CN109992881A - A kind of intelligent Nonlinear Process Planning Method towards STEP-NC manufacturing feature - Google Patents

Abstract

The present invention provides a kind of intelligent Nonlinear Process Planning Method towards STEP-NC manufacturing feature.The method of the present invention includes: the non-linear processing scheme that STEP-NC manufacturing feature is determined by improved BP；According to tooling step principle of ordering, double-strand table algorithm is used to be ranked up for all tooling steps in above-mentioned processing scheme, obtains several reasonable tooling step sequences；It is that each work step matches corresponding operation resource, and determines optimal tooling step sequence and working process parameter by rule-based operation resource matching algorithm and artificial bee colony algorithm.The method of the present invention, which combines BP neural network, double-strand table algorithm, artificial bee colony algorithm and rule, is applied to the process planning based on STEP-NC, the efficiency of process planning, accuracy, non-linear and intelligent can be improved, to realize that open, intelligent and networking STEP-NC digital control system is laid a good foundation, and to the further perfect of STEP-NC standard and implement that there is positive reference value.

Description

A kind of intelligent Nonlinear Process Planning Method towards STEP-NC manufacturing feature

Technical field

The present invention relates to process planning technical fields, specifically, more particularly to a kind of towards STEP-NC manufacturing feature Intelligent Nonlinear Process Planning Method.

Background technique

As a kind of novel NC programming data interface international standard, STEP-NC is by the way of based on manufacturing feature Description processing object and all information needed for covering processing, describe processing is what, what does, rather than how It does, this is realizes that open, intelligent, networking digital control system provides condition, development to intelligence manufacture or even entire Manufacturing industry will exert far reaching influence.And bridge and tie as connection design and manufacture, process planning are to implement STEP-NC An important step, and its intelligence have become realize it is open, intelligent and networking STEP-NC digital control system one The problem of a urgent need to resolve.

Domestic and foreign scholars have done many significant researchs to the process planning based on STEP-NC.For example, in STEP-NC In terms of tooling step sequence optimisation, Mokhtar and Xu propose rule-based STEP-NC tooling step sort algorithm, realize The sequence of STEP-NC tooling step, but the algorithm needs to formulate a large amount of rule, and many rules are difficult to formulate, therefore intelligence Energy property is poor, versatility is not strong.Later, the Ou Yanghuabing and Shen Bin of Tongji University propose it is a kind of towards STEP-NC based on mixing The tooling step sequence optimisation method of formula genetic algorithm.Since there is no solving, genetic algorithm is sensitive to initial value to be lacked this method Point is not high so as to cause its efficiency and accuracy.In addition, yellow wind is vertical et al. and Wang et al. is respectively from processing efficiency and processing energy It consumes angle and uses ant colony optimization for solving STEP-NC manufacturing feature work step sequence optimisation problem.Because there is convergence speed in ant group algorithm The problem of spending slowly and being easily trapped into local optimum, therefore the optimization efficiency of above-mentioned two method is not high and accuracy is poor.And In terms of STEP-NC machining parameters optimization, Shandong University Zhang Hongbo develops the Turning Features machining parameters optimization based on STEP-NC Module, which realizes the optimization of STEP-NC turning process parameter by linear weighted combination method, but the module can only Solve the problems, such as the machining parameters optimization of outer circle feature, therefore versatility is not strong.Zhang et al. constructs a kind of based on STEP-NC Machining parameters optimization system PPS, the machined parameters which inputs according to operator utilize theoretical residual height algorithm to calculate Surface roughness after cutting out, to judge whether machined parameters meet processing request, but the selection of the system machined parameters The know-how of technologist is excessively relied on, therefore intelligence and versatility be not strong.Ridwan et al. propose it is a kind of stage by stage Fe ed drive algorithm, the algorithm can distinguish the coarse-fine process segment, and adaptively excellent according to the different process segments Change machined parameters, but the algorithm can only realize the optimization of slabbing strategy lower feeding rate, versatility is poor.In conclusion mesh Although preceding done many significant grind to STEP-NC tooling step sequence optimisation and STEP-NC machining parameters optimization both at home and abroad Study carefully, but the method that most of research institute proposes has the disadvantages of intelligence is poor, versatility is not strong, low efficiency.Moreover, there are no Close the research to the non-linear process operation program decisions of STEP-NC.

Summary of the invention

According to it is above-mentioned it is existing based on the Process Planning Method of STEP-NC there are the problem of, the present invention provides one kind towards The intelligent Nonlinear Process Planning Method of STEP-NC manufacturing feature.

The technological means that the present invention uses is as follows:

A kind of intelligent Nonlinear Process Planning Method towards STEP-NC manufacturing feature, which is characterized in that including following step It is rapid:

Step S1: the non-linear processing scheme of STEP-NC manufacturing feature is determined by improved BP；Specific packet It includes:

Step S101: process operation method decision BP neural network model of the building towards STEP-NC manufacturing feature；

Step S102: artificial bee colony algorithm Optimized BP Neural Network initial weight and threshold value are utilized, optimal initial power is obtained Value and threshold value；

Step S103: it using the obtained optimal initial weight of above-mentioned steps S102 and Threshold-training BP neural network, obtains Improved BP for feature machining operating method decision；

Step S104: it will be input in improved BP after the machining information normalized of part, pass through improvement The nested reasoning of BP neural network exports the non-linear processing scheme of STEP-NC manufacturing feature；

Step S2: according to tooling step principle of ordering, use double-strand table algorithm for all processing in above-mentioned processing scheme Work step is ranked up, and obtains several reasonable tooling step sequences；It specifically includes:

Step S201: tooling step sequence any in above-mentioned non-linear processing scheme is stored in a doubly linked list, note For LL；

Step S202: picking out the tooling step without any relation constraint between other work steps in doubly linked list LL, Their positions in chained list are recorded simultaneously, and the above-mentioned tooling step without any relation constraint is successively stored in one newly Doubly linked list LL1 in, remaining tooling step sequence then forms another new doubly linked list LL2；If tooling step sequence In without mutual unallied work step, then direct duplication LL chained list；

Step S203: using the tooling step of tail portion as current traversal point, doubly linked list LL2 is traversed；Double All work steps that should be come after current traversal point according to work step the constraint relationship are found out into chained list LL2, by above-mentioned tooling step It takes out and forms sub- doubly linked list LL21 according to its original sequence；After doubly linked list LL21 is inserted into current traversal point, New doubly linked list LL2 is regenerated, current traversal point is shifted along at the tooling step of the tail portion new doubly linked list LL2, right New traversal point similarly uses the above method to be handled, and operates the tooling step satisfaction until doubly linked list most end repeatedly Until all work step the constraint relationships；

Step S204: current traversal point is moved at its previous tooling step, as new traversal point, processing Method is with step S203, until tooling step all in doubly linked list has all been handled, conversion process terminates；

Step S205: by be stored in above-mentioned steps S202 in doubly linked list LL1 between other work steps without any pass The tooling step of system's constraint is filled out again according to their original positions to be returned in chained list, and entire conversion process is completed；

Step S3: being that the matching of each work step is corresponding by rule-based operation resource matching algorithm and artificial bee colony algorithm Operation resource, and determine optimal tooling step sequence and working process parameter；It specifically includes:

Step S301: being that each work step matches corresponding operation resource by rule-based operation resource matching algorithm； Specific algorithm include the matching of rule-based machining tool, the matching of rule-based process tool and machining tool and cutter it Between matching；

A, rule-based machining tool matching algorithm is as follows:

A1, machine-tool database, if k=1, matching machine tool M are called_k；

A2, successively judgement " range of work of blank be less than each axis of lathe range? ", " blank weight be less than lathe Maximum capacity? ", " speed of mainshaft needed for part be less than lathe maximum (top) speed? ", " machine finish meets part Requirement on machining accuracy? ", " lathe miscellaneous function meets part processing request? " if all meeting the requirements, by lathe M_kDeposit In the candidate lathe set of workpiece, A3 is executed；If there is a condition to be unsatisfactory for requiring, A4 is executed；

A3, judge whether k is less than T.If so, k=k+1, returns to A2；If it is not, then exporting work step and its candidate lathe Set relation；

A4, judge whether k is less than T.If so, k=k+1, returns to A2；If it is not, then executing A5；

A5, judge whether candidate lathe set is empty.If so, it fails to match for lathe；If it is not, then exporting work step and its time The set relation of lathe is selected, matching machine tool terminates.

B, the matched specific algorithm of rule-based process tool is as follows:

B1, cutter database is called, obtains the cutter set T that can process the type feature_nIf n=1；

B2, cutter T is obtained_n；

B3, successively judgement " tool dimension be less than characteristic size? ", " hardness of the cutter material hardness than processing part It is high? ", " precision and surface roughness of part are met the requirements after tool sharpening? ", " surplus after Tool in Cutting meet work step spy Fixed machining allowance range? " if all meeting the requirements, by cutter T_nIt is stored in the candidate cutter set of workpiece, executes B4； If there is a condition to be unsatisfactory for requiring, B5 is executed；

B4, judge whether n is less than N.If so, n=n+1, returns to B2；If it is not, then exporting work step and its candidate cutter Set relation；

B5, judge whether n is less than N.If so, n=n+1, returns to B2；If it is not, then executing B6；

B6, judge whether candidate cutter set is empty.If so, it fails to match for cutter；If it is not, then exporting work step and its time The set relation of cutter is selected, matching cutter terminates.

C, the matching between machining tool and cutter, by comparing all available process tools for being selected with selected The cutter that machining tool is equipped with, for the corresponding candidate process tool of machining tool matching.

