CN114912346A - Skill planning configuration and workshop scheduling integrated optimization method based on learning ability - Google Patents

Abstract

The invention discloses a skill planning configuration and workshop scheduling integrated optimization method based on learning ability, which comprises the following steps: step 1, establishing an employee skill training decision and scheduling integrated optimization model considering learning capacity heterogeneity; and 2, solving the model established in the step 1 by adopting a genetic simulated annealing algorithm to obtain a final optimal solution, and outputting an optimal target value and an optimal scheduling scheme. Aiming at the workshop scheduling problem with staff heterogeneity, staff training decisions are introduced to plan staff skills in advance, the learning effect of staff in the production process is considered, a workshop scheduling optimization model considering staff training and learning effects is constructed, the completion time, the manufacturing cost and the staff load are taken as optimization targets, a three-layer coded genetic simulation annealing algorithm is adopted to solve, and simulation verification is carried out through a numerical case; the result shows that the staff training can enlarge the staff working range, increase the staff flexibility, reduce the completion time and the staff load and provide a more accurate and excellent scheduling scheme.

Description

Skill planning configuration and workshop scheduling integrated optimization method based on learning ability

Technical Field

The invention relates to a skill planning configuration and workshop scheduling integrated optimization method based on employee learning ability.

Background

In the production scheduling process of an enterprise, not only equipment resources but also worker resources are one of the basic elements. The manufacturing industry in China still commonly adopts a labor-intensive production mode, workers still have the mastery in production, and the industry has the problems of difficult recruitment and retention of workers and faces the difficult problem of wasted labor ^[1] ([1]Sun Baofeng, Ningxin, Zhengzisi, etc. multiple objective flow shop optimization scheduling considering worker load [ J]The university of Jilin journal (engineering edition), 2021,51(03): 900-. The method has great significance for analyzing the worker resources and researching the skill requirement problem of the workers ^[2] ([2]Dongwei, Zhang Mei, Wangshi and an and so on]Higher engineering education, 2018(06): 131-. Staff training is a mode for staff to obtain skills from enterprises, and in a manufacturing enterprise, the ability of operating different equipment or possessing a certain skill can be obtained through staff training, so that staff allocation schemes are increased, more flexible configuration can be performed on workers, and more flexible scheduling is realized.

Staff training has always been the focus of actual production in enterprises, and many scholars have studied it, such as Davis ^[3] Most of the benefit of staff flexibility is thought to be due to staff cross-training ([ 3)]Davis D J,Kher H V,Wagner B J.Influence of workload imbalances on the need for worker flexibility[J].Computers&Industrial Engineering,2009,57(1): 319-329), and the research on the same, under the conditions of high workload and unbalance, the more extensive worker cross training can obviously improve the workshop performance; slomp ^[4] Research has been conducted to identify the need for cross-training employees in a unitized manufacturing environment ([ 4]]Slomp J,Bokhorst J,Molleman E.Cross-training in a cellular manufacturing environment[J].Computers&Industrial Engineering,2005,48(3):609-624.)；Li ^[5] The emotional factor of the staff satisfaction is integrated into the training ([5 ]]Li Q,Tang J.Multi-objective optimal cross-training configuration models for an assembly cell using non-dominated sorting genetic algorithm-II[J]International Journal of Computer Integrated Manufacturing,2012,25(11):1-15.), studies how to determine the most suitable training program; koltai (R) K.K. ^[6] The assembly line balance problem was studied ([6 ]]Koltai T.Formulation of multi-level workforce skill constraints in assembly line balancing models[J]International Federation of Automatic Control,2013,46(9):772-777.) and classifies employee skills into three categories, low skills, high skills, and exclusive skills. Establishing a training model for acquiring the skill of the staff, and showing how to finish a simple assembly line balance model under the condition of skill loss; wang (Wang) ^[7] Skill enrichment of workers through employee training ([ 7)]Wang Y,Tang J.Optimized skill configuration for the seru production system under an uncertain demand[J]Annals of Operations Research,2020(1): 1-21), a robust production system is obtained that can efficiently respond to random demand. The researchers prove that the staff training brings benefits to enterprise production through research, but no scholars consider the combination of workshop scheduling and staff training for research, plan staff skill resources in advance through training and obtain high-quality scheduling results. And because the staff is influenced by the learning effect in the production process of the product, the staff capability can be changed, and therefore a more accurate scheduling result can be obtained by considering the learning effect. Many scholars study the learning effect and apply it to shop scheduling. Such as Wright ^[8] The concept of learning effect curves was first proposed in the analysis of productivity-influencing factors of the aeronautical industry ([8 ]]Wright T P.Factors affecting the cost of airplanes[J]Journal of Aeronoural Sciences,1936,3(4): 122-; subsequently, Biskup ^[9] A location-based learning effect model is proposed ([9 ]]Biskup D.A state-of-the-art review on scheduling with learning effects[J]European Journal of Operational Research,2008,188(2):315- ^[10-12] A logarithmic-linear model, an S-shaped model,Plateau model and Dejong model ^[13] ([10]Mousavi S M,Mahdavi I,Rezaeian J,et al.An efficient bi-objective algorithm to solve re-entrant hybrid flow shop scheduling with learning effect and setup times[J].Operational Research,2018,18(1):123-158.、[11]Wang J.Scheduling jobs with a general learning effect model[J].Applied Mathematical Modelling,2013,37(4):2364-2373.、[12]Cheng T,Wu C,Lee W.Some scheduling problems with sum-of-processing-times-based and job-position-based learning effects[J].Information Sciences,2008,178(11):2476-2487.、[13]Houfenglong, Yechuming, Gunn beautiful, learning effect production scheduling problem based on multi-target firefly membrane algorithm [ J]The journal of system management, 2018,27(04): 704-. Based on the research of the scholars, a plurality of scholars at home and abroad carry out continuous supplementary application on the scholars, such as Jiang and Kuo ^[14,15] The actual time-dependent and job-dependent learning effects are introduced into the single-machine scheduling problem ([14 ]]Jiang Z,Chen F,Kang H.Single-machine scheduling problems with actual time-dependent and job-dependent learning effect[J].European Journal of Operational Research,2013,227(1):76-80.、[15]Kuo W.Single-machine scheduling with an actual time-dependent learning effect by D-L yang and W-H kuo reply[J]Journal of the Operational facility, 2009,60(3): 435-; mosheiov and Sindey ^[16] Location and worker based learning effects are proposed ([16 ]]

