CN105760571A - Electric-heating variable mold temperature injection mold heating system design method - Google Patents

Abstract

The invention discloses an electric-heating variable mold temperature injection mold heating system design method based on BP-NSGA2 multi-target optimization and TOPSIS multi-attribute decision-making.The method comprises the steps of 1, establishing a heat response analysis model for the mold heating phase, and determining design variables and optimization targets; 2, adopting central composite testing design to obtain a testing sample; 3, conducting training on the sample, and establishing a BP neural network model related to heating time and the surface temperature difference of a heated cavity; 4, optimizing the BP neural network model through an NSGA2 multi-target genetic algorithm, so that a Pareto optimal solution set is obtained; 5, determining the objective weight of each optimization target through an entropy-based weight method; 6, ranking the Pareto optimal solution set in priority through a TOPSIS strategy to obtain an optimal heating scheme.By means of the electric-heating variable mold temperature injection mold heating system design method, heating efficiency of a mold is effectively improved, temperature evenness of the surface of the cavity is improved, development efficiency of the mold is improved, development cost is lowered, and the development cycle is shortened.

Description

Electric heating becomes mould temperature injection mold heating system design method

Technical field

The present invention relates to become the heating system optimization method field of mould temperature injection mold, particularly relate to a kind of based on BP- NSGA2 multiple-objection optimization becomes mould temperature injection mold heating system optimization method with the electric heating of TOPSIS multiple attribute decision making (MADM).

Background technology

Traditional injection moulding goods are increasingly difficult to meet people more to be come aspects such as injection-molded item mechanical property, presentation qualities The highest requirement.Quickly becoming mould temperature injection mo(u)lding (Rapid Heat Cycle Molding, RHCM) technology is in these years to send out A kind of new injection molding technology that exhibition is got up, this technology can eliminate the defects such as the weld mark of product surface, current mark, floating fibre, promote Products surface quality.RHCM moulding process is with the difference of traditional injection moulding moulding process maximum, fills out at polymer melt The temperature filling front mold mold cavity surface is heated rapidly on injection-molded item material glass transition temperature.The heating of mould After efficiency and heating, the temperature distribution evenness of cavity surface is two important technology indexs in RHCM moulding process, Qian Zheying Ringing and become mould temperature injection production efficiency, the latter affects change mould temperature injection-molded item quality.

At present, RHCM injection mold heating system design uses trial-and-error method adjusted design parameter mostly, passes through finite element fraction Analysis obtains and reaches the heat time needed for design temperature and the Temperature Distribution of mold cavity surface.But, meet rapid and uniform heating The optimal design parameters of mold cavity surface is not readily available, and this trial and error design method based on simulation has great limitation Property.Based on this, scholar both domestic and external has carried out some researchs to RHCM injection mold heating system optimization design.Document 1 (Xiao C L,Huang H X.Multiobjective optimization design of heating system in electric heating rapid thermal cycling mold for yielding high gloss parts[J] .Journal of Applied Polymer Science, 2014,131 (6): 596-602. i.e. Xiao Chenglong, Huang Hanxiong. Gao Guang Seamless electric heating becomes mould temperature injection mold heating system multiple-objection optimization [J]. journal of applied, and 2014,131 (6): 596-602.) with the spatial parameter of heating element heater as design variable, required heat time and the highest die cavity surface temperature are mesh Scalar functions, RHCM injection mold heating system has been carried out excellent by binding tests design, finite element analysis, response surface agent model Change.Document 2 (Gong Ningning, Li Xiping. high-gloss injection mold heating layout of beam line optimization design [J]. Tool and Die Technology, 2012:20- 23.) by integrated finite element simulation and intelligent algorithm, the mathematical optimization models of mould heating/cooling pipe is established, real Show the optimization design of RHCM injection mold heating/cooling pipe layout parameter.Document 3 (Wang GL, Zhao GQ, Li HP, Guan YJ.Research of thermal response simulation and mold structure optimization for rapid heat cycle molding processes,respectively,with steam The heating and i.e. Wang Guilong of electric heating [J] .Materials&Design, 2010,31:382-395., Zhao State group, Li Huiping, Guan Yanjin. steam heating and electrical heating become mould temperature injection mold thermal response research and Optimizing die structure [J]. material and design, 2010,31:382-395.) and document 4 (Wang G, Zhao G, Li H, Guan Y.Research on optimization design of the heating/cooling channels for rapid heat cycle molding based on response surface methodology and constrained particle swarm Optimization [J] .Expert Syst Appl, 2011,38:6705-6719. i.e. Wang Guilong, State of Zhao group, Li Huiping, pipe Yan Jin. rapid mould temperature variable injection mould heating/cooling pipe optimized design based on Response Surface Method and particle cluster algorithm is ground Study carefully [J]. expert system and application thereof, 2011,38:6705-6719.) combine FInite Element, response surface agent model and multiple target Particle cluster algorithm carries out multiple-objection optimization, thus realizes the optimization design of RHCM injection mold heating system, it is thus achieved that feasible adds The thermal efficiency and uniform mold cavity surface temperature.The optimum pareto disaggregation of the acquisition in document 1 finally needs manually to carry out subjectivity Judge, also exist subjective random.Optimization Design in document 2, the layout of beam line design variable part of heating is set as Definite value and Optimized Iterative process are complicated, and short time consumption is long.The employing second order polynomial model set up in document 3 and document 4 is set up Accurate Mathematical Modeling can not well reflect the relation between design parameter and optimization aim, is confined to its search volume.? In injection molding process, become mould temperature (Electrical Rapid Heat Cycle as a kind of widely used electric heating Molding, ERHCM) injection mold, it is necessary to form a kind of perfect heating system design method, that improves mould adds thermal effect Rate, improves the temperature homogeneity of mold cavity surface, improves the development efficiency of mould, shortens development cost and construction cycle.