Step S302: optimal tooling step sequence is determined by artificial bee colony algorithm；Specific algorithm is as follows:

Step S3021: initializing the parameter of artificial bee colony algorithm, and the parameter includes the number S of initial solution_N, limiting value Limit, maximum cycle Gen_Max lead the quantity N of bee₁, follow the quantity N of bee₂, and N₁=N₂=S_N；

Step S3022: initial nectar source is chosen from step S2 and it is encoded；

Step S3023: establishing fitness function, and calculates the fitness value in each nectar source；

Using cutter changing total time between work step as optimization aim, therefore objective function can be expressed as follows:

G (x)=Min (f₁,f₂,f₃,......f_m) (13)

In formula, m is the quantity in nectar source；f_iFor the corresponding tool change time in i-th of nectar source, i.e. i-th tooling step sequence institute Corresponding tool change time.And the tool change time in each nectar source is calculated by following equalities:

In formula, i=1,2,3 ... m, m are the quantity in nectar source；J=1, the work step that 2,3 ... n-1, n include by each nectar source Number；t₁The cutter changing time of [j] between continuous twice work step；T [j] is tool code corresponding to j-th of work step, right Above formula has following equation to set up:

In order to guarantee that the optimization direction of objective function corresponds to the direction that fitness value increases, fitness function and mesh are established The mapping relations of scalar functions, the fitness value after guaranteeing mapping is non-negative.Since above-mentioned objective function is to belong to minimum to ask Topic, therefore have mapping relations below for fitness function f (x) and objective function g (x):

In formula, c_MaxIt can be with relatively large number for one.

Step S3024: leading peak to carry out local search according to cross and variation strategy, and calculate the fitness value in new nectar source, If the fitness value of new explanation is bigger than the fitness value of old solution, otherwise the stagnation number of old solution is added 1 by new and old solution；

D, Crossover Strategy process is as follows:

Two work step sequences are arbitrarily chosen in population, and crossover operation is carried out by the way of two-point crossover.Assuming that two Work step sequence is respectively original process route 1 and original process route 2, the specific steps are as follows:

D1, two crosspoints are randomly selected.

D2, the element at 1 crosspoint both ends of original process route is copied in new process route same position.

D2, other elements obtain new process by the copy orderly to new process route rest position in original process route 2 Route.

E, Mutation Strategy process is as follows:

Mutation Strategy uses two o'clock exchange process, i.e., randomly chooses two operations in a process route and swap.Become If occurring invalid work step sequence after ETTHER-OR operation, it is necessary to invalid work step sequence be become effective by work step validation algorithm Work step sequence.

Step S3025: following bee according to roulette selection nectar source, and uses and the friendship for leading bee identical in step S3024 Fork Mutation Strategy carries out local search and calculates new nectar source fitness value, if the fitness value of new explanation is than the fitness value of old solution Greatly, then new and old solution, otherwise adds 1 for the stagnation number of old solution；

Step S3026: judge to stagnate whether number is greater than limiting value limit, if so, thening follow the steps S3027；If it is not, Then follow the steps S3028；

Step S3027: investigation bee generates a process route at random and carries out global search, to the process route generated at random Reasonableness check is carried out, if there is unreasonable route, algorithm is rationalized using work step and is translated into reasonable work step sequence Column, and calculate new nectar source fitness value；

Step S3028: judging whether to reach maximum cycle, if so, the tooling step sequence that output is optimal, if No, then the number of iterations adds 1, return step S3024；

Step S303: optimal working process parameter is determined by artificial bee colony algorithm；Specific algorithm is as follows:

Step S3031: multi-goal optimizing function model is established；It specifically includes:

F, determine variable: variable is feed engagement f_ZWith Milling Speed V；

G, it establishes multi-goal optimizing function: establishing multi-goal optimizing function using process time and processing cost as target；

G1: process time objective optimization function is established, when the process time includes cutting time, tool change time and auxiliary Between；The objective function of process time are as follows:

In formula, t is process time, t_mFor the cutting time, T is tool life, i.e. cutter life, t_m/ T is number of changing knife, t_ctTo change time consumed by a knife, t_oFor other unproductive times in addition to tool change time；

The calculation formula of cutting time and tool life is as follows:

In formula, l is length of cut, and h is thickness of cutting, and D is cutter diameter, and Z is the cutter number of teeth, f_zFor feed engagement, v For cutting speed, a_pFor back engagement of the cutting edge, a_eFor cutting width, K_t, a, b, c, d, e and f be empirical index number；

Formula (18) and (19), which are substituted into formula (17), to be obtained:

G2: establishing processing cost objective optimization function, and the processing cost includes time cost and the cost of charp tool；It is processed into This objective function are as follows:

In formula, c is processing cost, c₀For unit time cost, c_tFor cutter material cost；

Formula (18), (19) and (20) substitution formula (21) can be obtained:

G3: establishing process time and processing cost multi-goal optimizing function, if f₁(f_z, v) and=t, then have:

If f₂(f_z, v) and=c, then have:

The different demands for meeting different user, multiple objective function are adjusted by introducing weighting coefficient λ are as follows:

F(f_z, v) and=λ₁f₁(f_z,v)+λ₂f₂(f_z,v) (25)

In formula, λ₁And λ₂For weighting coefficient, and λ₁+λ₂=1,0≤λ₁≤ 1,0≤λ₂≤1；

Formula (23) and formula (24), which are substituted into formula (25), to be obtained:

H, determine constraint condition: according to the requirement of machining tool, process tool and processing quality, constraint condition is specifically included Maximum feeding cutting force constraint, maximum principal axis torque constraint, maximum working power constraint, workpiece surface quality constraint, cutting speed Degree constraint and feed engagement constraint；

H1, maximum feeding cutting force constraint:

In formula, F_fMaxFor maximum allowable cutting force, K_F, m, n, p and w be cutting force empirical coefficient；

H2, maximum principal axis torque constraint:

In formula, T_qMaxFor maximum permissible torque, K_qFor torque empirical coefficient；

H3, maximum working power constraint:

In formula, P_MaxFor maximum allowable power, K_pFor power empirical coefficient；

H4, workpiece surface quality constraint:

In formula,For maximum allowable surface roughness, K_sFor surface roughness empirical coefficient；

H5, cutting speed constraint:

In formula, N_MinAnd N_MaxFor the lathe minimum and maximum speed of mainshaft, v_MinAnd v_MaxFor minimum and maximum cutting speed；

H6, feed engagement constraint:

In formula,WithFor lathe minimum and maximum feed speed,WithFor the feeding of minimum and maximum per tooth Amount；

Step S3032: the technological parameter intelligent optimization based on artificial bee colony algorithm；Specific algorithm is as follows:

I1, according to the processing conditions given, each machined parameters in initialization function model；

I2, the parameter and initialization population for initializing artificial bee colony algorithm；The parameter of the artificial bee colony algorithm includes just The number S for the solution that begins_N, limiting value limit, maximum cycle Gen_Max；Lead the quantity N of bee₁, follow the quantity N of bee₂, and N₁=N₂=S_N；Initial nectar source, that is, initial solution X_i(i=1,2,3...S_N) by feed engagement f_ZIt is formed with Milling Speed V；

I3, fitness function is established, and calculates the fitness value in each nectar source；

Fitness function are as follows:

c_MaxIt is a biggish number, there will be constrained optimization problem to be converted into solution without constraint using compound penalty function Optimization problem, mixed penalty function such as following formula indicate；

In formula, F (x) is former objective function, h_iIt (x) is equality constraint, g_jIt (x) is the punishment of inequality constraints condition , k is the number of equality constraint, and l is the number of inequality constraints, and M is penalty factor and M⁰<M¹<M²...→∞；

The mathematical model that the cutting parameter of above-mentioned foundation optimizes is brought into formula (34) to obtain:

It brings formula (35) into formula (33) again and obtains fitness function；

I4, it leads bee to carry out neighborhood search, and carries out nectar source update；

It leads bee to carry out neighborhood search according to formula (36), and calculates the fitness value of new explanation according to formula (33)；If The fitness value of new explanation is bigger than the fitness value of old solution, then new and old solution, and the stagnation number of old solution is otherwise added 1；

V_ij=X_ij+rand(-1,1)(X_ij-X_kj) (36)

In formula, i=1,2,3...S_N, S_NFor population number, j=1,2,3...D, D are the dimension of solution, V_ijIt is after search i-th J-th of component value of a solution, X_ijFor j-th of component value for searching for preceding i-th of solution, X_kjJ-th of component for the solution being randomly generated Value, k ∈ { 1,2,3...S_N, and k ≠ i；If V_ijIt has been more than the maximum value allowed, then has been converted to boundary value according to the following formula:

In formula, lowerbound is value lower bound, and upperbound is the value upper bound；

I5, bee is followed to be updated according to roulette method choice nectar source, and to nectar source；

It follows bee according to formula (38) using roulette method choice nectar source, and nectar source is updated with formula (33)；Such as The fitness value of fruit new explanation is bigger than the fitness value of old solution, then new and old solution, and the stagnation number of old solution is otherwise added 1；

In formula, i=1,2,3...S_N, S_NFor population number, X_iIt is solved for i-th, P_iThe select probability solved for i-th, f (X_i) be The fitness value of i-th of solution；

I6, judge to stagnate whether number is greater than limiting value limit, be searched if so, search bee carries out the overall situation according to formula (39) Rope and the fitness value for calculating new explanation；If it is not, then executing I7；

In formula, i=1,2,3...S_N, S_NFor population number, j=1,2,3...D, D are the dimension of solution, X_i ^jIt is after search i-th J-th of component value of a solution,For the minimum value of j-th of component in population,For the maximum of j-th of component in population Value, rand (0,1) are the random number in (0,1) range；

I7, judge whether to reach maximum cycle, if so, the technological parameter that output is optimal, if it is not, then the number of iterations Add 1, returns to I4.