Mosheiov G,Sidney J B.Scheduling with general job-dependent learning curves[J]European Journal of Operational Research,2003,147(3): 665-; arbors with beautiful figures etc ^[17] The relationship between equipment resources and worker resources and the learning abilities of employees in a dual resource production process in this long-term dynamic process was studied ([17 ]]Double resource allocation method of lean production manufacturing system (J) considering learning curve]A computer integrated manufacturing system, 2016,22(12): 2800-; hujinchang tea ^[18] On sheetInfluence of employee learning ability in scheduling process of single-process multi-machine ([ 18)]Single-person job shop multi-target scheduling algorithm [ J ] considering learning effect]Computer integrated manufacturing system, 2021,27(05): 1361-; bin ^[19] How the effects of interactive learning among members of a research team affect the scheduling process ([19 ]]Li bin, and Metallum, consider a single-workgroup interruptible task scheduling model of team interactive learning effects [ J]Computer integrated manufacturing system, 2022,28(01): 161-; cao epi and the like ^[20] Taking into account the heterogeneity that exists between employees ([20 ]]Cao Lei Ming, Cao Xia, yellow Xia, based on employee learning behavior, multi-objective flexible workshop scheduling [ J]A computer integrated manufacturing system, 2018,24(08): 2023-. The researchers develop research on learning effect in the field of workshop scheduling, but do not consider the influence of staff training on the workshop scheduling result.

Disclosure of Invention

Based on the current situation, the invention provides a skill planning configuration and workshop scheduling integrated optimization method based on learning ability, which considers that staff training is introduced in the workshop scheduling process, the skill of the staff is increased through training, and the influence of the staff on the workshop scheduling is analyzed by combining the learning effect of the staff.

The invention adopts the following technical scheme:

the skill planning configuration and workshop scheduling integrated optimization method based on learning ability comprises the following steps:

step 1, establishing an employee skill training decision and scheduling integrated optimization model considering learning capacity heterogeneity;

and 2, solving the model established in the step 1 by adopting a genetic simulated annealing algorithm to obtain a final optimal solution, and outputting an optimal target value and an optimal scheduling scheme.

Preferably, in step 1, the staff skill training decision and scheduling integrated optimization model considering the heterogeneity of learning ability is described as follows: the work pieces with N different batches are processed by W workers with different proficiency levels on M devices with different types and operation capabilities at different positions of a workshopProcessing, wherein W is less than or equal to M; when an order comes, N kinds of workpieces are to be processed in a certain scheduling period, and the demand of each kind of workpiece is D _i And the workpiece has N _i The process comprises J processes, wherein each process can be carried out on a plurality of devices;

selecting the completion time of the product, the production cost and the workload balance of the staff as the objective function of the model;

(1) the completion time is minimal: the goal of minimizing the maximum time to completion of all products is a measure

f ₁ ＝min(maxT _is ) (4)

(2) The manufacturing cost is lowest: the cost of equipment, the wage cost of workers and the training cost of staff are divided into three major parts

(3) Staff workload balancing: taking the staff workload balance coefficient as a measurement target

The dimensions of the three targets of product completion time, manufacturing cost and load balancing coefficient in the model are different, so normalization processing is required to be carried out in the solving process to eliminate the dimensions for comparison of target functions, and the following formula is adopted:

wherein f is _max And f _min The maximum value and the minimum value of the objective function are obtained;

combining the above target weights as:

f＝α·f ₁ +β·f ₂ +γ·f ₃ (8)

wherein α, β, and γ are weights of completion time, manufacturing cost, and employee load balancing coefficients, and α + β + γ is 1;

the constraints are as follows:

(1) the process of workpieces of any batch is started at 0 moment when the conditions allow

S _ijskw ≥0 (9)

(2) The process completion time of each batch of workpieces is determined by the start time of the batch of processes and the actual processing time of the batch

E _ijskw ≥S _ijskw +d _i x _ijskw T _ijkw (10)

(3) The completion time of any process of any workpiece is the completion time considering the learning effect

(4) Whether to train constraints, if the worker is assigned to the equipment k, if it has the ability to operate the equipment k, no training is required, if it does not, training is required

(5) Ensuring that the workers w assigned to each equipment in each batch have the ability to operate the equipment

x _ijskw ＝x _ijsk [x _kw +(1-x _kw )y _kw ] (13)

(6) For different processes of the same batch of the same workpiece, the operation of the next process can be carried out only after the previous process is finished

S _ijskw ≥E _i(j-1)slo (14)

(7) The working procedure j of the workpiece i can only select one equipment from the equipment set capable of processing the working procedure for processing, and a worker operating the equipment can only select one operating equipment from the selectable worker set

(8) Total processing time of each worker in production

(9) The uniqueness of the equipment is restricted, one equipment cannot simultaneously process two procedures at a certain time

S _ijskw X _ijskw -E _i'j'sko X _i'j'sko ≥0 (17)

(10) The uniqueness of workers is restricted, one worker can only operate one device and cannot simultaneously operate a plurality of devices

S _ijskw X _ijskw -E _i'j'slw X _i'j'slw ≥0 (18)

