the parameter that the network precision is had the greatest impact, at first value (0.5～3.0) is carried out tentative calculation on a large scale, contrast its error, and then segment (0.6～1.4) as central point, can determine that the secondary segmentation optimal value of SPREAD is 0.7 with optimal value (1.0) in this group.
The RBF neural network that trains based on previous step, as input, quality index is as output with the conduit parameter after the normalization and technological parameter, predicted value carried out obtaining after the anti-normalization predicted value of quality index.Exemplarily, when input vector is: D=20mm, t=1mm, R=30mm, α=45.55 °, pressing mold pressure=8kN, pressing mold boosting power=2.5kN, when mandrel gap=0.8mm, mandrel overhang=3.57mm, utilize the quality index that the neural network that trains obtains and utilize quality index value that finite element model obtains, and both compare errors as shown in table 2 below:
| Wall thickness reduction | Wall thickness thickens rate | The elliptical distortion rate | |
| Neural network | 12.4992 | 18.6464 | 2.6081 |
| Finite element | 11.22 | 19.04 | 3.98 |
| Compare error | 11.4％ | 2.07％ | 34.5％ |
Table 2
Guiding-tube bend qualitative forecasting method based on artificial neural network, overcome the defective of poor in timeliness of the finite element analogy Forecasting Methodology of prior art, make full use of existing empirical data, reduced the time of prediction of quality, but it still can have preferable precision of prediction.
The guiding-tube bend qualitative forecasting method of the embodiment of the invention: selected the RBF neural network model for use, improved accuracy of predicting; Plastic Forming mechanism according to conduit, four the catheter design parameters of input parameter for the conduit forming quality is had the greatest impact of neural network model have been determined: tangible four technological parameters of conduit external diameter, wall thickness, bending radius and angle of bend and influence: pressing mold pressure, boosting power, mandrel overhang and mandrel gap, and ignore of the influence of other secondary cause to the conduit forming quality, under the prerequisite that guarantees precision of prediction, simplified model; The sample data of numerical stability, reflection conduit shaping law feature is provided for neural network, initial sample data is with the format organization of orthogonal test, guarantee the gamut that limited training sample covering process parameter changes, and reflected the shaping law feature of conduit with higher precision.
The numerical control bending forming quality of conduit neural network prediction model is effectively replenishing of finite element model, when process planning, utilizes method of the present invention can in seconds provide the predicted value of conduit forming quality.Certainly, because neural network prediction model only can reflect the relation between the data, thereby when taking place, serious mass defect need find out the reason that defective takes place by finite element analogy.
The embodiment of the invention also discloses a kind of prediction unit of numerical control bending forming quality of conduit, comprising: finite element model is set up module, is used to set up the finite element model in order to simulate catheter numerical control bending forming technological process; The parameter determination module, be used to utilize described finite element model, the finite element numerical simulation test of tissue tract numerical control bending forming obtains first test findings, and determining according to described first test findings influences significant technological parameter and catheter design parameter to numerical control bending forming quality of conduit; Neural network module, be used for described determine numerical control bending forming quality of conduit is influenced significant technological parameter and the catheter design parameter is set up artificial nerve network model as input parameter, predetermined quality index as output parameter; Prediction module is used for described artificial nerve network model is trained, and utilizes described housebroken artificial nerve network model to carry out the prediction of numerical control bending forming quality of conduit.
Preferably, described prediction unit, wherein, in the described prediction module, the initial training sample that is used for described artificial nerve network model is trained is:
By with described numerical control bending forming quality of conduit influenced significant technological parameter and catheter design parameter obtaining of determining as the experiment of the conduit numerical control bending forming of design factor, tissue or the numerical experiments of finite element.
Preferably, in the described prediction unit, based on the orthogonal design method tissue during numerical experiments of described conduit numerical control bending forming finite element.
Preferably, in the described prediction unit, the described artificial nerve network model that described neural network module is set up is: radially basic artificial nerve network model.