Summary of the invention

In order to the efficiency of heating surface temperature relatively low, mold cavity surface overcoming existing electric heating change mould temperature injection mold mode of heating is equal The deficiency that even property is poor, development efficiency is relatively low, the present invention provides one based on BP-NSGA2 multiple-objection optimization and many attributes The electric heating of TOPSIS decision-making becomes mould temperature injection mold heating system optimization method, it is achieved electrical bar specification and spatial configuration optimal, Meet the efficiency of heating surface and the mold cavity surface temperature homogeneity requirement of mould, use trial-and-error method to adjust during making up traditional design and set The weak point of meter parameter, improves development efficiency, shortens development cost and construction cycle.

In order to solve above technical problem, the present invention provides following technical scheme:

A kind of electric heating based on BP-NSGA2 multiple-objection optimization with TOPSIS multiple attribute decision making (MADM) becomes the heating of mould temperature injection mold System optimization method, these heating means comprise the following steps:

Step S100, sets up the thermal response of mould heating period and analyzes model, wherein with specification and the space layout of electrical bar It is optimization aim for design variable, the efficiency of heating surface and mold cavity surface temperature homogeneity；

Step S200, uses Central Composite experimental design CCD to obtain test sample；

Step S300, is trained sample, and sets up the BP nerve net of die cavity surface temperature difference after heat time and heating Network model；

Step S400, uses NSGA2 multi-objective genetic algorithm to be optimized BP neural network model, it is thus achieved that Pareto is Excellent disaggregation；

Step S500, uses the entropy method of weighting to determine the objective weight of each optimization aim；

Step S600, carries out priority ordered based on TOPSIS strategy to Pareto optimal solution set and obtains optimal heating side Case.

Further, in described step S100, use 4 parameters to be described, be that the diameter of electrical bar, power are close respectively Degree, electrical bar wall are away from the distance between mold cavity surface distance and electrical bar wall.

Further, in described step S200, use the Central Composite experimental design CCD planning sample of four factor five levels This.

Further, in described step S300, BP Establishment of Neural Model step is as follows:

Step S301, determines training sample, with the diameter of electrical bar, power density, electrical bar wall away from mold cavity surface away from From and electrical bar wall between four design parameters of distance be network input, with the heat time and heating after mold cavity surface temperature Difference exports for network；

Step S302, creates BP neutral net and carries out network training；

Step S303, according to mean square error MSE, it is thus achieved that the BP neural network model needed for required multiple-objection optimization.