(11) Constraint of binary variable

x _ijsk ＝{0,1}，x _ijskw ＝{0,1}，y _wk ＝{0,1} (19)。

Preferably, in step 2, the model is solved by using a genetic simulated annealing algorithm, and the specific steps are as follows:

s1: generating an initialization population using a genetic algorithm;

s2: calculating the fitness value of the population individuals, and enabling the iteration number g to be 0;

s3: obtaining a temporary population through selection, crossing and variant genetic operations by comparing individual fitness values;

s4: performing simulated annealing algorithm operation in the temporary population, determining the initial temperature of the simulated annealing algorithm, generating an initial solution S, and calculating a target function value corresponding to the solution S;

s5: generating a new solution S' through a disturbance strategy, calculating a target function value corresponding to the new solution, judging whether the new solution is accepted or not by using a Metropolis criterion, and switching to S6 if the new solution is accepted, and switching to S4 if the new solution is not accepted;

s6: detecting whether sufficient search is carried out at the temperature, if the sufficient search is carried out, turning to S7, and if the sufficient search is not carried out, turning to S4;

s7: judging whether the simulated annealing termination condition is met, if so, memorizing the current optimal solution in a memory table to be converted into S8 for continuous iteration; if not, the annealing operation is carried out to S4;

s8: judging whether the iteration process reaches the maximum iteration number, if not, turning to S3 to perform a new iteration process; if yes, go to S9;

s9: and comparing all the optimal solutions in the memory table to obtain a final optimal solution, and outputting an optimal target value and an optimal scheduling scheme.

Preferably, step S1 is specifically as follows: initializing population size P, cross probability P _c Probability of variation P _m Annealing initiation temperature T ₀ Cooling rate λ, maximum number of iterations G _max Threshold temperature T _min Chain length L;

at solution demand of D _i The dispatching time of the workpieces is determined by the batch number of each product, and the batched products are dispatched; in the genetic algorithm, a computer cannot directly identify parameters in a model, so that the parameters need to be converted into individuals consisting of a plurality of genes, and the process of converting the parameters into the identifiable individuals is called encoding; chromosomes are divided into three parts to code: a process sequence code based on a batch, an allocation code based on equipment resources, and an allocation code based on worker resources; workpiece procedure processing sequence information, machine equipment resource allocation information and worker resource allocation information; the length of the chromosome is

N denotes the number of workpieces, N _i Indicates that the workpiece i has N _i Procedure, three layers of coding, each layer being of length

Lot-based process order coding: the code consists of several groups of repeated numbers, the first number represents a workpiece, the second number represents a batch, the number of times of occurrence of the repeated numbers represents the number of processes of each workpiece, and the sequence of occurrence of the numbers represents the processing sequence;

encoding based on the allocation of device resources: coding according to the types and batches of the workpieces, wherein the internal codes of the workpieces are the process priority and batch sequence of each workpiece from the workpiece 1 to the workpiece n; each procedure is processed on a plurality of devices, and the coding gene number represents the sequential position of the selected processing device in the selectable machine set;

worker resource based allocation coding: the layer of codes corresponds to equipment resource allocation codes, wherein the numbers represent workers, and the equipment resources are matched with the worker resources through allocation of the workers; the layer code comprises two kinds of information, namely, a worker for operating the second layer equipment is selected; secondly, judging whether the selected worker on the layer needs to add new skills, if the selected worker has the capability of operating the equipment on the second layer, namely x _ijkw 1 and x _kw 1, the employee does not need training y _kw 0 if the selected worker does not have the ability to operate the second level of equipment, i.e. x _ijkw 1 and x _kw If 0, the employee needs training, y _kw 1, generating training costs;

the chromosome decoding process is described as: the process sequence and the batch production sequence of the processes, the machine equipment allocated for the work process of each batch, and the workers are determined step by step starting from the first layer.

Preferably, step S2 is specifically as follows:

evaluating a skill planning and workshop scheduling optimization scheme by using a fitness function, wherein the larger the fitness value is, the more chance is to select the candidate as a parent; in the model, the inverse of a weighted objective function of eliminating dimensions among the objective functions by using normalization processing is compared, and a fitness function is as follows

The genetic algorithm selects a proper number of good individuals as parent breeding filial generations through selection operation, offspring is selected by using a roulette method, the higher the fitness value of the individual is, the higher the survival probability is, and the higher the probability of the good individuals selected as parent breeding filial generations is;the selection mode is as follows: assuming that n individuals exist in the population, the fitness value of the ith individual is F _i Then the probability that the individual is selected is:

preferably, the interleaving operation in step S3 is specifically as follows:

the sequence crossing operators are adopted to respectively cross the working procedure sequence of the workpieces; or respectively carrying out intersection on equipment resource allocation and worker resource allocation by using a multipoint intersection operator.

Preferably, the mutation operation in step S3 is specifically as follows:

carrying out mutation on the sequence of the work piece working procedures by adopting a reverse mutation operator; alternatively, equipment resource allocation and worker resource allocation are mutated using random single point mutations.

Aiming at the workshop scheduling problem with staff heterogeneity, staff training decisions are introduced to plan staff skills in advance, the learning effect of staff in the production process is considered, a workshop scheduling optimization model considering staff training and learning effects is constructed, the completion time, the manufacturing cost and the staff load are taken as optimization targets, a three-layer coded genetic simulation annealing algorithm is adopted to solve, and simulation verification is carried out through a numerical case; the result shows that staff training can expand the working range of staff, increase staff flexibility, reduce completion time and staff load and provide a more accurate and excellent scheduling scheme.

Drawings

FIG. 1 is a flow chart of a genetic annealing algorithm according to the present invention.

FIG. 2 is a chromosome coding map.

Fig. 3 is a cross-plot of the workpiece process sequence using a sequence cross operator (POX).

FIG. 4 is a cross-over diagram for device resource allocation and worker resource allocation, respectively, using a multipoint cross-over operator.

FIG. 5 is a diagram of a variation of a sequence of work processes using a reverse mutation operator.

Fig. 6 is a graph of variation of equipment resource allocation and worker resource allocation using random single point variation.

FIG. 7 is a time chart of a workshop dispatch protocol completion in three methods.

FIG. 8 is a manufacturing cost diagram for a shop scheduling scheme under three methods.