Preferably, described prediction unit, wherein, described neural network module is further used for setting up a plurality of artificial nerve network models according to the different segmentations of the geometric parameter of the difference of tube material and/or conduit.
Preferably, described prediction unit, wherein, described parameter determination module is determined describedly influences significant technological parameter to numerical control bending forming quality of conduit and comprises: pressing mold pressure, boosting power, mandrel overhang and mandrel gap; Describedly numerical control bending forming quality of conduit is influenced significant catheter design parameter comprise: conduit external diameter, wall thickness, bending radius and angle of bend.
The above only is a preferred implementation of the present invention; should be pointed out that for those skilled in the art, under the prerequisite that does not break away from the principle of the invention; can also make some improvements and modifications, these improvements and modifications also should be considered as protection scope of the present invention.
Claims (14)
1. the Forecasting Methodology of a numerical control bending forming quality of conduit is characterized in that, comprises the steps:
Step 1 is set up the finite element model in order to simulate catheter numerical control bending forming technological process;
Step 2, utilize described finite element model, the finite element numerical simulation test of tissue tract numerical control bending forming obtains first test findings, and determining according to described first test findings influences significant technological parameter and catheter design parameter to numerical control bending forming quality of conduit;
Step 3, with described determine numerical control bending forming quality of conduit is influenced significant technological parameter and catheter design parameter as input parameter, as output parameter, set up artificial nerve network model with predetermined quality index;
Step 4 is trained described artificial nerve network model, and utilizes housebroken artificial nerve network model to carry out the prediction of numerical control bending forming quality of conduit.
2. method according to claim 1 is characterized in that, before the described step 4, also comprises:
With described determine numerical control bending forming quality of conduit is influenced significant technological parameter and catheter design parameter as design factor, the experiment of tissue tract numerical control bending forming or the numerical experiments of finite element think that described artificial nerve network model provides the initial training sample.
3. method according to claim 1 and 2 is characterized in that, organizes the numerical experiments of described conduit numerical control bending forming finite element based on orthogonal design method.
4. method according to claim 1 is characterized in that, in the described step 3, described artificial nerve network model is: radially basic artificial nerve network model.
5. method according to claim 2 is characterized in that, in the described step 4,
Before described artificial nerve network model is trained, also comprise: described initial training sample data is carried out normalized;
The described step of utilizing described housebroken artificial nerve network model to carry out the prediction of numerical control bending forming quality of conduit comprises: before the input parameter value is imported described artificial nerve network model, carry out normalized earlier; And, after obtaining the output valve of described artificial nerve network model, described output valve is carried out obtaining after the anti-normalized quality index value of the numerical control bending forming of described conduit.
6. method according to claim 1 and 2 is characterized in that, in the described step 3, the described step of setting up artificial nerve network model comprises:
A plurality of artificial nerve network models are set up in different segmentations according to the geometric parameter of the difference of tube material and/or conduit.
7. method according to claim 6 is characterized in that, the geometric parameter of described conduit comprises: the conduit external diameter.
8. method according to claim 1 and 2 is characterized in that, describedly numerical control bending forming quality of conduit is influenced significant technological parameter comprises: pressing mold pressure, boosting power, mandrel overhang and mandrel gap; Describedly numerical control bending forming quality of conduit is influenced significant catheter design parameter comprise: conduit external diameter, wall thickness, bending radius and angle of bend.
9. the prediction unit of a numerical control bending forming quality of conduit is characterized in that, comprising:
Finite element model is set up module, is used to set up the finite element model in order to simulate catheter numerical control bending forming technological process;
The parameter determination module, be used to utilize described finite element model, the finite element numerical simulation test of tissue tract numerical control bending forming obtains first test findings, and determining according to described first test findings influences significant technological parameter and catheter design parameter to numerical control bending forming quality of conduit;
Neural network module, be used for described determine numerical control bending forming quality of conduit is influenced significant technological parameter and catheter design parameter as input parameter, as output parameter, set up artificial nerve network model with predetermined quality index;
Prediction module is used for described artificial nerve network model is trained, and utilizes housebroken artificial nerve network model to carry out the prediction of numerical control bending forming quality of conduit.