Further, the process that BP neural network model is constructed in the step S101 is as follows:

Step S1011: the input layer of neural network is determined according to the factor number for influencing feature machining operating method decision Number n₁；

Step S1012: rule of thumb formula n₂=2n₁+ 1 determines implicit number of plies n₂；

Step S1013: the output number of plies of neural network is determined according to Milling Process operation data model in STEP-NC standard m。

Further, artificial bee colony algorithm Optimized BP Neural Network initial weight and threshold value are utilized in the step S102 Process is as follows:

Step S1021: initializing the parameter and initialization population of artificial bee colony algorithm, and the parameter includes initial solution Number S_N, limiting value limit, maximum cycle Gen_Max lead the quantity N of bee₁, follow the quantity N of bee₂, and N₁=N₂= S_N；The initialization population, that is, initial solution X_i(i=1,2,3...S_N) input layer of BP neural network of creation connect with hidden layer Weight matrix w_ij, hidden layer and output layer connection weight matrix w_jk, hidden layer threshold value matrix a_j, output layer threshold matrix b_kFour It is grouped as；The initial solution of each part is a D dimensional vector, and D is calculated by formula (1):

D=N_input×N_hidden+N_hidden+N_hidden×N_output+N_output (1)

In formula, N_input、N_hidden、N_outputIt is the mind of the input layer of BP neural network of building, hidden layer, output layer respectively Through first number；

Step S1022: establishing fitness function, and calculates the fitness value in each nectar source；

Fitness function are as follows:

In formula, i=1,2,3...S_N, MSE_iIt indicates i-th of BP network mean square error solved, is when fitness value reaches 1 Optimal state；

Step S1023: it leads bee to carry out neighborhood search according to following formula (3), and calculates the suitable of new explanation according to formula (2) Answer angle value；If the fitness value of new explanation is bigger than the fitness value of old solution, otherwise new and old solution adds the stagnation number of old solution 1；

V_ij=X_ij+rand(-1,1)(X_ij-X_kj) (3)

In formula, i=1,2,3...S_N, S_NFor population number, j=1,2,3...S_ND, D are the dimension of solution, V_ijIt is after search i-th J-th of component value of a solution, X_ijFor j-th of component value for searching for preceding i-th of solution, X_kjJ-th of component for the solution being randomly generated Value, k ∈ { 1,2,3...S_N, and k ≠ i；If V_ijIt has been more than the maximum value allowed, then has been converted to boundary value according to the following formula:

In formula, lowerbound is value lower bound, and upperbound is the value upper bound；

Step S1024: it follows bee according to formula (5) using roulette method choice nectar source, and nectar source is carried out with formula (3) It updates；If the fitness value of new explanation is bigger than the fitness value of old solution, otherwise the stagnation number of old solution is added 1 by new and old solution；

In formula, i=1,2,3...S_N, S_NFor population number, X_iIt is solved for i-th, P_iThe select probability solved for i-th, f (X_i) be The fitness value of i-th of solution；

Step S1025: judging whether the stagnation number of solution is greater than limiting value limit, if so, search bee is according to formula (6) It carries out global search and calculates the fitness value of new explanation；If it is not, thening follow the steps S1026；

In formula, i=1,2,3...S_N, S_NFor population number, j=1,2,3 ... D, D are the dimension of solution, X_i ^jIt is i-th after search J-th of component value of solution,For the minimum value of j-th of component in population,For the maximum value of j-th of component in population, Rand (0,1) is the random number in (0,1) range；

Step S1026: judge whether to reach maximum cycle or meet required precision；If so, output is optimal just Beginning weight and threshold value execute step S103；If it is not, then the number of iterations adds 1, return step S1023.

Further, the mistake of obtained optimal initial weight and Threshold-training BP neural network is utilized in the step S103 Journey is as follows:

Step S1031: obtained initial weight and threshold value are brought into BP neural network, sequentially input sample data into Row training, the output of hidden layer and output layer is calculated according to formula (7) and (8), calculates each sample training according to formula (9) Mean square error calculates sample overall error according to formula (10)；Then have:

Hidden layer exports H_j:

In formula, f₁It (x) is general hidden layer excitation function, n₁To input the number of plies, w_ijFor input layer and hidden layer connection weight square Battle array, x_iFor input sample, a_jFor hidden layer threshold value matrix, n₂To imply the number of plies；The output O of output layer_K:

In formula, f₂It (x) is output layer excitation function, n₂To imply the number of plies, H_jFor hidden layer output, w_jkFor hidden layer with it is defeated Layer connection weight matrix out, b_kFor output layer threshold matrix, m is the output number of plies；

The error function of p-th of sample:

In formula, m is the output number of plies,For the kth layer desired output of sample p,It is actually defeated for the kth layer of sample p Out；

The overall error of P sample:

In formula, P is sample total, and p is current input sample, and m is the output number of plies,It is defeated for the kth layer expectation of sample p Out,For the kth layer reality output of sample p；

Step S1032: judging whether to meet error requirements or reach frequency of training, if satisfied, then training finishes, is used In the improved BP of process operation method decision, step S104 is executed；If not satisfied, then according to formula (11) and (12) The anti-pass error for calculating each layer executes step S1031 according to the anti-pass error update weight and threshold value of each layer；

The anti-pass error of output layer:

In formula,For the anti-pass error of the jth layer of the kth layer and hidden layer of sample p output layer,It is the of sample p K layers of desired output,For the kth layer reality output of sample p, m is the output number of plies；

The anti-pass error of hidden layer:

In formula,For the jth layer of sample p hidden layer and i-th layer of anti-pass error of input layer,For sample p output The anti-pass error of the jth layer of the kth layer and hidden layer of layer, w_jkFor hidden layer and output layer connection weight matrix,It is hidden for this p Jth layer output containing layer；

Further, the nested reasoning process of the improved BP in the step S104 is as follows:

Step S1041: the process operation method and its selecting priority coefficient of final step are exported according to input vector；

Step S1042: obtaining intermediate features according to machining allowance, and judge whether intermediate features correspond to blank, if so, Processing scheme spanning tree is exported, traverses feature machining schemes generation tree from back to front, the selection of each process operation method is excellent First weight coefficient is multiplied, and obtains the non-linear processing scheme of each feature and its select probability；BP nerve net is improved if it is not, then reusing Network obtains the process operation method and its selecting priority coefficient of previous step, and so on until intermediate features are characterized blank When output processing scheme spanning tree and export the non-linear processing scheme of STEP-NC manufacturing feature.

Compared with the prior art, the invention has the following advantages that

1, the invention proposes a kind of intelligent Nonlinear Process Planning Method towards STEP-NC manufacturing feature, this method Realize the decision of the non-linear process operation scheme of STEP-NC manufacturing feature.

2, the invention proposes a kind of intelligent Nonlinear Process Planning Method towards STEP-NC manufacturing feature, this method BP neural network, double-strand table algorithm, artificial bee colony algorithm and rule are combined and are applied to the process planning based on STEP-NC, The efficiency of process planning, accuracy, non-linear and intelligent can be improved, it is open, intelligent and networking to realize STEP-NC digital control system is laid a good foundation.

3, method proposed by the present invention to the further perfect of STEP-NC standard and implements have positive reference value.

The present invention can be widely popularized in fields such as process plannings based on the above reasons.

Detailed description of the invention

In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below There is attached drawing needed in technical description to do simply to introduce, it should be apparent that, the accompanying drawings in the following description is this hair Bright some embodiments for those of ordinary skill in the art without any creative labor, can be with It obtains other drawings based on these drawings.

Fig. 1 is the embodiment of the present invention based on the intelligent Nonlinear processing scheme decision flow diagram for improving neural network.

Fig. 2 is doubly linked list of embodiment of the present invention LL.

Fig. 3 is doubly linked list of embodiment of the present invention LL1 and doubly linked list LL2.

Fig. 4 is the new doubly linked list LL2 of the embodiment of the present invention.

Fig. 5 is the new doubly linked list LL of the embodiment of the present invention.

Fig. 6 is the rule-based machining tool matching algorithm flow chart of the embodiment of the present invention.

Fig. 7 is the rule-based process tool matching algorithm flow chart of the embodiment of the present invention.

Fig. 8 is tooling step of embodiment of the present invention sequence and coding.

Fig. 9 is Crossover Strategy of embodiment of the present invention figure.

Figure 10 is Mutation Strategy of embodiment of the present invention figure.

Figure 11 is tooling step sequence intelligent optimization algorithm flow chart of the embodiment of the present invention based on artificial bee colony algorithm.

Figure 12 is technological parameter intelligent optimization algorithm flow chart of the embodiment of the present invention based on artificial bee colony algorithm.

Figure 13 is threedimensional model of the embodiment of the present invention.

Figure 14 is cavity feature of embodiment of the present invention process operation method decision BP neural network model.

Figure 15 is open type of embodiment of the present invention type chamber feature machining operating method decision BP neural network model.

Figure 16 is step-feature of embodiment of the present invention process operation method decision BP neural network model.

Figure 17 is hole characteristic of embodiment of the present invention process operation method decision BP neural network model.

Figure 18 is closed type of embodiment of the present invention type chamber feature machining operating method decision BP neural network model.

Figure 19 is two kinds of neural metwork training error convergence curve graphs of the embodiment of the present invention.

Figure 20 is step-feature of the embodiment of the present invention 1,5 processing scheme trees.

Figure 21 is cavity feature of the embodiment of the present invention 2,3 processing scheme trees.

Figure 22 is 4 processing scheme tree of closed type of embodiment of the present invention type chamber feature.

Figure 23 is open type of embodiment of the present invention type chamber feature 6,7 processing scheme trees.

Figure 24 is 8 processing scheme tree of hole characteristic of the embodiment of the present invention.

Figure 25 is that the present invention is based on the tooling step sequences after artificial bee colony algorithm optimization.

Figure 26 is the method for the present invention flow chart.

Specific embodiment

In order to enable those skilled in the art to better understand the solution of the present invention, below in conjunction in the embodiment of the present invention Attached drawing, technical scheme in the embodiment of the invention is clearly and completely described, it is clear that described embodiment is only The embodiment of a part of the invention, instead of all the embodiments.Based on the embodiments of the present invention, ordinary skill people The model that the present invention protects all should belong in member's every other embodiment obtained without making creative work It encloses.

It should be noted that description and claims of this specification and term " first " in above-mentioned attached drawing, " Two " etc. be to be used to distinguish similar objects, without being used to describe a particular order or precedence order.It should be understood that using in this way Data be interchangeable under appropriate circumstances, so as to the embodiment of the present invention described herein can in addition to illustrating herein or Sequence other than those of description is implemented.In addition, term " includes " and " having " and their any deformation, it is intended that cover Cover it is non-exclusive include, for example, the process, method, system, product or equipment for containing a series of steps or units are not necessarily limited to Step or unit those of is clearly listed, but may include be not clearly listed or for these process, methods, product Or other step or units that equipment is intrinsic.

Embodiment

To part intelligent Nonlinear process planning as shown in fig. 13 that, the machining information of part is as follows:

(1) step 1,5: its characteristic type be step-feature, feature limit having a size of cross-sectional dimension 120mm × 30mm, blank type are forging, and material type is 45# carbon steel, and surface roughness is 3.2 μm, and heat treatment method is not need heat Processing, grade of tolerance IT8, batch size are small lot production.