Detailed Description

The preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

The learning process of the staff can be divided into two stages, wherein the first stage is that a training process is carried out to master new skill standards and operation key points through learning so as to meet the training end requirement; and in the second stage, one operation is repeatedly performed on a post, and the skill proficiency is improved by continuously learning and accumulating experiences under the influence of the learning effect.

The staff training is an important way for enterprises to cultivate talents, is a way for the staff to obtain skills from the enterprises, and is a process which can independently work after technical specifications and skill standards meeting the skills are not known to a certain skill. The staff knows the skilled working procedure operation in this stage, realizes the acquisition of skill. The learning effect is a phenomenon that the subsequent production efficiency is continuously improved due to the fact that workers continuously repeat certain operation and experience is continuously accumulated, and the learning effect generally exists in production.

In the production operation process, the learning effect of the staff is influenced by four factors, which are respectively: employee initial competency level, learning ability, number of task operation repetitions, and difficulty of the task. The initial ability and learning ability of the staff are restricted by the staff when the staff officially go on duty, the initial ability is influenced by the training before the staff's duty, the initial ability of the staff with deep knowledge base, strong receptivity and high manual operation ability is stronger when the staff goes on duty, otherwise, the initial ability is weaker; the number of times of work repetition is determined by the order requirement; the task difficulty is determined by the job itself. Dejong in 1957 proposes a learning effect model, which considers the learning effects of machines and workers, is called as Dejong learning effect model, and better accords with the production mode of manual operation or semi-automatic operation in China, and the formula is as follows:

T _ijkw ＝A[F+(1-F)X ^a ] (1)

a＝lg ^l /lg ² (2)

in the formula, T _ijkw The actual time when the process j of the workpiece i is finished by the worker w on the equipment k for X pieces, A is the processing time of the first workpiece, F is an incompressible factor (F is more than or equal to 0 and less than or equal to 1) represents the influence of the automation degree of the production line on the learning effect, the larger the value is, the higher the automation degree is, the lower the influence of the learning effect of the worker is, X represents the processing number of the processes on the processing machine, and a is the learning factor (a is less than or equal to 0) of the worker. In this model, when F is equal to 1, the calculation method is a process actual time when there is no learning effect for the employee, and when F is equal to 0, the model is a conventional learning effect model.

Considering the initial capability heterogeneity and learning rate heterogeneity among employees, the embodiment adopts Cao Lei to correct the model ^[20] ([20]Cao Lei Ming, Cao Xia, yellow Xia, based on employee learning behavior, multi-objective flexible workshop scheduling [ J]The computer integrated manufacturing system, 2018,24(08): 2023-2034) analyzes the influence of the employee learning effect with initial capability heterogeneity and learning rate heterogeneity on the working time, and the modified model is as follows:

in the formula, the first step is that,

for the initial operating capability of the staff w for operating the device k,/ _wk The learning ability of staff w in different devices. In this model, there is heterogeneity in initial competency and learning rate of employees, and different workers, even if they have the same initial competencyThe same work is carried out, and due to different learning rates, the completion time can be different. To clearly express the heterogeneity that exists between employees, an employee initial capability matrix is used

And a learning rate matrix L _wk The initial capacity matrix represents the operation capacity of the equipment when the staff independently go on duty after training, and the learning rate matrix represents the operation and learning capacity of different staff to each equipment.

Therefore, according to the above description, the skill planning configuration and workshop scheduling integrated optimization method based on learning ability according to the embodiment includes the following steps:

step 1, establishing an employee skill training decision and scheduling integrated optimization model considering learning capacity heterogeneity;

and 2, solving the model established in the step 1 by adopting a genetic simulated annealing algorithm to obtain a final optimal solution, and outputting an optimal target value and an optimal scheduling scheme.

In step 1, the staff skill training decision and scheduling integrated optimization model considering the heterogeneity of learning ability can be described as follows: the N kinds of workpieces with different batches are processed by W workers with different proficiency levels (generally W is less than or equal to M) on M devices with different types and operation capabilities at different positions of a workshop. A perfect staff training mechanism is established in an enterprise, and can train staff for any equipment operation, and heterogeneity exists among the staff, so that the staff after training has initial capability difference when going on duty formally, the staff learning rate has difference, and the staff is influenced by learning effect with heterogeneous learning capability, and obvious difference exists among different staff. When an order comes, N kinds of workpieces are to be processed in a certain scheduling period, and the demand of each kind of workpiece is D _i And the workpiece has N _i The working procedures are J working procedures in total, each working procedure can be used for processing on a plurality of devices, the device types and the processing capabilities are different, employees can increase new skills through skill training, the skill proficiency level is continuously improved through the learning effect, the processing time of the working procedures is different along with the selection of the devices and the difference of workers, and the adjustment is carried outThe degree scheme also changes with the matching scheme of different equipment and personnel, and worker resources, equipment resources, process processing sequences and staff skill training decisions need to be planned to minimize completion time, processing cost and equipment load balance.

The following assumptions were made for the model:

(1) the standard time of each process processed by a standard worker on the equipment is known, and the actual processing time is only related to the equipment and the worker and is not related to batches;

(2) all raw materials and production resources (machines and workers) can be used at time 0;

(3) workers have different initial abilities and learning rates and are known;

(4) the arrival time of all the workpieces is the same, the processes among different workpieces do not have a priority relation, and the priority relation with the processes of the workpieces is known;

(5) at the same time, each device can only process one procedure, and each worker can only operate one device;

(6) machine faults are not considered, and the process cannot be interrupted after the process is started;

(7) a staff training system is built in an enterprise, but staff training needs cost, and the training cost of each skill is assumed to be the same, namely the training cost of skill 1 and skill 2 is the same, and the cost for training one skill is C, so that the cost for training two skills is 2C;

(8) all employees can participate in training of all skills. The heterogeneity exists among the employees, so that different employees have different skill initial abilities when the employees are formally and independently on duty after training;

(9) the demand for orders is relatively stable, regardless of temporary inserts, order cancellation, and changes.