In described step S400, being optimized BP neural network model by NSGA2 multi-objective genetic algorithm, process is such as Under:

Step S401, determine two object functions that ERHCM injection mold heating system optimization exports, i.e. heat time with Die cavity surface temperature difference after heating；

Step S402, confirms the constrained type of multi-objective optimization question, and this is nonlinear restriction, confirms to optimize design ginseng The bound of number；

Step S403, initializes population M, and one size of stochastic generation is the parent population P of N_t；

Step S404, carries out target function value calculating to current population at individual；

Step S405, carries out non-bad layer sorting to population at individual；

Step S406, uses binary algorithm of tournament selection, intersection and mutation operation to produce N number of progeny population Q_t；

Step S407, population P_tWith population Q_tIt is incorporated into R_tIn, R_t=P_t∪Q_t；

Step S408, to new population R_tMiddle individuality carries out target function value calculating；

Step S409, carries out non-dominated ranking to population at individual；

Step S410, selects top n individuality to produce parent population P_t+1；

Step S411, if reaching the condition of convergence, terminates；Otherwise, algebraically increases by 1, turns step S404 step；

Step S412, exports Pareto optimal solution set.

In described step S500, using the entropy method of weighting to determine the objective weight of each optimization aim, process is as follows:

Pareto optimal solution set data are done standardization by step S501, obtain specified decision:

$y_{ij}=\frac{max\left\{X_j\right\}-x_{ij}}{max\left\{X_j\right\}-min\left\{X_j\right\}},(1\≤i\≤m,1\≤j\≤n),$

Wherein X_ijRepresent the numerical value of i-th optional program jth optimization aim, max{X_jAnd min{X_jAll sides to be selected The maximum of jth item evaluation index and minimum of a value in case, m is optional program number, and n is optimization aim number；

Step S502, processes decision matrix, obtains matrix P=(p_ij)_m×n: $p_{ij}=\frac{y_{ij}}{\underset{i=1}{\overset{m}{\Σ}}{y}_{ij}},$ $(1\≤i\≤m,1\≤j\≤n);$

Step S503, comentropy e of parameter output_j: $e_j=-\frac{1}{lnm}\underset{i=1}{\overset{m}{\Σ}}{p}_{ij}ln{p}_{ij},$ $(1\≤i\≤m,1\≤j\≤n);$

Step S504, computation attribute weight vectors ω_j: ${\mathrm{\ω}}_j=\frac{1-e_j}{\underset{j=1}{\overset{n}{\Σ}}(1-e_j)},\underset{j=1}{\overset{n}{\Σ}}{\mathrm{\ω}}_j=1,$ $(1\≤i\≤m,1\≤j\≤n).$

In described step S600, based on TOPSIS strategy, Pareto optimal solution set is carried out priority ordered and obtain optimal Heating control scheme, process is as follows:

Step S601, makees standardization processing to decision matrix, obtains specified decision matrix Y=(y_ij)_m×n, wherein $y_{ij}=\frac{max\left\{X_j\right\}-x_{ij}}{max\left\{X_j\right\}-min\left\{X_j\right\}},(1\≤i\≤m,1\≤j\≤n);$

Step S602, utilizes step S500 to obtain objective weight ω of optimization aim_j；

Step S603, calculates weighted normal decision matrix Z=(z_ij)_m×n, wherein z_ij=ω_jy_ij,(1≤i≤m,1≤j ≤n)；

Step S604, determines positive ideal solution Z⁺With minus ideal result Z^-: $Z^{-}=(z_1^{-},z_2^{-},...,z_n^{-},)=0,$ Wherein $z_j^{+}=\underset{1\≤i\≤m}{\max }{z}_{ij}={\mathrm{\ω}}_j,z_j^{-}=\underset{1\≤i\≤m}{\min }{z}_{ij}=0;$

Step S605, calculates each scheme Euclid distance to positive ideal solution and minus ideal resultWith

$d_i^{+}=\sqrt{\underset{i=1}{\overset{n}{\Σ}}{(z_{ij}-z_j^{+})}^2},d_i^{-}=\sqrt{\underset{i=1}{\overset{n}{\Σ}}{(z_{ij}-z_j^{-})}^2},1\≤i\≤m;$

Step S605, calculates the relative similarity degree of each scheme

$C_i^{+}=\frac{d_i^{-}}{d_i^{+}+d_i^{-}},1\≤i\≤m;$

Step S606, arranges the precedence of each scheme: relative similarity degree is the biggest the most excellent, relatively according to relative similarity degree Approach degree is the least the most bad.