(2) slot 2,3: its characteristic type is cavity feature, and feature is limited having a size of cross-sectional dimension 30mm × 20mm, hair Base type is forging, and material type is 45# carbon steel, and surface roughness is 3.2 μm, and heat treatment method is not need to be heat-treated, public Poor grade is IT8, and batch size is small lot production.

(3) closed type type chamber 4: its characteristic type is closed type type chamber feature, and feature is limited having a size of maximum cross section ruler Very little 60mm × 60mm, blank type are forging, and material type is 45# carbon steel, and surface roughness is 6.3 μm, and heat treatment method is It does not need to be heat-treated, grade of tolerance IT10, batch size is small lot production.

(4) open type type chamber 6,7: its characteristic type is open type type chamber feature, and feature is limited having a size of maximum cross section Size 20mm × 20mm, blank type are forging, and material type is 45# carbon steel, and surface roughness is 3.2 μm, heat treatment method Not need to be heat-treated, grade of tolerance IT8, batch size is small lot production.

(5) hole 8: its characteristic type is hole characteristic, and feature is limited having a size of aperture size 20mm, and blank type is forging, Material type is 45# carbon steel, and surface roughness is 0.8 μm, and heat treatment method is not need to be heat-treated, grade of tolerance IT8, batch Size is measured as small lot production.

Based on the machining information to above-mentioned part, as shown in figure 26, the present invention provides one kind to manufacture spy towards STEP-NC The intelligent Nonlinear Process Planning Method of sign, comprising the following steps:

Step S1: the non-linear processing scheme of STEP-NC manufacturing feature is determined by improved BP；Specific packet It includes:

Step S101: as seen in figs. 14-18, the process operation method decision BP mind towards STEP-NC manufacturing feature is constructed Through network model；

Step S1011: the input layer of neural network is determined according to the factor number for influencing feature machining operating method decision Number n₁；It is 8 typical machining attributes: feature class present invention primarily contemplates the factor that feature machining operating method selects is influenced Type, feature limit size, blank type, material type, surface roughness, heat treatment method, the grade of tolerance, batch size, really Surely the input number of plies is 8；

Step S1012: rule of thumb formula n₂=2n₁+ 1 determines implicit number of plies n₂It is 17；

Step S1013: the output number of plies of neural network is determined according to Milling Process operation data model in STEP-NC standard m；Present invention primarily contemplates planes to rough mill, plane finish milling, side are rough milled and 6 kinds of Milling Process operating methods such as side finish-milling or general 9 kinds of drilling process operation methods such as logical brill, centre drill, spot-facing and thread milling determine that the output number of plies is 6 or 9；

Step S102: artificial bee colony algorithm Optimized BP Neural Network initial weight and threshold value are utilized, optimal initial power is obtained Value and threshold value；

Step S1021: initializing the parameter and initialization population of artificial bee colony algorithm, and the parameter includes initial solution Number S_N, limiting value limit, maximum cycle Gen_Max lead the quantity N of bee₁, follow the quantity N of bee₂, and N₁=N₂= S_N；N is taken in the present embodiment₁=N₂=S_N=50, limit=10, Gen_Max=500.Initialization population, that is, initial solution X_i(i=1, 2,3...S_N) by the input layer and hidden layer connection weight matrix w of the BP neural network created_ij, hidden layer and output layer connect Weight matrix w_jk, hidden layer threshold value matrix a_j, output layer threshold matrix b_kFour parts composition；The initial solution of each part is one A D=278 or 318 dimensional vectors, D are calculated by formula (1):

D=N_input×N_hidden+N_hidden+N_hidden×N_output+N_output (1)

In formula, N_input、N_hidden、N_outputIt is the mind of the input layer of BP neural network of building, hidden layer, output layer respectively Through first number；

Step S1022: establishing fitness function, and calculates the fitness value in each nectar source；

Fitness function are as follows:

In formula, i=1,2,3...S_N, MSE_iIt indicates i-th of BP network mean square error solved, is when fitness value reaches 1 Optimal state；

Step S1023: it leads bee to carry out neighborhood search according to following formula (3), and calculates the suitable of new explanation according to formula (2) Answer angle value；If the fitness value of new explanation is bigger than the fitness value of old solution, otherwise new and old solution adds the stagnation number of old solution 1；

V_ij=X_ij+rand(-1,1)(X_ij-X_kj) (3)

In formula, i=1,2,3...S_N, S_NFor population number, j=1,2,3 ... D, D are the dimension of solution, V_ijIt is i-th after search J-th of component value of solution, X_ijFor j-th of component value for searching for preceding i-th of solution, X_kjJ-th of component for the solution being randomly generated Value, k ∈ { 1,2,3...S_N, and k ≠ i；If V_ijIt has been more than the maximum value allowed, then has been converted to boundary value according to the following formula:

In formula, lowerbound is value lower bound, and upperbound is the value upper bound；

Step S1024: it follows bee according to formula (5) using roulette method choice nectar source, and nectar source is carried out with formula (3) It updates；If the fitness value of new explanation is bigger than the fitness value of old solution, otherwise the stagnation number of old solution is added 1 by new and old solution；

In formula, i=1,2,3...S_N, S_NFor population number, X_iIt is solved for i-th, P_iThe select probability solved for i-th, f (X_i) be The fitness value of i-th of solution；

Step S1025: judging whether the stagnation number of solution is greater than limiting value 10, if so, search bee carries out according to formula (6) Global search and the fitness value for calculating new explanation；If it is not, thening follow the steps S1026；

In formula, i=1,2,3...S_N, S_NFor population number, j=1,2,3 ... D, D are the dimension of solution, X_i ^jIt is i-th after search J-th of component value of solution,For the minimum value of j-th of component in population,For the maximum value of j-th of component in population, Rand (0,1) is the random number in (0,1) range；

Step S1026: judge whether to reach maximum cycle 500；If so, optimal initial weight and threshold value are exported, Execute step S103；If it is not, then the number of iterations adds 1, return step S1023.

Step S103: it using the obtained optimal initial weight of above-mentioned steps S102 and Threshold-training BP neural network, obtains Improved BP for feature machining operating method decision；

Step S1031: obtained initial weight and threshold value are brought into BP neural network, sequentially input sample data into Row training, the output of hidden layer and output layer is calculated according to formula (7) and (8), calculates each sample training according to formula (9) Mean square error calculates sample overall error according to formula (10)；Then have:

Hidden layer exports H_j:

In formula, f₁It (x) is general hidden layer excitation function, n₁To input the number of plies, w_ijFor input layer and hidden layer connection weight square Battle array, x_iFor input sample, a_jFor hidden layer threshold value matrix, n₂To imply the number of plies；

The output O of output layer_K:

In formula, f₂It (x) is output layer excitation function, n₂To imply the number of plies, H_jFor hidden layer output, w_jkFor hidden layer with it is defeated Layer connection weight matrix out, b_kFor output layer threshold matrix, m is the output number of plies；

The error function of p-th of sample:

In formula, m is the output number of plies,For the kth layer desired output of sample p,It is actually defeated for the kth layer of sample p Out；

The overall error of P sample:

In formula, P is sample total, and p is current input sample, and m is the output number of plies,It is defeated for the kth layer expectation of sample p Out,For the kth layer reality output of sample p；

Step S1032: judging whether to meet error requirements or reach frequency of training, if satisfied, then training finishes, is used In the improved BP of process operation method decision, step S104 is executed；If not satisfied, then according to formula (11) and (12) The anti-pass error for calculating each layer executes step S1031 according to the anti-pass error update weight and threshold value of each layer；

The anti-pass error of output layer:

In formula,For the anti-pass error of the jth layer of the kth layer and hidden layer of sample p output layer,It is the of sample p K layers of desired output,For the kth layer reality output of sample p, m is the output number of plies；

The anti-pass error of hidden layer:

In formula,For the jth layer of sample p hidden layer and i-th layer of anti-pass error of input layer,For sample p output The anti-pass error of the jth layer of the kth layer and hidden layer of layer, w_jkFor hidden layer and output layer connection weight matrix,It is hidden for this p Jth layer output containing layer.

In the present embodiment, general hidden layer excitation function uses S type tangent function tansig (), and output layer excitation function uses S Type logarithmic function logsig (), training function are traingdx algorithms, and learning rate 0.01 frequency of training 500 times, is trained Error target is 0.001.In MATLAB respectively to traditional BP neural network and improved BP proposed in this paper into Row training, Figure 19 is the error convergence curve of both the above neural metwork training.Traditional BP neural network exists as shown in Figure 19 Again without reaching permissible accuracy after 500 training, and improved BP when training 202 times just Training mission is completed, the precision of fast convergence rate, solution is high.

Step S104: it will be input in improved BP after the machining information normalized of part, pass through improvement The nested reasoning of BP neural network exports the non-linear processing scheme of STEP-NC manufacturing feature；

According to the process data of the normalization rule process part of proposition, and the input vector as neural network is input to In neural network.Normalization rule is as follows:

(1) characteristic type (x₁):

Cavity feature type, slot class only have a kind of feature, i.e. x₁=1.0.

Type chamber characteristic type, if closed type type chamber, then x₁=0.0；If open type type chamber, then x₁=1.0.

Step-feature type, step only have a kind of feature, i.e. x₁=1.0.

Hole characteristic type, if unthreaded hole, then x₁=0.0；If threaded hole, then x₁=1.0.

(2) feature limits size (x₂):

The restriction size of slot, if slot maximum cross section area is 1mm²~500mm², then x₂=0.0；If slot maximum cross section Area is 500mm²~4000mm², then x₂=0.5；If slot maximum cross section area is 4000mm²~30000mm², then x₂= 1.0。

The restriction size of type chamber, if type chamber maximum cross section area is 1mm²~500mm², then x₂=0.0；If type chamber is maximum Cross-sectional area is 500mm²~4000mm², then x₂=0.5；If type chamber the maximum cross-section area 4000mm²~30000mm², then x₂=1.0.

The restriction size in hole, if aperture size is 1mm~30mm, x₂=0.0；If aperture size is 30mm~80mm, Then x₂=0.5；If aperture size is 80mm~200mm, x₂=1.0.

The restriction size of step, if step maximum cross section area is 1mm²~500mm², then x₂=0.0；If step is maximum Cross-sectional area is 500mm²~4000mm², then x₂=0.5；If step maximum cross section area is 4000mm²~30000mm², Then x₂=1.0.