Next, the model is first constructed:

the symbols and variables of the design are defined as follows:

i, i': the number of workpieces to be processed is 1 … N;

k, l: the number of devices { k ═ 1 … M };

w, o: the number of workers { W ═ 1.. W };

j, j': a step { J ═ 1 … J };

s: lot index, denoted as the s-th sublot of workpiece i, s ═ 1 … l _i }；

C _k : the cost of use of the device k per unit time;

C _w : salary cost for worker w per unit time;

D _i : a demand for a workpiece i;

t _ijk : the standard time of the process j of processing the workpiece i by a worker using the device k;

e _wk : proficiency level of worker w using device k;

T _is : the finishing time of the s-th batch of workpieces i;

T _w : represents the total processing time of worker w in one production cycle;

d _i : a batch of each product;

l _i : number of batches of i-th workpiece,/ _i ＝[D _i /d _i ]，“[]"means rounding up;

S _ijskw : the start time of the s-th batch of process j in which a worker w processes a workpiece i using the apparatus k;

E _ijskw : the end time of the s-th batch of process j in which the worker w processes the workpiece i using the apparatus k;

T _ijskw : the actual processing time of the s-th lot of process j of processing the workpiece i by the worker w using the apparatus k;

x _ijskw : the value is 1, which indicates that a worker w uses the equipment k to process the s batch of the working procedure j of the workpiece i, otherwise, the value is 0;

x _ijsk : a binary variable of 1 indicates that the s-th batch of process j of workpiece i is allocated to be processed on equipment k, and otherwise, 0;

x _kw : a binary variable, 1, indicating that worker w has the ability to operate device k, otherwise 0;

y _wk : a binary variable, whether worker w needs to be trained in the use of device k, 1 indicates that it needs to be trainedTo, a value of 0 means not required.

Establishing an objective function:

in the actual production process, enterprises can set different expectations for different targets according to own interests, and can set some targets, for example, a quality department can take product quality as a target, a production department can take completion time and yield and productivity as targets, a cost department can take production cost as a target, an energy consumption department can take energy consumption and pollution generated by the energy consumption department as targets, and the enterprises can select different targets according to different purposes. In the embodiment, the completion time of the product, the production cost and the workload balance of the staff are selected as the objective function of the model.

(1) The completion time is minimal: the maximum time to minimize the completion of all products is targeted as a measure.

f ₁ ＝min(maxT _is ) (4)

(2) The manufacturing cost is lowest: the method is divided into three major parts, namely the use cost of equipment, the wage cost of workers and the training cost of staff.

(3) Staff workload balancing: staff working time is a factor directly related to staff, staff can cause the fact that staff is busy and idle unevenly in the workshop scheduling process due to the fact that the staff has different skill quantity and skill level, and part of staff idle time is too long. Therefore, the staff workload balancing coefficient is used as a measurement target.

The linear weighting method is a common method, determines the weight of each target according to the importance degree of different targets, and then adds the weights to form a weighted target function, and the method contains all original data and can be directly used for practical problems, so that the linear weighting method is selected to solve the multi-target problem. In combination with the expectations of the enterprise and the importance degrees of different targets, but the dimensions of the product completion time, the manufacturing cost and the load balancing coefficient in the model are different, so that normalization processing is required to be carried out in the solving process to eliminate the dimensions, and the objective functions are compared, wherein the method comprises the following steps:

wherein f is _max And f _min The maximum and minimum values of the objective function.

The above target weights can be combined as:

f＝α·f ₁ +β·f ₂ +γ·f ₃ (8)

where α, β, and γ are weights of completion time, manufacturing cost, and employee load balancing coefficients, and α + β + γ is 1.

The constraints are as follows:

(1) the process of any batch of workpieces can be started at time 0 when the conditions allow.

S _ijskw ≥0 (9)

(2) The process completion time of each batch of workpieces is determined by the start time of the process of the batch and the actual processing time of the batch.

E _ijskw ≥S _ijskw +d _i x _ijskw T _ijkw (10)

(3) Since there are different initial abilities and learning rates among workers, the completion time of any one process for any one kind of work is the completion time considering the learning effect.

(4) If the worker is assigned to the equipment k, no training is required if the worker has the ability to operate the equipment k, and if the worker does not have the ability, training is required.

(5) It is ensured that the workers w assigned to each equipment in each batch have the ability to operate the equipment.

x _ijskw ＝x _ijsk [x _kw +(1-x _kw )y _kw ] (13)

(6) And (4) constraint of precedence relationship. For different processes of the same workpiece in the same batch, the operation of the next process can be performed only after the previous process is finished.

S _ijskw ≥E _i(j-1)slo (14)

(7) For the work piece i, the process j can only select one from the equipment set which can process the process to process, and the worker operating the equipment can only select one from the optional worker set to operate the equipment.

(8) Total processing time of each worker in production.

(9) The uniqueness of the equipment restricts that one equipment cannot simultaneously process two procedures at a certain time.

S _ijskw X _ijskw -E _i'j'sko X _i'j'sko ≥0 (17)

(10) The uniqueness of workers is restricted, and the same worker can only operate one device and cannot simultaneously operate a plurality of devices.

S _ijskw X _ijskw -E _i'j'slw X _i'j'slw ≥0 (18)

(11) A binary variable constraint.

x _ijsk ＝{0,1}，x _ijskw ＝{0,1}，y _wk ＝{0,1} (19)

The model solution is as follows:

the genetic algorithm is a parallel search algorithm, has the advantages of stronger global search capability, simple process, good parallel processing capability, good robustness, higher general flexibility, high search speed and the like, but also has the defects of poorer local search capability, premature condition, easy premature convergence trapping in local optimum and failure of obtaining a global optimum solution; the simulated annealing algorithm is widely applied, can effectively solve the NP complete problem, has strong local search capability, can receive inferior solutions with certain probability and jump out of a local optimal trap to obtain a global optimal solution, but also has the defect of poor global search capability. From the above, it can be found that the two algorithms can mutually compensate weak terms of each other to form an algorithm having advantages in both global search and local search, and the two algorithms are mutually nested to mutually compensate respective defects, so that a global optimal solution can be quickly obtained.