The technology of the present invention is contemplated that: BP neutral net has the strongest fault-tolerance and processing speed, it may not be necessary to essence True computation model, it is achieved sample input layer and the Nonlinear Mapping of output layer data, finally obtains reliable and precision is higher Input and output forecast model.Nondominated sorting genetic algorithm II (NSGA2) is the many mesh of a kind of effectively solution The method of mark optimization problem, has preferable space optimal solution search capability and convergence faster.TOPSIS method is limited One of integrated evaluating method of Scheme Multiple Attribute Decision Making, mainly determines the most excellent of each scheme by the positive and negative ideal solution of structure First spending, the gap reflected between each scheme of energy objective reality, information loss is few, has wide range of applications.The present invention is based on foundation The ERHCM injection mold heat time and heating after the BP neural network model of die cavity surface temperature difference, excellent in conjunction with NSGA2 multiple target Change method and TOPSIS multiple attribute decision making (MADM) technology obtain the best design of ERHCM injection mold heating system, it is achieved that electricity Hot pin specification and spatial configuration optimal, meet the efficiency of heating surface and the mold cavity surface temperature homogeneity requirement of mould, compensate for passing System design process in use trial-and-error method adjusted design parameter weak point, improve development efficiency, shorten development cost with And the construction cycle.

The present invention provide technical scheme provide the benefit that: the present invention construct ERHCM thermal response analyze model, with in Heart composite testing method (CCD) carries out test planning, obtains test data；With test data as training sample, establish design ginseng The BP neutral net mould of Nonlinear Mapping relation between die cavity surface temperature uniformity objective after number and the efficiency of heating surface and heating Type；Use NSGA2 multi-objective genetic algorithm that BP neural network model is optimized, it is thus achieved that Pareto optimal solution set；Introduce entropy The value method of weighting determines the objective weight of die cavity surface temperature uniformity two optimization aim after the efficiency of heating surface and heating；Weigh based on entropy Weight method, uses TOPSIS multiple attribute decision making (MADM) technology that the Pareto optimal solution set obtained is carried out priority ordered, so that it is determined that Good design.This heating system optimization method, is effectively improved the efficiency of heating surface of mould, improves mold cavity surface Temperature homogeneity, improves the development efficiency of mould, shortens development cost and construction cycle.

Accompanying drawing explanation

Fig. 1 electric heating becomes mould temperature injection mold heating system optimization method flow chart.

Model is analyzed in Fig. 2 ERHCM injection mold heating period thermal response.

Fig. 3 BP neural metwork training target convergence curve；Wherein, abscissa Epochs represents frequency of training, ordinate Mean Squared Error represents that mean square error, curve Train represent that training curve, straight line Best and Goal represent the most respectively Good value and desired value.

Fig. 4 BP Neural Network Data fitted figure；Wherein, abscissa Target represents measurement normalized value, ordinate Output represents prediction normalized value, and Data is data point, and Fit is fitting a straight line, and Training:R represents the coefficient of determination.

The Pareto optimal solution set front end figure that Fig. 5 present invention obtains；Wherein, abscissa Time represents the heat time, vertical seat Mark Δ T represents the mold cavity surface temperature difference.

Detailed description of the invention

Clearer for make the object, technical solutions and advantages of the present invention express, below in conjunction with the accompanying drawings and tool The present invention is described further by body embodiment again.

With reference to Fig. 1～Fig. 5, a kind of electric heating based on BP-NSGA2 multiple-objection optimization with TOPSIS multiple attribute decision making (MADM) becomes mould temperature Injection mold heating system optimization method, comprises the steps:

Step S100, ERHCM injection mold heating period thermal response analyzes model as shown in Figure 2.In Fig. 2, d, q, h, p divide Do not represent that the diameter of electrical bar, power density, electrical bar wall are away from the distance between mold cavity surface distance and electrical bar wall Half, is ERHCM injection mold heating system important design parameter.The efficiency of heating surface can be with heating the heating reaching temperature required Time (t) represents, after heating, die cavity surface temperature uniformity can represent with die cavity surface temperature difference (Δ T) after heating, this enforcement The temperature reached needed for heating in example is 120 DEG C；

Step S200, uses Central Composite experimental design (CCD) the planning sample of four factor five levels, it is possible to less Testing site reaction designing space characteristics, the design parameter factor is as shown in table 1 with level.Mold cavity surface after heat time and heating The temperature difference is analyzed by finite element thermal response and is obtained.