(3) blank type (x₃):

If blank is casting, x₃=0.0；If blank is forging, x₃=0.33；If blank is profile, x₃= 0.67；If blank is weldment, x₃=1.0.

(4) material type (x₄):

If material is cast iron, x₄=0.0；If material is carbon steel, x₄=0.5；If material is steel alloy, x₄= 1.0。

(5) surface roughness (x₅):

Various figuratrix roughness 0.05~12.5 are divided into 9 grades, enable x₅=(log₂12.5-log₂Ra)/ (log₂12.5-log₂0.05), corresponding value respectively 12.5 (0.0), 6.3 (0.12), 3.2 (0.25), 1.6 (0.37), 0.8 (0.50), 0.4 (0.62), 0.2 (0.75), 0.1 (0.87), 0.05 (1.0).

(6) heat treatment method (x₆):

If desired it is heat-treated, then x₆=0.0；If not needing to be heat-treated, x₆=1.0.

(7) grade of tolerance (x₇):

The grade of tolerance of feature is divided into IT3~IT13 totally 11 grades, if feature tolerances grade is IT, x₇=(13- IT)/10, corresponding to be worth respectively IT3 (1.0), IT4 (0.9), IT5 (0.8), IT6 (0.7), IT7 (0.6), IT8 (0.5), IT9 (0.4), IT10 (0.3), IT11 (0.2), IT12 (0.1), IT13 (0.0).

(8) batch size (x₈):

It is produced if small lot, then x₈=0.0；If medium quantity batch processing, then x₈=0.5；If mass production, then x₈ =1.0.

The normalized input vector of the present embodiment part is as shown in table 1:

The input of the improved BP neural network of table 1

Step S1041: the process operation method and its selecting priority coefficient of final step are exported according to input vector；

Step S1042: obtaining intermediate features according to machining allowance, and judge whether intermediate features correspond to blank, if so, Processing scheme spanning tree is exported, traverses feature machining schemes generation tree from back to front, the selection of each process operation method is excellent First weight coefficient is multiplied, and obtains the non-linear processing scheme of each feature and its select probability；BP nerve net is improved if it is not, then reusing Network obtains the process operation method and its selecting priority coefficient of previous step, and so on until intermediate features are characterized blank When output processing scheme spanning tree and export the non-linear processing scheme of STEP-NC manufacturing feature.The present embodiment is through nested reasoning After obtain processing scheme tree as shown in Figure 20~Figure 24, the non-linear processing scheme and its select probability such as table that the present embodiment obtains Shown in 2~table 6:

2 step-feature 1 of table, 5 non-linear processing schemes and its select probability

3 cavity feature 2 of table, 3 non-linear processing schemes and its select probability

The non-linear processing scheme of 4 closed type type chamber feature of table 4 and its select probability

5 open type type chamber feature 6 of table, 7 non-linear processing schemes and its select probability

The non-linear processing scheme of 6 hole characteristic of table 8 and its select probability

Processing scheme of the highest processing scheme of probability as feature is chosen, remaining processing scheme gives over to alternatively.Therefore The processing scheme that each feature can finally be obtained is as shown in table 7:

The processing scheme of each feature of table 7

Step S2: according to other, first face metapore after after essence, the former head after first thick times, first benchmark, it is first small after big, minimum indexing, The medium Working-step sequencing rule of cutter collection, uses double-strand table algorithm to be ranked up for all tooling steps in above-mentioned processing scheme, Obtain several reasonable tooling step sequences；It specifically includes:

Step S201: as shown in Fig. 2, tooling step sequence any in above-mentioned non-linear processing scheme deposit one is two-way In chained list, it is denoted as LL；

Step S202: as shown in figure 3, being picked out in doubly linked list LL between other work steps without any relation constraint Tooling step, while their positions in chained list are recorded, and the above-mentioned tooling step without any relation constraint is successively protected There are in a new doubly linked list LL1, remaining tooling step sequence then forms another new doubly linked list LL2；If plus Without mutual unallied work step in work work step sequence, then LL chained list is directly replicated；

Step S203: using the tooling step of tail portion as current traversal point, doubly linked list LL2 is traversed；Double All work steps that should be come after current traversal point according to work step the constraint relationship are found out into chained list LL2, by above-mentioned tooling step It takes out and forms sub- doubly linked list LL21 according to its original sequence；After doubly linked list LL21 is inserted into current traversal point, New doubly linked list LL2 is regenerated, current traversal point is shifted along at the tooling step of the tail portion new doubly linked list LL2, right New traversal point similarly uses the above method to be handled, and operates the tooling step satisfaction until doubly linked list most end repeatedly Until all work step the constraint relationships.

Step S204: current traversal point is moved at its previous tooling step, as new traversal point, processing Method is with step S203, until tooling step all in doubly linked list has all been handled, conversion process terminates；In the present embodiment, As shown in figure 3, work step 5,12 and 2, which belongs between other work steps, does not have correlative work step.It is now assumed that work step 3,9 and 1 is pressed After work step 6 should be come according to the constraint relationship, so work step 3,9 and 1 takes out one sub- doubly linked list LL21 of composition, and place it in After work step 6, new doubly linked list LL2 is regenerated.A pointer will currently be traversed and be directed toward work step 1, using above-mentioned identical side Method carries out constraint judgement to operation and processing, concrete operations are as shown in Figure 4.

Step S205: by be stored in above-mentioned steps S202 in doubly linked list LL1 between other work steps without any pass The tooling step of system's constraint is filled out again according to their original positions to be returned in chained list, and entire conversion process is completed；Such as Fig. 5 institute Show, after work step all in new doubly linked list LL2 has traversed, work step 5,12 and 2 is relay according to their original positions It returns in the new doubly linked list ultimately produced, the new doubly linked list LL finally obtained.

For above-mentioned part machining information, 576 reasonable tooling step sequences are obtained using double-strand table algorithm.

Step S3: being that the matching of each work step is corresponding by rule-based operation resource matching algorithm and artificial bee colony algorithm Operation resource, and determine optimal tooling step sequence and working process parameter；It specifically includes:

Step S301: being that each work step matches corresponding operation resource by rule-based operation resource matching algorithm； Specific algorithm include the matching of rule-based machining tool, the matching of rule-based process tool and machining tool and cutter it Between matching, algorithm flow chart is as shown in Figure 6 and Figure 7；

A, rule-based machining tool matching algorithm is as follows:

A1, machine-tool database, if k=1, matching machine tool M are called_k；

A2, successively judgement " range of work of blank be less than each axis of lathe range? ", " blank weight be less than lathe Maximum capacity? ", " speed of mainshaft needed for part be less than lathe maximum (top) speed? ", " machine finish meets part Requirement on machining accuracy? ", " lathe miscellaneous function meets part processing request? " if all meeting the requirements, by lathe M_kDeposit In the candidate lathe set of workpiece, A3 is executed；If there is a condition to be unsatisfactory for requiring, A4 is executed；

A3, judge whether k is less than T.If so, k=k+1, returns to A2；If it is not, then exporting work step and its candidate lathe Set relation；

A4, judge whether k is less than T.If so, k=k+1, returns to A2；If it is not, then executing A5；

A5, judge whether candidate lathe set is empty.If so, it fails to match for lathe；If it is not, then exporting work step and its time The set relation of lathe is selected, matching machine tool terminates.

Lathe matching result and its parameter are as shown in table 8:

The matched machining tool specifying information of table 8

B, the matched specific algorithm of rule-based process tool is as follows:

B1, cutter database is called, obtains the cutter set T that can process the type feature_nIf n=1；

B2, cutter T is obtained_n；

B3, successively judgement " tool dimension be less than characteristic size? ", " hardness of the cutter material hardness than processing part It is high? ", " precision and surface roughness of part are met the requirements after tool sharpening? ", " surplus after Tool in Cutting meet work step spy Fixed machining allowance range? " if all meeting the requirements, by cutter T_nIt is stored in the candidate cutter set of workpiece, executes B4； If there is a condition to be unsatisfactory for requiring, B5 is executed；

B4, judge whether n is less than N.If so, n=n+1, returns to B2；If it is not, then exporting work step and its candidate cutter Set relation；

B5, judge whether n is less than N.If so, n=n+1, returns to B2；If it is not, then executing B6；

B6, judge whether candidate cutter set is empty.If so, it fails to match for cutter；If it is not, then exporting work step and its time The set relation of cutter is selected, matching cutter terminates.

Cutter matching result and its parameter are as shown in table 9:

The matched process tool specifying information of 9 cutter of table (unit: mm)

C, the matching between machining tool and cutter, by comparing all available process tools for being selected with selected The cutter that machining tool is equipped with, for the corresponding candidate process tool of machining tool matching.Finally, lathe and cutter matching result and Its parameter is as shown in table 10:

10 lathe of table and the matched process tool specifying information of cutter (unit: mm)

Step S302: optimal tooling step sequence is determined by artificial bee colony algorithm；Specific algorithm is as follows:

Step S3021: initializing the parameter of artificial bee colony algorithm, and the parameter includes the number S of initial solution_N, limiting value Limit, maximum cycle Gen_Max lead the quantity N of bee₁, follow the quantity N of bee₂, and N₁=N₂=S_N；In the present embodiment N₁=N₂=S_N=50, limit=10, Gen_Max=500.

Step S3022: initial nectar source is chosen from step S2 and it is encoded；The coding of each work step is by manufacturing feature generation Code, process operation code, process tool code, four part of direction of feed code composition, tooling step sequence and coding are such as Fig. 8 institute Show；

Step S3023: establishing fitness function, and calculates the fitness value in each nectar source；

Using cutter changing total time between work step as optimization aim, therefore objective function can be expressed as follows:

G (x)=Min (f₁,f₂,f₃,......f_m) (13)

In formula, m is the quantity in nectar source；f_iFor the corresponding tool change time in i-th of nectar source, i.e. i-th tooling step sequence institute Corresponding tool change time.And the tool change time in each nectar source is calculated by following equalities:

In formula, i=1,2,3 ... m, m are the quantity in nectar source；J=1, the work step that 2,3 ... n-1, n include by each nectar source Number；t₁The cutter changing time of [j] between continuous twice work step；T [j] is tool code corresponding to j-th of work step, right Above formula has following equation to set up:

In order to guarantee that the optimization direction of objective function corresponds to the direction that fitness value increases, fitness function and mesh are established The mapping relations of scalar functions, the fitness value after guaranteeing mapping is non-negative.Since above-mentioned objective function is to belong to minimum to ask Topic, therefore have mapping relations below for fitness function f (x) and objective function g (x):

In formula, c_MaxIt can be with relatively large number for one.