10. prediction unit according to claim 9 is characterized in that, in the described prediction module, the initial training sample that is used for described artificial nerve network model is trained is:
By with described numerical control bending forming quality of conduit influenced significant technological parameter and catheter design parameter obtaining of determining as the experiment of the conduit numerical control bending forming of design factor, tissue or the numerical experiments of finite element.
11. according to claim 9 or 10 described prediction units, it is characterized in that, based on the orthogonal design method tissue during numerical experiments of described conduit numerical control bending forming finite element.
12. prediction unit according to claim 9 is characterized in that, the described artificial nerve network model that described neural network module is set up is: radially basic artificial nerve network model.
13., it is characterized in that described neural network module is further used for setting up a plurality of artificial nerve network models according to the different segmentations of the geometric parameter of the difference of tube material and/or conduit according to claim 9 or 10 described prediction units.
14. according to claim 9 or 10 described prediction units, it is characterized in that described parameter determination module is determined describedly influences significant technological parameter to numerical control bending forming quality of conduit and comprise: pressing mold pressure, boosting power, mandrel overhang and mandrel gap; Describedly numerical control bending forming quality of conduit is influenced significant catheter design parameter comprise: conduit external diameter, wall thickness, bending radius and angle of bend.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN 201010205834 CN101901283A (en) | 2010-06-22 | 2010-06-22 | Prediction method of numerical control bending forming quality of conduit and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN 201010205834 CN101901283A (en) | 2010-06-22 | 2010-06-22 | Prediction method of numerical control bending forming quality of conduit and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN101901283A true CN101901283A (en) | 2010-12-01 |
Family
ID=43226816
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN 201010205834 Pending CN101901283A (en) | 2010-06-22 | 2010-06-22 | Prediction method of numerical control bending forming quality of conduit and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN101901283A (en) |
Cited By (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102567577A (en) * | 2011-12-16 | 2012-07-11 | 沈阳飞机工业（集团）有限公司 | Rapid die compensation method considering rebound of bent part |
| CN104614156A (en) * | 2013-11-05 | 2015-05-13 | 珠海格力电器股份有限公司 | Method for detecting sealing clamp performance |
| CN103514325B (en) * | 2013-09-18 | 2017-07-21 | 华侨大学 | Numerical simulation method of the spinning roller of spoke three mistake away from shear spinning technique |
| CN108956322A (en) * | 2018-04-25 | 2018-12-07 | 成都飞机工业（集团）有限责任公司 | A method of for testing S-shaped material for test bending property parameter |
| CN109153186A (en) * | 2016-05-12 | 2019-01-04 | 惠普发展公司，有限责任合伙企业 | Predict the quality of 3D object part |
| CN109684753A (en) * | 2018-12-28 | 2019-04-26 | 西北工业大学 | A kind of bending pipes springback angle backward-predicted and compensation method |
| CN109732450A (en) * | 2019-02-27 | 2019-05-10 | 重庆理工大学 | A kind of blade polishing processing method neural network based |
| CN110688793A (en) * | 2019-09-23 | 2020-01-14 | 成都飞机工业（集团）有限责任公司 | Secondary shape correction quality checking method during assembly of aviation elbow based on finite element |
| CN112632810A (en) * | 2020-11-30 | 2021-04-09 | 江苏科技大学 | Method for predicting pressing amount rule of upper die for bending rod piece |