Table 1

Step S300, BP Establishment of Neural Model, its step includes:

Step S301, determines training sample, uses three layers of BP neutral net, with tetra-design parameters of d, q, h, p as network Input, after heat time and heating, die cavity surface temperature difference exports for network；

Step S302, creates BP neutral net and carries out network training.In the present embodiment, the input layer of BP neutral net and Hidden layer uses tangent s type function (tansig) to use linear function (purelin) as transmission function, output layer, trains letter Number uses function (trainglm)；Hidden layer is 9, and network structure is 4-9-2, and training precision is 0.001, maximum training training time Number is 500, and learning rate is 0.05；

Step S303, according to mean square error (MSE) and the coefficient of determination (R²) carry out training of judgement effect.Fig. 3 is BP nerve net Network training objective convergence curve, reaches desired value 0.001 in the 13rd generation.Fig. 4 is BP neural metwork training fitted figure, the coefficient of determination R²Be 0.99935, i.e. BP network has preferable matching to test data, after the BP neural network model obtained can be used for Multiple-objection optimization.

Further, BP neural network model is optimized by described step S400 by NSGA2 multi-objective genetic algorithm Step as follows:

Step S401, determine two object functions that ERHCM injection mold heating system optimization exports, i.e. heat time with The BP neural network model of die cavity surface temperature difference after heating；

Step S402, confirms the constrained type of multi-objective optimization question, and this is nonlinear restriction, confirms optimal design parameter Bound, lower limit lb=[4 15 4 5], upper limit ub=[8 35 8 9]；

Step S403, initializes population M, and Population Size is 100, and one size of stochastic generation is the parent population P of N_t；

Step S404, carries out target function value calculating to current population at individual；

Step S405, carries out non-bad layer sorting to population at individual；

Step S406, uses binary algorithm of tournament selection, intersection and mutation operation to produce N number of progeny population Q_t；

Step S407, population P_tWith population Q_tIt is incorporated into R_tIn, R_t=P_t∪Q_t；

Step S408, to new population R_tMiddle individuality carries out target function value calculating；

Step S409, carries out non-dominated ranking to population at individual；

Step S410, selects top n individuality to produce parent population P_t+1；

Step S411, if reaching the condition of convergence, terminates；Otherwise algebraically increases by 1, goes to step S404 step；

Step S412, exports Pareto optimal solution set, and Fig. 5 is optimal solution set Pareto front end figure, and corresponding Pareto solves Collection data are as shown in table 2.

Table 2

Further, described step S500 uses the entropy method of weighting determine that the objective weight step of each optimization aim is as follows:

Pareto optimal solution set data are done standardization by step S501, obtain specified decision: $y_{ij}=\frac{max\left\{X_j\right\}-x_{ij}}{max\left\{X_j\right\}-min\left\{X_j\right\}},(1\≤i\≤m,1\≤j\≤n),$

Wherein X_ijRepresent the numerical value of i-th optional program jth optimization aim, max{X_jAnd min{X_jAll sides to be selected The maximum of jth item evaluation index and minimum of a value in case, m is optional program number, and n is optimization aim number.In the present embodiment, m is 30, n is 2；

Step S502, processes decision matrix further, obtains matrix P=(p_ij)_m×n: $p_{ij}=\frac{y_{ij}}{\underset{i=1}{\overset{m}{\Σ}}{y}_{ij}},$ $(1\≤i\≤m,1\≤j\≤n);$

Step S503, comentropy e of parameter output_j: $e_j=-\frac{1}{lnm}\underset{i=1}{\overset{m}{\Σ}}{p}_{ij}\ln p_{ij},$ $(1\≤i\≤m,1\≤j\≤n);$

Step S504, computation attribute weight vectors ω_j: After the two optimization aim heat times obtained in the present embodiment and heating, the weight of die cavity surface temperature difference is respectively 0.4486, 0.5514。

Further, described step S600 carries out priority ordered acquisition based on TOPSIS strategy to Pareto optimal solution set Optimal design step is as follows:

Step S601, makees standardization processing to decision matrix, obtains specified decision matrix Y=(y_ij)_m×n, wherein $y_{ij}=\frac{max\left\{X_j\right\}-x_{ij}}{max\left\{X_j\right\}-min\left\{X_j\right\}},(1\≤i\≤m,1\≤j\≤n);$

Step S602, utilizes step S500 to obtain objective weight ω of optimization aim_j；

Step S603, calculates weighted normal decision matrix Z=(z_ij)_m×n, wherein z_ij=ω_jy_ij,(1≤i≤m,1≤j ≤n)；