Step S3024: leading peak to carry out local search according to cross and variation strategy, and calculate the fitness value in new nectar source, If the fitness value of new explanation is bigger than the fitness value of old solution, otherwise the stagnation number of old solution is added 1 by new and old solution；

D, Crossover Strategy process is as follows:

As shown in figure 9, arbitrarily choosing two work step sequences in population, intersection behaviour is carried out by the way of two-point crossover Make.Assuming that two work step sequences are respectively original process route 1 and original process route 2, the specific steps are as follows:

D1, two crosspoints are randomly selected.

D2, the element at 1 crosspoint both ends of original process route is copied in new process route same position.

D3, other elements obtain new process by the copy orderly to new process route rest position in original process route 2 Route.

E, Mutation Strategy process is as follows:

As shown in Figure 10, Mutation Strategy uses two o'clock exchange process, i.e., two operations of random selection in a process route It swaps.If occurring invalid work step sequence after mutation operation, it is necessary to by work step validation algorithm by invalid work step Sequence becomes effective work step sequence.

Step S3025: following bee according to roulette selection nectar source, and uses and the friendship for leading bee identical in step S3024 Fork Mutation Strategy carries out local search and calculates new nectar source fitness value, if the fitness value of new explanation is than the fitness value of old solution Greatly, then new and old solution, otherwise adds 1 for the stagnation number of old solution

Step S3026: judge to stagnate whether number is greater than limiting value 10, if so, thening follow the steps S3027；If it is not, then holding Row step S3028；

Step S3027: investigation bee generates a process route at random and carries out global search, to the process route generated at random Reasonableness check is carried out, if there is unreasonable route, algorithm is rationalized using work step and is translated into reasonable work step sequence Column, and calculate new nectar source fitness value；

Step S3028: judging whether to reach maximum cycle 500, if so, the tooling step sequence that output is optimal, If it is not, then the number of iterations adds 1, return step S3024；

Finally, as shown in figure 25, obtaining 10 optimal tooling step sequences.

Step S303: optimal working process parameter is determined by artificial bee colony algorithm；Specific algorithm is as follows:

The present embodiment carries out working process parameter optimization by taking first tooling step sequence in Figure 25 as an example.The technique road Tooling step and its corresponding process tool information in line is as shown in table 11:

11 tooling step of table and its corresponding process tool information

It is as follows that processing conditions is set:

Work step 11 (bottom surface and side of step 1 are rough milled): processing request: machining allowance 1mm, 6.3 μm of surface roughness；Machine Bed: XK5042A CNC milling machine, lathe maximum power P_Max=11kw, lathe permissible revolution: N_Min=18r/min, N_Max= 1400r/min；Cutter: Morse taper-shank end-milling cutter；Technical parameter: t_ct=58s, t_o=62s, l=240mm, h=18mm, D= 18mm, Z=3, a_p=2mm, a_e=18mm, K_t=0.2, a=1.36, b=3.03, c=2.15, d=0.65, e=0.15, f= 0.2, c₀=50 yuan, c_t=14.57 yuan, K_F=0.25, m=1.01, n=1.09, p=1.1, u=1.3, w=0.2, K_q= 0.15, K_p=0.2, K_s=0.25.

Work step 3 (bottom surface and side of slot 2,3 are rough milled): processing request: machining allowance 1mm, 6.3 μm of surface roughness；Machine Bed: XK5042A CNC milling machine, lathe maximum power P_Max=11kw, lathe permissible revolution: N_Min=18r/min, N_Max= 1400r/min；Cutter: Morse taper-shank end-milling cutter；Technical parameter: t_ct=58s, t_o=62s, l=120mm, h=18mm, D= 18mm, Z=3, a_p=2mm, a_e=9mm, K_t=0.2, a=1.36, b=3.03, c=2.15, d=0.65, e=0.15, f= 0.2, c₀=50 yuan, c_t=14.57 yuan, K_F=0.25, m=1.01, n=1.09, p=1.1, u=1.3, w=0.2, K_q= 0.15, K_p=0.2, K_s=0.25.

Work step 5 (bottom surface and side of closed type type chamber 4 are rough milled): processing request: machining allowance 1mm, surface roughness 6.3 μm；Lathe: XK5042A CNC milling machine, lathe maximum power P_Max=11kw, lathe permissible revolution: N_Min=18r/min, N_Max= 1400r/min；Cutter: Morse taper-shank end-milling cutter；Technical parameter: t_ct=58s, t_o=62s, l=360mm, h=58mm, D= 18mm, Z=3, a_p=2mm, a_e=18mm, K_t=0.2, a=1.36, b=3.03, c=2.15, d=0.65, e=0.15, f= 0.2, c₀=50 yuan, c_t=14.57 yuan, K_F=0.25, m=1.01, n=1.09, p=1.1, u=1.3, w=0.2, K_q= 0.15, K_p=0.2, K_s=0.25.

Work step 7 (bottom surface and side of open type type chamber 6 are rough milled): processing request: machining allowance 1mm, surface roughness 6.3 μm；Lathe: XK5042A CNC milling machine, lathe maximum power P_Max=11kw, lathe permissible revolution: N_Min=18r/min, N_Max= 1400r/min；Cutter: Morse taper-shank end-milling cutter；Technical parameter: t_ct=58s, t_o=62s, l=9mm, h=18mm, D= 18mm, Z=3, a_p=2mm, a_e=9mm, K_t=0.2, a=1.36, b=3.03, c=2.15, d=0.65, e=0.15, f= 0.2, c₀=50 yuan, c_t=14.57 yuan, K_F=0.25, m=1.01, n=1.09, p=1.1, u=1.3, w=0.2, K_q= 0.15, K_p=0.2, K_s=0.25.

Work step 9 (bottom surface and side of open type type chamber 7 are rough milled): processing request: machining allowance 1mm, surface roughness 6.3 μm；Lathe: XK5042A CNC milling machine, lathe maximum power P_Max=11kw, lathe permissible revolution: N_Min=18r/min, N_Max= 1400r/min；Cutter: Morse taper-shank end-milling cutter；Technical parameter: t_ct=58s, t_o=62s, l=9mm, h=14mm, D= 18mm, Z=3, a_p=2mm, a_e=9mm, K_t=0.2, a=1.36, b=3.03, c=2.15, d=0.65, e=0.15, f= 0.2, c₀=50 yuan, c_t=14.57 yuan, K_F=0.25, m=1.01, n=1.09, p=1.1, u=1.3, w=0.2, K_q= 0.15, K_p=0.2, K_s=0.25.

Work step 13 (bottom surface and side of step 5 are rough milled): processing request: machining allowance 1mm, 6.3 μm of surface roughness；Machine Bed: XK5042A CNC milling machine, lathe maximum power P_Max=11kw, lathe permissible revolution: N_Min=18r/min, N_Max= 1400r/min；Cutter: Morse taper-shank end-milling cutter；Technical parameter: t_ct=58s, t_o=62s, l=240mm, h=18mm, D= 18mm, Z=3, a_p=2mm, a_e=18mm, K_t=0.2, a=1.36, b=3.03, c=2.15, d=0.65, e=0.15, f= 0.2, c₀=50 yuan, c_t=14.57 yuan, K_F=0.25, m=1.01, n=1.09, p=1.1, u=1.3, w=0.2, K_q= 0.15, K_p=0.2, K_s=0.25.

Work step 12 (bottom surface of step 1 and side finish-milling): processing request: machining allowance 0mm, 3.2 μm of surface roughness；Machine Bed: XK5042A CNC milling machine, lathe maximum power P_Max=11kw, lathe permissible revolution: N_Min=18r/min, N_Max= 1400r/min；Cutter: straight shank end mill 1；Technical parameter: t_ct=58s, t_o=62s, l=240mm, h=1mm, D=10mm, Z =2, a_p=1mm, a_e=20mm, K_t=0.2, a=1.36, b=3.03, c=2.15, d=0.65, f=0.2, c₀=50 yuan, c_t =3 yuan, K_F=0.25, m=1.01, n=1.09, p=1.1, u=1.3, w=0.2, K_q=0.15, K_p=0.2, K_s=0.25.

Work step 4 (bottom surface of slot 2,3 and side finish-milling): processing request: machining allowance 0mm, 3.2 μm of surface roughness；Machine Bed: XK5042A CNC milling machine, lathe maximum power P_Max=11kw, lathe permissible revolution: N_Min=18r/min, N_Max= 1400r/min；Cutter: straight shank end mill 1；Technical parameter: t_ct=58s, t_o=62s, l=120mm, h=1mm, D=10mm, Z =2, a_p=1mm, a_e=10mm, K_t=0.2, a=1.36, b=3.03, c=2.15, d=0.65, e=0.15, f=0.2, c₀ =50 yuan, c_t=3 yuan, K_F=0.25, m=1.01, n=1.09, p=1.1, u=1.3, w=0.2, K_q=0.15, K_p=0.2, K_s=0.25.

Work step 6 (bottom surface of closed type type chamber 4 and side finish-milling): processing request: machining allowance 0mm, surface roughness 3.2 μm；Lathe: XK5042A CNC milling machine, lathe maximum power P_Max=11kw, lathe permissible revolution: N_Min=18r/min, N_Max= 1400r/min；Cutter: straight shank end mill 1；Technical parameter: t_ct=58s, t_o=62s, l=360mm, h=1mm, D=10mm, Z =2, a_p=1mm, a_e=10mm, K_t=0.2, a=1.36, b=3.03, c=2.15, d=0.65, e=0.15, f=0.2, c₀ =50 yuan, c_t=3 yuan, K_F=0.25, m=1.01, n=1.09, p=1.1, u=1.3, w=0.2, K_q=0.15, K_p=0.2, K_s=0.25.