| CN113011055A (en) * | 2021-02-03 | 2021-06-22 | 中国船级社 | Motor high-temperature fault simulation method and system based on finite element technology |
| CN114558918A (en) * | 2022-03-25 | 2022-05-31 | 昌河飞机工业（集团）有限责任公司 | Bending forming method for large-pipe-diameter thin-walled pipe without transition in middle |
| CN114863063A (en) * | 2022-07-07 | 2022-08-05 | 南京智欧智能技术研究院有限公司 | Springback prediction method for forming single-point gradual change surface topography |
| US11524461B2 (en) | 2016-05-12 | 2022-12-13 | Hewlett-Packard Development Company, L.P. | Data units for additive manufacturing |
| CN116910944A (en) * | 2023-09-13 | 2023-10-20 | 中基科技(武汉)有限公司 | Technological parameter determining method and device of marine profile straightening machine and electronic equipment |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101226606A (en) * | 2007-11-16 | 2008-07-23 | 燕山大学 | Method for optimization of socket-shaped part machinery expanding technological parameter |
-
2010
- 2010-06-22 CN CN 201010205834 patent/CN101901283A/en active Pending
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101226606A (en) * | 2007-11-16 | 2008-07-23 | 燕山大学 | Method for optimization of socket-shaped part machinery expanding technological parameter |
Non-Patent Citations (2)
| Title |
|---|
| 《塑性工程学报》 20091231 刘婧瑶等 "管材数控绕弯回弹实验研究及BP网络预测模型" 第4页到第5页 1-14 第6卷, 第16期 2 * |
| 《机械设计与研究》 20081031 刘婧瑶等 "薄壁管数控弯曲成形中芯轴参数的确定" 第2页到第3页 1-14 第24卷, 第5期 2 * |
Cited By (22)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102567577A (en) * | 2011-12-16 | 2012-07-11 | 沈阳飞机工业（集团）有限公司 | Rapid die compensation method considering rebound of bent part |
| CN102567577B (en) * | 2011-12-16 | 2014-02-26 | 沈阳飞机工业（集团）有限公司 | Rapid die compensation method considering rebound of bent part |
| CN103514325B (en) * | 2013-09-18 | 2017-07-21 | 华侨大学 | Numerical simulation method of the spinning roller of spoke three mistake away from shear spinning technique |
| CN104614156A (en) * | 2013-11-05 | 2015-05-13 | 珠海格力电器股份有限公司 | Method for detecting sealing clamp performance |
| CN104614156B (en) * | 2013-11-05 | 2017-08-29 | 珠海格力电器股份有限公司 | Method for detecting sealing clamp performance |
| US11524461B2 (en) | 2016-05-12 | 2022-12-13 | Hewlett-Packard Development Company, L.P. | Data units for additive manufacturing |
| CN109153186A (en) * | 2016-05-12 | 2019-01-04 | 惠普发展公司，有限责任合伙企业 | Predict the quality of 3D object part |
| US11364685B2 (en) | 2016-05-12 | 2022-06-21 | Hewlett-Packard Development Company, L.P. | Predicting quality of a 3D object part |
| CN108956322B (en) * | 2018-04-25 | 2020-12-01 | 成都飞机工业（集团）有限责任公司 | Method for testing bending performance parameters of S-shaped test piece material |
| CN108956322A (en) * | 2018-04-25 | 2018-12-07 | 成都飞机工业（集团）有限责任公司 | A method of for testing S-shaped material for test bending property parameter |
| CN109684753A (en) * | 2018-12-28 | 2019-04-26 | 西北工业大学 | A kind of bending pipes springback angle backward-predicted and compensation method |
| CN109732450A (en) * | 2019-02-27 | 2019-05-10 | 重庆理工大学 | A kind of blade polishing processing method neural network based |
| CN109732450B (en) * | 2019-02-27 | 2021-02-26 | 重庆理工大学 | Blade polishing and grinding method based on neural network |
| CN110688793B (en) * | 2019-09-23 | 2022-04-22 | 成都飞机工业（集团）有限责任公司 | Secondary shape correction quality checking method during assembly of aviation elbow based on finite element |
| CN110688793A (en) * | 2019-09-23 | 2020-01-14 | 成都飞机工业（集团）有限责任公司 | Secondary shape correction quality checking method during assembly of aviation elbow based on finite element |
| CN112632810A (en) * | 2020-11-30 | 2021-04-09 | 江苏科技大学 | Method for predicting pressing amount rule of upper die for bending rod piece |
| CN113011055A (en) * | 2021-02-03 | 2021-06-22 | 中国船级社 | Motor high-temperature fault simulation method and system based on finite element technology |
| CN114558918A (en) * | 2022-03-25 | 2022-05-31 | 昌河飞机工业（集团）有限责任公司 | Bending forming method for large-pipe-diameter thin-walled pipe without transition in middle |