The algorithm adopted in the embodiment is described as follows:

the genetic simulated annealing algorithm integrates the advantages of the genetic algorithm and the advantages of the simulated annealing algorithm, makes up the respective disadvantages, and can quickly obtain the global optimal solution, so that based on the advantages, the embodiment adopts the genetic simulated annealing algorithm to solve the model, and the detailed steps are as follows:

s1: generating an initialization population using a genetic algorithm;

s2: calculating the fitness value of the population individuals, and enabling the iteration number g to be 0;

s3: obtaining a temporary population by genetic operations such as selection, crossing, mutation and the like by comparing the individual fitness values;

s4: performing simulated annealing algorithm operation in the temporary population, determining the initial temperature of the simulated annealing algorithm, generating an initial solution S, and calculating a target function value corresponding to the solution S;

s5: generating a new solution S' through a disturbance strategy, calculating a target function value corresponding to the new solution, judging whether to accept the new solution by using a Metropolis criterion, and accepting to switch to S6 or not accepting to switch to S4;

s6: detecting whether sufficient search is performed at the temperature, if the search is performed sufficiently, turning to S7, and if the search is not performed sufficiently, turning to S4;

s7: judging whether the simulated annealing termination condition is met, if so, memorizing the current optimal solution in a memory table to be converted into S8 for continuous iteration; if not, performing annealing operation and transferring to S4;

s8: judging whether the iteration process reaches the maximum iteration number, if not, making the iteration number g equal to g +1, and turning to Step3 to perform a new iteration process; if yes, go to S9;

s9: and comparing all the optimal solutions in the memory table to obtain a final optimal solution, and outputting an optimal target value and an optimal scheduling scheme.

The flow chart of the genetic simulated annealing algorithm is shown in FIG. 1.

In the above algorithm steps, the population initialization and initial solution generation are specifically as follows:

initializing population size P, cross probability P _c Probability of mutation P _m Annealing initiation temperature T ₀ Cooling rate λ, maximum number of iterations G _max Threshold (annealing end) temperature T _min Chain length L.

At solution demand of D _i The dispatching time of the workpieces can be determined by the batch of each batch, and the batch of products can be dispatched. In genetic algorithms, a computer cannot directly identify parameters in a model, and thus it is required to convert them into individuals having a plurality of genetic compositions, and the process of converting the parameters into identifiable individuals is called encoding. Chromosomes are divided into three parts to code: lot-based process sequence coding, equipment resource-based allocation coding, and worker resource-based allocation coding. Workpiece procedure processing sequence information, machine equipment resource allocation information and worker resource allocation information. The length of the chromosome is

(N represents the number of workpieces, N _i Indicates that the workpiece i has N _i Channel process), three-layer coding is shown in FIG. 2, and each layer has the length

Lot-based process order coding: the code consists of several sets of repeating numbers, the first number representing a workpiece, the second number representing a batch, the number of occurrences of the repeating numbers representing the number of processes per workpiece, the order of occurrences of the numbers from left to right representing the processing order, as shown in fig. 2, the number 12 representing the first batch of workpieces 1, the number 12 representing the first processing batch of processes 1 of workpieces 1, the coded segment [ 12312122 ] meaning in terms of priority between product processes, the processing order being: work 1, work 3, work 1, work 2, work, and work, and work, work.

Device resource based allocation coding: the codes are encoded according to the types and the batches of the workpieces, from left to right, the workpieces 1 to n, and the codes inside the workpieces are the process priority and the batch order of each workpiece. Each process can be processed on a plurality of devices, the coding gene number represents the sequential position of the selected processing device in the selectable machine group, as shown in fig. 2, for example, the first number of the device code is 1, the 1 st device M1 is selected from the first selectable device groups [ M1, M3 and M4] of the process 1 of the workpiece 1 for processing, and the second number 2 represents the 2 nd device M4 is selected from the first selectable device groups [ M3 and M4] of the process 2 of the workpiece 1 for processing.

Worker resource based allocation coding: this layer of coding corresponds to the equipment resource allocation coding, where the numbers represent workers and the equipment resources are matched to the worker resources by the allocation of workers. In this embodiment, the staff can acquire the capability of operating the equipment through training in addition to the skills of the staff themselves, and expand the skill range of the staff themselves, so that for the worker resources, the existing worker resources are removed, and new resources can be added through training, so that the resource range of the worker is changed. The layer code comprises two kinds of information, namely, a worker for operating the second layer equipment is selected; secondly, whether the workers selected by the layer need to be increased or not is judgedAdding new skills if the selected worker has the ability to operate the second level equipment, i.e. x _ijkw 1 and x _kw 1, the employee does not need training y _kw 0 if the selected worker does not have the capability to operate the second floor equipment, i.e. x _ijkw 1 and x _kw If 0, the employee needs training, y _kw 1, training costs are incurred.

The chromosome decoding process can be described as: the process sequence and the batch production sequence of the processes, the machine equipment allocated for the work process of each batch, and the workers are determined step by step starting from the first layer.

In the above algorithm steps, the calculation and selection of the fitness value are specifically as follows:

and evaluating the skill planning and workshop scheduling optimization scheme by using a fitness function, wherein the larger the fitness value is, the more chance is to be selected as a parent. In the model, the inverse of the weighted objective function of eliminating the dimension among the objective functions by using normalization processing is compared, and the fitness function is shown as a formula (20).

The genetic algorithm selects a proper number of good individuals as parent breeding offspring through selection operation, the roulette method is the simplest and most common method, the random selection method is used for quickly selecting needed individuals through proportion, the roulette method is used for selecting offspring, the fitness value of the individuals is higher, the survival probability is higher, and the probability of selecting the good individuals as parent breeding offspring is higher. The selection mode is as follows: assuming that n individuals exist in the population, the fitness value of the ith individual is F _i Then the probability that the individual is selected is:

in the above algorithm steps, the crossover and mutation are specifically as follows:

(1) crossover operation

The cross-over functions to transmit the gene segments of the parent generation to the offspring and also functions as a global search. In the embodiment, the sequence crossover operators (POX) are adopted to respectively crossover the sequence of the work piece processes, as shown in fig. 3; the device resource allocation and the worker resource allocation are respectively intersected by using a multipoint intersection operator, as shown in fig. 4.