Work step 14 (bottom surface of step 5 and side finish-milling): processing request: machining allowance 0mm, 3.2 μm of surface roughness；Machine Bed: XK5042A CNC milling machine, lathe maximum power P_Max=11kw, lathe permissible revolution: N_Min=18r/min, N_Max= 1400r/min；Cutter: straight shank end mill 1；Technical parameter: t_ct=58s, t_o=62s, l=240mm, h=1mm, D=10mm, Z =2, a_p=1mm, a_e=20mm, K_t=0.2, a=1.36, b=3.03, c=2.15, d=0.65, f=0.2, c₀=50 yuan, c_t =3 yuan, K_F=0.25, m=1.01, n=1.09, p=1.1, u=1.3, w=0.2, K_q=0.15, K_p=0.2, K_s=0.25.

Work step 8 (bottom surface of open type type chamber 6 and side finish-milling): processing request: machining allowance 0mm, surface roughness 3.2 μm；Lathe: XK5042A CNC milling machine, lathe maximum power P_Max=11kw, lathe permissible revolution: N_Min=18r/min, N_Max= 1400r/min；Cutter: straight shank end mill 1；Technical parameter: t_ct=58s, t_o=62s, l=50mm, h=1mm, D=10mm, Z =2, a_p=1mm, a_e=10mm, K_t=0.2, a=1.36, b=3.03, c=2.15, d=0.65, e=0.15, f=0.2, c₀ =50 yuan, c_t=3 yuan, K_F=0.25, m=1.01, n=1.09, p=1.1, u=1.3, w=0.2, K_q=0.15, K_p=0.2, K_s=0.25.

Work step 10 (bottom surface of open type type chamber 7 and side finish-milling): processing request: machining allowance 0mm, surface roughness 3.2μm；Lathe: XK5042A CNC milling machine, lathe maximum power P_Max=11kw, lathe permissible revolution: N_Min=18r/min, N_Max=1400r/min；Cutter: straight shank end mill 1；Technical parameter: t_ct=58s, t_o=62s, l=50mm, h=1mm, D= 10mm, Z=2, a_p=1mm, a_e=10mm, K_t=0.2, a=1.36, b=3.03, c=2.15, d=0.65, e=0.15, f= 0.2, c₀=50 yuan, c_t=3 yuan, K_F=0.25, m=1.01, n=1.09, p=1.1, u=1.3, w=0.2, K_q=0.15, K_p =0.2, K_s=0.25.

Work step 1 (the common brill in hole 8): processing request: machining allowance 0.1mm (unilateral), 3.2 μm of surface roughness；Lathe: XK5042A CNC milling machine, lathe maximum power P_Max=11kw, lathe permissible revolution: N_Min=18r/min, N_Max=1400r/ min；Cutter: Morse taper shank twist drill；Technical parameter: t_ct=58s, t_o=62s, l=30mm, h=19.8mm, D=19.8mm, Z=2, a_p=9.25mm, a_e=30mm, K_t=0.2, a=1.36, b=3.03, c=2.15, d=0.65, e=0.15, f= 0.2, c₀=50 yuan, c_t=14.57 yuan, K_F=0.25, m=1.01, n=1.09, p=1.1, u=1.3, w=0.2, K_q= 0.15, K_p=0.2, K_s=0.25.

Work step 2 (fraising in hole 8): processing request: machining allowance 0mm (unilateral), 0.8 μm of surface roughness；Lathe: XK5042A CNC milling machine, lathe maximum power P_Max=11kw, lathe permissible revolution: N_Min=18r/min, N_Max=1400r/ min；Cutter: Morse taper shank machine reamer；Technical parameter: t_ct=58s, t_o=62s, l=30mm, h=0.1mm, D=20mm, Z =8, a_p=0.1mm, a_e=30mm, K_t=0.2, a=1.36, b=3.03, c=2.15, d=0.65, e=0.15, f=0.2, c₀=50 yuan, c_t=14.57 yuan, K_F=0.25, m=1.01, n=1.09, p=1.1, u=1.3, w=0.2, K_q=0.15, K_p =0.2, K_s=0.25.

Step S3031: multi-goal optimizing function model is established；It specifically includes:

F, determine variable: variable is feed engagement f_ZWith Milling Speed V；

G, it establishes multi-goal optimizing function: establishing multi-goal optimizing function using process time and processing cost as target；

G1: process time objective optimization function is established, when the process time includes cutting time, tool change time and auxiliary Between；The objective function of process time are as follows:

In formula, t is process time, t_mFor the cutting time, T is tool life, i.e. cutter life, t_m/ T is number of changing knife, t_ctTo change time consumed by a knife, t_oFor other unproductive times in addition to tool change time；

The calculation formula of cutting time and tool life is as follows:

In formula, l is length of cut, and h is thickness of cutting, and D is cutter diameter, and Z is the cutter number of teeth, f_zFor feed engagement, v For cutting speed, a_pFor back engagement of the cutting edge, a_eFor cutting width, K_t, a, b, c, d, e and f be empirical index number；

Formula (18) and (19), which are substituted into formula (17), to be obtained:

G2: establishing processing cost objective optimization function, and the processing cost includes time cost and the cost of charp tool；It is processed into This objective function are as follows:

In formula, c is processing cost, c₀For unit time cost, c_tFor cutter material cost；

Formula (18), (19) and (20) substitution formula (21) can be obtained:

G3: establishing process time and processing cost multi-goal optimizing function, if f₁(f_z, v) and=t, then have:

If f₂(f_z, v) and=c, then have:

The different demands for meeting different user, multiple objective function are adjusted by introducing weighting coefficient λ are as follows:

F(f_z, v) and=λ₁f₁(f_z,v)+λ₂f₂(f_z,v) (25)

In formula, λ₁And λ₂For weighting coefficient, and λ₁+λ₂=1,0≤λ₁≤ 1,0≤λ₂≤1；

Formula (23) and formula (24), which are substituted into formula (25), to be obtained:

H, constraint condition is determined: according to the requirement of machining tool, process tool and processing quality, the constraint item of optimization aim Part specifically includes maximum feeding cutting force constraint, maximum principal axis torque constraint, maximum working power constraint, workpiece surface quality about Beam, cutting speed constraint and feed engagement constraint；

H1, maximum feeding cutting force constraint:

In formula, F_fMaxFor maximum allowable cutting force, K_F, m, n, p and w be cutting force empirical coefficient；

H2, maximum principal axis torque constraint:

In formula, T_qMaxFor maximum permissible torque, K_qFor torque empirical coefficient；

H3, maximum working power constraint:

In formula, P_MaxFor maximum allowable power, K_pFor power empirical coefficient；

H4, workpiece surface quality constraint:

In formula,For maximum allowable surface roughness, K_sFor surface roughness empirical coefficient；

H5, cutting speed constraint:

In formula, N_MinAnd N_MaxFor the lathe minimum and maximum speed of mainshaft, v_MinAnd v_MaxFor minimum and maximum cutting speed；

H6, feed engagement constraint:

In formula,WithFor lathe minimum and maximum feed speed,WithFor the feeding of minimum and maximum per tooth Amount；

Step S3032: the technological parameter intelligent optimization based on artificial bee colony algorithm；Specific algorithm is as follows:

I1, according to the processing conditions given, each machined parameters in initialization function model；

I2, the parameter and initialization population for initializing artificial bee colony algorithm；The parameter of the artificial bee colony algorithm includes just The number S for the solution that begins_N, limiting value limit, maximum cycle Gen_Max lead the quantity N of bee₁, follow the quantity N of bee₂, and N₁=N₂=S_N；Initial nectar source, that is, initial solution X_i(i=1,2,3...S_N) by feed engagement f_ZIt is formed with Milling Speed V；This reality Apply N in example₁=N₂=S_N=50, limit value limit=10, maximum number of iterations Gen_Max=500；

I3, fitness function is established, and calculates the fitness value in each nectar source；

Fitness function are as follows:

c_MaxIt is a biggish number, there will be constrained optimization problem to be converted into solution without constraint using compound penalty function Optimization problem, mixed penalty function such as following formula indicate；

In formula, F (x) is former objective function, h_iIt (x) is equality constraint, g_j(x) be inequality constraints condition punishment , k is the number of equality constraint, and l is the number of inequality constraints, and M is penalty factor and M⁰<M¹<M²...→∞；

The mathematical model that the cutting parameter of above-mentioned foundation optimizes is brought into formula (34) to obtain:

It brings formula (35) into formula (33) again and obtains fitness function；

I4, it leads bee to carry out neighborhood search, and carries out nectar source update；

It leads bee to carry out neighborhood search according to formula (36), and calculates the fitness value of new explanation according to formula (33)；If The fitness value of new explanation is bigger than the fitness value of old solution, then new and old solution, and the stagnation number of old solution is otherwise added 1；

V_ij=X_ij+rand(-1,1)(X_ij-X_kj) (36)

In formula, i=1,2,3...S_N, S_NFor population number, j=1,2,3 ... D, D are the dimension of solution, V_ijIt is i-th after search J-th of component value of solution, X_ijFor j-th of component value for searching for preceding i-th of solution, X_kjJ-th of component for the solution being randomly generated Value, k ∈ { 1,2,3...S_N, and k ≠ i；If V_ijIt has been more than the maximum value allowed, then has been converted to boundary value according to the following formula:

In formula, lowerbound is value lower bound, and upperbound refers to as the value upper bound；

I5, bee is followed to be updated according to roulette method choice nectar source, and to nectar source；

It follows bee according to formula (38) using roulette method choice nectar source, and nectar source is updated with formula (33)；Such as The fitness value of fruit new explanation is bigger than the fitness value of old solution, then new and old solution, and the stagnation number of old solution is otherwise added 1；

I=1 in formula, 2,3...S_N S_NFor population number, X_iIt is solved for i-th, P_iThe select probability solved for i-th, f (X_i) be The fitness value of i-th of solution；

I6, judge to stagnate whether number is greater than limiting value 10, if so, search bee carries out global search simultaneously according to formula (39) Calculate the fitness value of new explanation；If it is not, then executing I7；

In formula, i=1,2,3...S_N, S_NFor population number, j=1,2,3 ... D, D are the dimension of solution, X_i ^jIt is i-th after search J-th of component value of solution,For the minimum value of j-th of component in population,For the maximum value of j-th of component in population, Rand (0,1) is the random number in (0,1) range；

I7, judge whether to reach maximum cycle 500；If so, the technological parameter that output is optimal, if it is not, then iteration Number adds 1, returns to I4.