Step S604, determines positive ideal solution Z⁺With minus ideal result Z^-: $Z^{-}=(z_1^{-},z_2^{-},...,z_n^{-},)=0,$ Wherein $z_j^{+}=\underset{1\≤i\≤m}{\max }{z}_{ij}={\mathrm{\ω}}_j,z_j^{-}=\underset{1\≤i\≤m}{\min }{z}_{ij}=0;$

Step S605, calculates each scheme Euclid distance to positive ideal solution and minus ideal resultWith

$d_i^{+}=\sqrt{\underset{i=1}{\overset{n}{\Σ}}{(z_{ij}-z_j^{+})}^2},d_i^{-}=\sqrt{\underset{i=1}{\overset{n}{\Σ}}{(z_{ij}-z_j^{-})}^2},1\≤i\≤m;$

Step S605, calculates the relative similarity degree of each schemeResult is as shown in table 2:

$C_i^{+}=\frac{d_i^{-}}{d_i^{+}+d_i^{-}},1\≤i\≤m;$

Step S606, arranges the precedence of each scheme: relative similarity degree is the biggest the most excellent, relatively pastes according to relative similarity degree Recency is the least the most bad.In table 3,Maximum scheme number is 29, each design parameter d, q, h, p difference that this scheme number is corresponding Being 5.7821,32.4783,6.1072,5.0056, the corresponding heat time is 7.7153s, and after heating, die cavity surface temperature difference is 5.5441 DEG C, meet the efficiency of heating surface and the mold cavity surface temperature homogeneity requirement of mould, adopt during compensate for traditional design Weak point by trial-and-error method adjusted design parameter.

Table 3

The above is the desirable embodiment of the present invention, is not limited to the present invention, and relevant staff is the most permissible According to the technical scheme of this invention, carry out suitable change and amendment or equivalent.The technical model of this invention Enclose the content being not limited on specification, its technical scope must be confirmed with reference to right.

Claims (7)

1. an electric heating based on BP-NSGA2 multiple-objection optimization with TOPSIS multiple attribute decision making (MADM) becomes mould temperature injection mold and heats system System optimization method, it is characterised in that these heating means comprise the following steps:

Step S100, sets up the thermal response of mould heating period and analyzes model, wherein with the specification of electrical bar and space layout for setting Meter variable, the efficiency of heating surface and mold cavity surface temperature homogeneity are optimization aim；

Step S200, uses Central Composite experimental design CCD to obtain test sample；

Step S300, is trained sample, and sets up the BP neutral net mould of die cavity surface temperature difference after heat time and heating Type；

Step S400, uses NSGA2 multi-objective genetic algorithm to be optimized BP neural network model, it is thus achieved that Pareto optimal solution Collection；

Step S500, uses the entropy method of weighting to determine the objective weight of each optimization aim；

Step S600, carries out priority ordered based on TOPSIS strategy to Pareto optimal solution set and obtains optimal heat protocol.

Electric heating the most according to claim 1 becomes mould temperature injection mold heating system optimization method, it is characterised in that described step In rapid S100, use 4 parameters to be described, be that the diameter of electrical bar, power density, electrical bar wall are away from mold cavity surface respectively Distance between distance and electrical bar wall.

Electric heating the most according to claim 1 and 2 becomes mould temperature injection mold heating system optimization method, it is characterised in that institute State in step S200, use the Central Composite experimental design CCD planning sample of four factor five levels.

Electric heating the most according to claim 2 becomes mould temperature injection mold heating means, it is characterised in that described step S300 In, BP Establishment of Neural Model step is as follows:

Step S301, determines training sample, with the diameter of electrical bar, power density, electrical bar wall away from mold cavity surface distance with And four design parameters of the distance between electrical bar wall are network input, with die cavity surface temperature difference after heat time and heating it is Network exports；

Step S302, creates BP neutral net and carries out network training；

Step S303, according to mean square error MSE, it is thus achieved that the BP neural network model needed for required multiple-objection optimization.