(2) Mutation operations

Similar to the crossover operation, the sequence of the work-piece process is mutated using a reverse mutation operator as shown in FIG. 5; equipment resource allocation and worker resource allocation are mutated using random single point mutations, as shown in fig. 6.

The state function generates a new entity:

(1) generating an initial solution S, and calculating an objective function value of the solution S;

(2) disturbing to generate a new solution S' by using a neighborhood searching technology;

(3) using Metropolis criterion to judge delta E ═ f (S ') -f (S)) and execute a memory function, if delta E is less than 0, using new solution S ' to replace old solution S, if delta E is more than 0, then generating random real number rand between [0, 1), calculating acceptance rate r ═ exp (-delta E/T) of new solution, if r is more than rand, then accepting new solution S ' as new current solution. After judging whether a new solution is accepted or not, setting an additional variable for storing the global optimal solution, and storing the global optimal solution in each iteration so as to avoid the loss of the optimal solution in the Metropolis criterion judgment process.

And (3) annealing operation:

the genetic algorithm has the problem of easy precocity and is easy to fall into the local optimal solution, so the embodiment is integrated into the simulated annealing operation to solve the problem. The simulated annealing operation comprises setting initial temperature, and judging whether the current temperature is higher than threshold temperature T _min Above the threshold temperature T _min And starting cooling operation, and selecting a linear cooling strategy.

And when the iteration times are reached or the obtained optimal solution is kept unchanged, the iteration is terminated, and the optimal solution is output.

The above embodiments were simulated and verified:

the invention provides a skill planning configuration and workshop scheduling integrated optimization problem model considering the heterogeneity of employee learning abilities, provides a genetic algorithm embedded with simulated annealing for solving, and verifies the optimization model provided by the embodiment through an example.

Case description: company a is a machining enterprise, which receives an order and distributes it to a workshop for production, in which 11 machines, 8 workers, 5 types of which are classified into a lathe, a milling machine, a grinding machine, a boring machine, a planer and a machining center capable of performing machining, milling and drilling operations, 2 lathes, 2 milling machines, 1 grinding machine, 1 boring machine, 2 planing machines and 3 machining centers are present. 8 workers in the workshop, old staff who had the existing work for many years also have temporary work or new staff that newly brought in, and every worker possesses different initial ability after the training, and because there is learning rate heterogeneity between every staff, there is the difference in the aspect of learning operation equipment, and different staff possess different learning efficiency.

The company distributes the received order to the workshop for production, the order comprises 7 workpieces, each workpiece comprises all operations in the workshop, the 11 equipment 8 workers in the workshop are used for production, and reasonable arrangement is needed to minimize the completion time, the manufacturing cost and the workload of the workers. The order information is shown in table 2, the operation that each equipment can process and the equipment information that each worker can operate are shown in table 1, and the initial ability of the employee after training and the learning efficiency of the employee are shown in tables 3 and 4.

Table 1 workpiece processing information table

TABLE 2 staff Requirements and production lots

Workpiece	1	2	3	4	5	6	7
								Demand volume	80	55	100	200	120	160	50
Batch size	16	10	20	40	25	30	10

TABLE 3 employee initial competency

TABLE 4 employee learning Rate

The bold font in the table represents the initial ability and the learning rate of the existing staff, and the bold font does not represent the initial ability of the trained staff.

And (4) analyzing results:

according to the data, the Matlab 2017a software is used for solving the employee skill planning configuration and workshop scheduling integrated optimization model considering the heterogeneity of employee learning abilities, the solving algorithm is a genetic simulated annealing algorithm, and the related initialization parameters are set as: p is 50, G _max ＝200，P _c ＝0.6，P _m ＝0.4，T ₀ ＝100，T _min The using cost of the equipment is 20 yuan/hour, the wage of workers is 20 yuan/hour, the cost of single training is 15 yuan/time, the weight alpha is 0.5, beta is 0.3, and gamma is 0.2.

Table 5 considers employee learning capabilities the same scheduling scheme results

Solution (II)	Total target value	Completion time (h)	Manufacturing cost (Yuan)	Coefficient of load balancing
					1	0.29	137.57	14523.33	2.29
2	0.30	148.44	14363.60	2.51
					3	0.32	152.16	14372.50	2.44
4	0.33	161.93	14239.91	2.52
					5	0.36	137.89	14570.22	2.72
6	0.41	145.32	14542.88	2.86
					7	0.43	151.48	14505.87	2.88
8	0.44	160.52	14435.20	2.74
					9	0.45	164.65	14389.07	2.85
10	0.46	147.81	14628.61	2.82

TABLE 6 scheduling scenario results considering employee learning capability differences

TABLE 7 Schedule plan results without accounting for employee learning effects

Solution (II)	Total target value	Completion time (h)	Manufacturing cost (Yuan)	Coefficient of load balancing
					1	0.36	159.02	14367.00	2.90
2	0.38	161.68	14375.83	2.89
					3	0.40	163.30	14506.00	2.32
4	0.41	157.68	14476.25	3.00
					5	0.41	155.90	14544.75	2.80
6	0.42	158.95	14573.00	2.48
					7	0.43	171.23	14373.67	2.62
8	0.45	160.46	14547.50	2.80
					9	0.49	161.52	14626.75	2.71
10	0.50	165.69	14556.75	2.81

Table 5, table 6, and table 7 are results obtained by solving using the genetic simulated annealing algorithm, respectively, table 5 is a skill training decision and workshop scheduling integrated optimization scheme considering the same learning effect of employee learning ability, table 6 is a skill training decision and workshop scheduling integrated optimization scheme considering a heterogeneous learning effect of employee learning ability, and table 7 is a skill training decision and workshop scheduling integrated optimization scheme not considering employee learning effect. Fig. 7 and 8 compare the completion time and the manufacturing cost of different schemes in three cases, and it can be seen from fig. 7 and 8 that the learning ability considering the heterogeneity of employees has a positive influence on the scheduling result, the scheduling scheme is optimal, the completion time and the manufacturing cost are obviously superior to the other two schemes, and the scheduling scheme not considering the learning effect of employees is the worst.