It is as shown in table 12 that the present embodiment obtains process parameter optimizing result:

12 process parameter optimizing result of table

Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention., rather than its limitations；To the greatest extent Pipe present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that: its according to So be possible to modify the technical solutions described in the foregoing embodiments, or to some or all of the technical features into Row equivalent replacement；And these are modified or replaceed, various embodiments of the present invention technology that it does not separate the essence of the corresponding technical solution The range of scheme.

Claims (5)

1. a kind of intelligent Nonlinear Process Planning Method towards STEP-NC manufacturing feature, which is characterized in that including following step It is rapid:

Step S1: the non-linear processing scheme of STEP-NC manufacturing feature is determined by improved BP；It specifically includes:

Step S101: process operation method decision BP neural network model of the building towards STEP-NC manufacturing feature；

Step S102: utilize artificial bee colony algorithm Optimized BP Neural Network initial weight and threshold value, obtain optimal initial weight and Threshold value；

Step S103: it using the obtained optimal initial weight of above-mentioned steps S102 and Threshold-training BP neural network, is used for The improved BP of feature machining operating method decision；

Step S104: will be input in improved BP after the machining information normalized of part, by improving BP mind Nested reasoning through network exports the non-linear processing scheme of STEP-NC manufacturing feature；

Step S2: according to tooling step principle of ordering, use double-strand table algorithm for all tooling steps in above-mentioned processing scheme It is ranked up, obtains several reasonable tooling step sequences；It specifically includes:

Step S201: tooling step sequence any in above-mentioned non-linear processing scheme is stored in a doubly linked list, LL is denoted as；

Step S202: the tooling step without any relation constraint between other work steps is being picked out in doubly linked list LL, simultaneously Their positions in chained list are recorded, and the above-mentioned tooling step without any relation constraint is successively stored in a new pair Into chained list LL1, remaining tooling step sequence then forms another new doubly linked list LL2；If not having in tooling step sequence There is mutual unallied work step, then directly replicates LL chained list；

Step S203: using the tooling step of tail portion as current traversal point, doubly linked list LL2 is traversed；In Two-way Chain All work steps that should be come after current traversal point according to work step the constraint relationship are found out in table LL2, and above-mentioned tooling step is taken out And sub- doubly linked list LL21 is formed according to its original sequence；After doubly linked list LL21 is inserted into current traversal point, again New doubly linked list LL2 is generated, current traversal point is shifted along at the tooling step of the tail portion new doubly linked list LL2, to new Traversal point is similarly handled using the above method, operates own until the tooling step of doubly linked list most end meets repeatedly Work step the constraint relationship until；

Step S204: current traversal point is moved at its previous tooling step, as new traversal point, processing method With step S203, until tooling step all in doubly linked list has all been handled, conversion process terminates；

Step S205: by be stored in above-mentioned steps S202 in doubly linked list LL1 between other work steps without any relationship about The tooling step of beam is filled out again according to their original positions and is returned in chained list, and entire conversion process is completed；

Step S3: being that the matching of each work step is corresponding by rule-based operation resource matching algorithm and artificial bee colony algorithm adds Wage source, and determine optimal tooling step sequence and working process parameter；It specifically includes:

Step S301: being that each work step matches corresponding operation resource by rule-based operation resource matching algorithm；Specifically Algorithm includes that rule-based machining tool matches, between the matching of rule-based process tool and machining tool and cutter Matching；

Step S302: optimal tooling step sequence is determined by artificial bee colony algorithm；Specific algorithm is as follows:

Step S3021: initializing the parameter of artificial bee colony algorithm, and the parameter includes the number S of initial solution_N, limiting value limit, Maximum cycle Gen_Max leads the quantity N of bee₁, follow the quantity N of bee₂, and N₁=N₂=S_N；

Step S3022: initial nectar source is chosen from step S2 and it is encoded；

Step S3023: establishing fitness function, and calculates the fitness value in each nectar source；

Step S3024: leading peak to carry out local search according to cross and variation strategy, and calculate the fitness value in new nectar source, if The fitness value of new explanation is bigger than the fitness value of old solution, then new and old solution, and the stagnation number of old solution is otherwise added 1；

Step S3025: following bee according to roulette selection nectar source, and using it is identical with step S3024 lead bee intersect change Different strategy carries out local search and calculates new nectar source fitness value, if the fitness value of new explanation is bigger than the fitness value of old solution, Otherwise the stagnation number of old solution is added 1 by then new and old solution；

Step S3026: judge to stagnate whether number is greater than limiting value limit, if so, thening follow the steps S3027；If it is not, then holding Row step S3028；

Step S3027: investigation bee generates a process route at random and carries out global search, carries out to the process route generated at random Reasonableness check rationalizes algorithm using work step and is translated into reasonable work step sequence if there is unreasonable route, and Calculate new nectar source fitness value；

Step S3028: judging whether to reach maximum cycle, if so, the tooling step sequence that output is optimal, if it is not, then The number of iterations adds 1, return step S3024；

Step S303: optimal working process parameter is determined by artificial bee colony algorithm；Specific algorithm is as follows:

Step S3031: multi-goal optimizing function model is established；It specifically includes:

F, determine variable: variable is feed engagement f_ZWith Milling Speed V；

G, it establishes multi-goal optimizing function: establishing multi-goal optimizing function using process time and processing cost as target；

H, determine constraint condition: according to the requirement of machining tool, process tool and processing quality, constraint condition specifically includes maximum Feeding cutting force constraint, maximum principal axis torque constraint, maximum working power constraint, workpiece surface quality constraint, cutting speed are about Beam and feed engagement constraint；

Step S3032: the technological parameter intelligent optimization based on artificial bee colony algorithm；Specific algorithm is as follows:

I1, according to the processing conditions given, each machined parameters in initialization function model；

I2, the parameter and initialization population for initializing artificial bee colony algorithm；The parameter of the artificial bee colony algorithm includes initial solution Number S_N, limiting value limit, maximum cycle Gen_Max lead the quantity N of bee₁, follow the quantity N of bee₂, and N₁=N₂ =S_N；Initial nectar source, that is, initial solution X_i(i=1,2,3...S_N) by feed engagement f_ZIt is formed with Milling Speed V；

I3, fitness function is established, and calculates the fitness value in each nectar source；

I4, it leads bee to carry out neighborhood search, and carries out nectar source update；

I5, bee is followed to be updated according to roulette method choice nectar source, and to nectar source；

I6, judge to stagnate whether number is greater than limiting value limit；If so, search bee carries out global search and calculates new explanation Fitness value；If it is not, then executing I7；

I7, judge whether to reach maximum cycle, if so, the technological parameter that output is optimal, if it is not, then the number of iterations adds 1, Return to I4.

2. the intelligent Nonlinear Process Planning Method according to claim 1 towards STEP-NC manufacturing feature, feature exist In the process for constructing BP neural network model in the step S101 is as follows:

Step S1011: the input number of plies n of neural network is determined according to the factor number for influencing feature machining operating method decision₁；

Step S1012: rule of thumb formula n₂=2n₁+ 1 determines implicit number of plies n₂；

Step S1013: the output number of stories m of neural network is determined according to Milling Process operation data model in STEP-NC standard.

3. the intelligent Nonlinear Process Planning Method according to claim 1 towards STEP-NC manufacturing feature, feature exist In the process in the step S102 using artificial bee colony algorithm Optimized BP Neural Network initial weight and threshold value is as follows:

Step S1021: initializing the parameter and initialization population of artificial bee colony algorithm, and the parameter includes the number of initial solution S_N, limiting value limit, maximum cycle Gen_Max lead the quantity N of bee₁, follow the quantity N of bee₂, and N₁=N₂=S_N； The initialization population, that is, initial solution X_i(i=1,2,3...S_N) connect by the input layer of the BP neural network created with hidden layer Weight matrix w_ij, hidden layer and output layer connection weight matrix w_jk, hidden layer threshold value matrix a_j, output layer threshold matrix b_kFour It is grouped as；

Step S1022: establishing fitness function, and calculates the fitness value in each nectar source；

Step S1023: it leads bee to carry out neighborhood search, and calculates the fitness value of new explanation；If the fitness value of new explanation is than old The fitness value of solution is big, then new and old solution, and the stagnation number of old solution is otherwise added 1；

Step S1024: it follows bee using roulette method choice nectar source, and nectar source is updated；If the fitness of new explanation Value is bigger than the fitness value of old solution, then new and old solution, otherwise adds 1 for the stagnation number of old solution；

Step S1025: judging whether the stagnation number of solution is greater than limiting value limit, if so, search bee carries out global search simultaneously Calculate the fitness value of new explanation；If it is not, thening follow the steps S1026；

Step S1026: judge whether to reach maximum cycle or meet required precision；If so, the optimal initial power of output Value and threshold value execute step S103；If it is not, then the number of iterations adds 1, return step S1023.

4. the intelligent Nonlinear Process Planning Method according to claim 1 towards STEP-NC manufacturing feature, feature exist In the process in the step S103 using obtained optimal initial weight and Threshold-training BP neural network is as follows:

Step S1031: obtained initial weight and threshold value are brought into BP neural network, are sequentially input sample data and are instructed Practice, calculate the output of hidden layer and output layer, and calculates the mean square error and sample overall error of each sample training；

Step S1032: judging whether to meet error requirements or reach frequency of training, if satisfied, then training finishes, obtains for adding The improved BP of work operating method decision executes step S104；If not satisfied, then being missed according to the anti-pass for calculating each layer Difference executes step S1031 according to the anti-pass error update weight and threshold value of each layer.

5. the intelligent Nonlinear Process Planning Method according to claim 1 towards STEP-NC manufacturing feature, feature exist In the nested reasoning process of the improved BP in the step S104 is as follows:

Step S1041: the process operation method and its selecting priority coefficient of final step are exported according to input vector；

Step S1042: obtaining intermediate features according to machining allowance, and judge whether intermediate features correspond to blank, if so, output Processing scheme spanning tree traverses feature machining schemes generation tree, by the selecting priority of each process operation method from back to front Multiplication obtains the non-linear processing scheme of each feature and its select probability；If it is not, improved BP is then reused, Obtain the process operation method and its selecting priority coefficient of previous step, and so on defeated when intermediate features are characterized blank Processing scheme spanning tree and export the non-linear processing scheme of STEP-NC manufacturing feature out.