5. become mould temperature injection mold heating system optimization method according to the electric heating described in claim 1,2 or 4, it is characterised in that In described step S400, being optimized BP neural network model by NSGA2 multi-objective genetic algorithm, process is as follows:

Step S401, determines two object functions that ERHCM injection mold heating system optimization exports, i.e. heat time and heating Rear die cavity surface temperature difference；

Step S402, confirms the constrained type of multi-objective optimization question, and this is nonlinear restriction, confirms the upper of optimal design parameter Lower limit；

Step S403, initializes population M, and one size of stochastic generation is the parent population P of N_t；

Step S404, carries out target function value calculating to current population at individual；

Step S405, carries out non-bad layer sorting to population at individual；

Step S406, uses binary algorithm of tournament selection, intersection and mutation operation to produce N number of progeny population Q_t；

Step S407, population P_tWith population Q_tIt is incorporated into R_tIn, R_t=P_t∪Q_t；

Step S408, to new population R_tMiddle individuality carries out target function value calculating；

Step S409, carries out non-dominated ranking to population at individual；

Step S410, selects top n individuality to produce parent population P_t+1；

Step S411, if reaching the condition of convergence, terminates；Otherwise, algebraically increases by 1, turns step S404 step；

Step S412, exports Pareto optimal solution set.

Electric heating the most according to claim 1 and 2 becomes mould temperature injection mold heating system optimization method, it is characterised in that institute Stating in step S500, use the entropy method of weighting to determine the objective weight of each optimization aim, process is as follows:

Pareto optimal solution set data are done standardization by step S501, obtain specified decision:

$y_{ij}=\frac{max\left\{X_j\right\}-x_{ij}}{max\left\{X_j\right\}-min\left\{X_j\right\}},(1\≤i\≤m,1\≤j\≤n),$

Wherein X_ijRepresent the numerical value of i-th optional program jth optimization aim, max{X_jAnd min{X_jIn all optional programs The maximum of jth item evaluation index and minimum of a value, m is optional program number, and n is optimization aim number；

Step S502, processes decision matrix, obtains matrix P=(p_ij)_m×n:

Step S503, comentropy e of parameter output_j: $e_j=-\frac{1}{\ln m}\underset{i=1}{\overset{m}{\Σ}}{p}_{ij}\ln p_{ij},$ $(1\≤i\≤m,1\≤j\≤n);$

Step S504, computation attribute weight vectors ω_j: ${\mathrm{\ω}}_j=\frac{1-e_j}{\underset{j=1}{\overset{n}{\Σ}}(1-e_j)},\underset{j=1}{\overset{n}{\Σ}}{\mathrm{\ω}}_j=1,(1\≤i\≤m,1\≤j\≤n).$

Electric heating the most according to claim 1 and 2 becomes mould temperature injection mold heating system optimization method, it is characterised in that institute State in step S600, based on TOPSIS strategy, Pareto optimal solution set is carried out priority ordered and obtain optimal computer heating control side Case, process is as follows:

Step S601, makees standardization processing to decision matrix, obtains specified decision matrix Y=(y_ij)_m×n, wherein $y_{ij}=\frac{max\left\{X_j\right\}-x_{ij}}{max\left\{X_j\right\}-min\left\{X_j\right\}},(1\≤i\≤m,1\≤j\≤n);$

Step S602, utilizes step S500 to obtain objective weight ω of optimization aim_j；

Step S603, calculates weighted normal decision matrix Z=(z_ij)_m×n, wherein z_ij=ω_jy_ij,(1≤i≤m,1≤j≤ n)；

Step S604, determines positive ideal solution Z⁺With minus ideal result Z^-: $Z^{-}=(z_1^{-},z_2^{-},...,z_n^{-},)=0,$ Wherein $z_j^{+}=\underset{1\≤i\≤m}{max}{z}_{ij}={\mathrm{\ω}}_j,z_j^{-}=\underset{1\≤i\≤m}{min}{z}_{ij}=0;$

Step S605, calculates each scheme Euclid distance to positive ideal solution and minus ideal resultWith

$d_i^{+}=\sqrt{\underset{i=1}{\overset{n}{\Σ}}{(z_{ij}-z_j^{+})}^2},d_i^{-}=\sqrt{\underset{i=1}{\overset{n}{\Σ}}{(z_{ij}-z_j^{-})}^2},1\≤i\≤m;$

Step S605, calculates the relative similarity degree of each scheme

$C_i^{+}=\frac{d_i^{-}}{d_i^{+}+d_i^{-}},1\≤i\≤m;$

Step S606, arranges the precedence of each scheme: relative similarity degree is the biggest the most excellent, relative similarity degree according to relative similarity degree The least the most bad.