| CN114863063B (en) * | 2022-07-07 | 2022-10-11 | 南京智欧智能技术研究院有限公司 | Springback prediction method for single-point gradual change surface topography forming |
| CN114863063A (en) * | 2022-07-07 | 2022-08-05 | 南京智欧智能技术研究院有限公司 | Springback prediction method for forming single-point gradual change surface topography |
| CN116910944A (en) * | 2023-09-13 | 2023-10-20 | 中基科技(武汉)有限公司 | Technological parameter determining method and device of marine profile straightening machine and electronic equipment |
| CN116910944B (en) * | 2023-09-13 | 2023-12-26 | 中基科技(武汉)有限公司 | Technological parameter determining method and device of marine profile straightening machine and electronic equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN101901283A (en) | Prediction method of numerical control bending forming quality of conduit and device | |
| Pilani et al. | A hybrid intelligent systems approach for die design in sheet metal forming | |
| Zhu et al. | Springback prediction for rotary-draw bending of rectangular H96 tube based on isotropic, mixed and Yoshida–Uemori two-surface hardening models | |
| Lee et al. | Shape optimization of the workpiece in the forging process using equivalent static loads | |
| Praveen Kumar et al. | Influence of forming parameters on the crash performance of capped cylindrical tubes using LS-DYNA follow-on simulations | |
| CN104307939A (en) | Springback compensation method of thin-wall stamped part with large relative bending radius | |
| Lipinski et al. | System for monitoring and optimization of micro-and nano-machining processes using intelligent voice and visual communication | |
| CN105844025A (en) | Non-probability aerothermoelasticity reliability designing method for hypersonic control surface | |
| CN103366056A (en) | Machine part cylindricity error measuring method based on finite element analysis | |
| Zhang et al. | A hierarchical prediction method based on hybrid-kernel GWO-SVM for metal tube bending wrinkling detection | |
| JP2003340529A (en) | Analysis system for springback of press-formed product | |
| Wang et al. | Influence of thickness reduction on the forming results in the three-dimensional surface rolling process with rigid arc-shaped rollers | |
| Zhang et al. | A new method for automatic optimization of drawbead geometry in the sheet metal forming process based on an iterative learning control model | |
| Kim et al. | A Study on Improving Formability of Stamping Processes with Segmented Blank Holders using Artificial Neural Network and Genetic Algorithm | |
| Chen | An analysis of forming limit in the elliptic hole-flanging process of sheet metal | |
| Zhao et al. | Springback theory of plane bending and the progress of study on its engineering application | |
| CN104820729A (en) | Temperature succession method of whole rolling process simulation analysis of steel rail | |
| Wang et al. | A multi-state fusion informer integrating transfer learning for metal tube bending early wrinkling prediction | |
| CN111209694B (en) | Structure identification method for rigidity and axial force of truss structure | |
| Adrian et al. | Curating Datasets of Flexible Assemblies to Predict Spring-Back Behavior for Machine Learning Purposes | |
| Yeh et al. | Application of adaptive network fuzzy inference system to die shape optimal design in sheet metal bending process | |
| Zhang | Design for uncertainties of sheet metal forming process | |
| Kim et al. | Computational approach to analysis and design of hydroforming process for an automobile lower arm | |
| Chebbah et al. | A fast algorithm for strain prediction in tube hydroforming based on one-step inverse approach | |
| Strano et al. | Hierarchical metamodeling of the air bending process |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C02 | Deemed withdrawal of patent application after publication (patent law 2001) | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Open date: 20101201 |