The learning ability has a positive effect on the scheduling result from the whole view, the completion time and the manufacturing cost can be obviously reduced, the solving result is stable, and the optimal results of the three schemes are compared by tables 5, 6, 7 and 8 and figures 7 and 8, so that the scheme is optimal when the heterogeneity learning of the staff is considered.

Compared with the optimal results of the three schemes, the completion time of the dispatching scheme can be effectively reduced by considering the learning capacity of the staff, wherein compared with the dispatching scheme without considering the learning effect, the completion time of the integrated optimization scheme of the skill training decision and the workshop dispatching considering the same learning capacity of the staff is reduced by about 13.5 h, and the completion time of the dispatching scheme considering the heterogeneity of the learning capacity of the staff is reduced by 27.7h and about 17.4%. At this time, equipment-worker matching and employee skill resource planning of an optimal scheduling scheme considering heterogeneous employee learning abilities are shown in tables 9 and 10, wherein the table 9 is a process-equipment-worker distribution scheme, a first number in the table is selected equipment, and a second number is a worker matched with the equipment; table 10 shows the skill matrix after the skill resource planning of the worker, 1 indicates that the skill is required to be possessed, 0 indicates that the skill is not required, and bold font indicates that the skill is possessed before the skill resource planning.

Table 8 optimal scheduling results of three methods

TABLE 9 optimal scheduling scheme procedure-equipment-worker assignment scheme

TABLE 10 optimal scheduling plan worker skill resource planning

In conclusion, the invention constructs a workshop scheduling optimization model considering staff training and learning effects and designs a simulated annealing algorithm for solving. The enterprise realizes the increase of the skills of the staff through the training of the skills of the staff, realizes the multi-technical of the staff, increases the staff distribution scheme in the dispatching process and improves the staff utilization rate; the learning effect in the operation process of the staff is considered, the calculated operation time of the workers can be closer to the actual working time, the configuration of the resources of the workers is facilitated, the labor utilization rate is improved, and the obtained production scheduling scheme is more accurate and excellent.

The foregoing shows and describes the general principles and features of the present invention, together with the advantages thereof. It will be understood by those skilled in the art that the present invention is not limited to the embodiments described above, which are described in the specification and illustrated only to illustrate the principle of the present invention, but that various changes and modifications may be made therein without departing from the spirit and scope of the present invention, which fall within the scope of the invention as claimed.

Claims (7)

1. The skill planning configuration and workshop scheduling integrated optimization method based on learning ability is characterized by comprising the following steps:

step 1, establishing an employee skill training decision and scheduling integrated optimization model considering learning capacity heterogeneity;

and 2, solving the model established in the step 1 by adopting a genetic simulated annealing algorithm to obtain a final optimal solution, and outputting an optimal target value and an optimal scheduling scheme.

2. The integrated learning-ability-based optimization method for skill planning configuration and shop scheduling of claim 1, wherein:

in the step 1, the staff skill training decision and scheduling integrated optimization model considering the heterogeneity of the learning ability is described as follows: the method comprises the following steps that N kinds of workpieces with different batches are processed on M devices which are located in different positions of a workshop and have different types and operation capabilities by W workers with different proficiency levels, wherein W is less than or equal to M; when an order comes, N kinds of workpieces are to be processed in a certain scheduling period, and the demand of each kind of workpiece is D _i And the workpiece has N _i The process comprises J processes, wherein each process can be carried out on a plurality of devices;

selecting product completion time, production cost and employee work load balance as a target function of the model;

(1) the completion time is minimal: the goal of minimizing the maximum time to completion of all products is a measure

f ₁ ＝min(maxT _is ) (4)

(2) The manufacturing cost is lowest: the cost of the equipment, the wage cost of workers and the training cost of staff are divided into three major parts

(3) Staff workload balancing: taking the staff workload balance coefficient as a measurement target

The dimensions of the three targets of product completion time, manufacturing cost and load balancing coefficient in the model are different, so normalization processing is required to be carried out in the solving process to eliminate the dimensions for comparison of target functions, and the following formula is adopted:

wherein f is _max And f _min The maximum value and the minimum value of the objective function are obtained;

combining the above target weights as:

f＝α·f ₁ +β·f ₂ +γ·f ₃ (8)

wherein α, β, and γ are weights of completion time, manufacturing cost, and employee load balancing coefficients, and α + β + γ is 1;

the constraints are as follows:

(1) the process of workpieces of any batch is started at 0 moment when the conditions allow

S _ijskw ≥0 (9)

(2) The process completion time of each batch of workpieces is determined by the start time of the process of the batch and the actual processing time of the batch

E _ijskw ≥S _ijskw +d _i x _ijskw T _ijkw (10)

(3) The completion time of any process of any workpiece is the completion time considering the learning effect

(4) Whether to train constraints, if a worker is assigned to a device k, if the worker has the ability to operate the device k, no training is required, and if the worker does not have the ability, training is required

(5) Ensuring that the workers w assigned to each equipment in each batch have the ability to operate the equipment

x _ijskw ＝x _ijsk [x _kw +(1-x _kw )y _kw ] (13)

(6) For different processes of the same workpiece in the same batch, the operation of the next process can be carried out only after the previous process is finished

S _ijskw ≥E _i(j-1)slo (14)

(7) The process j of the workpiece i can only select one from the equipment set capable of processing the process for processing, and the worker for operating the equipment can only select one to operate the equipment from the selectable worker set

(8) Total processing time of each worker in production

(9) The uniqueness of the equipment is restricted, one equipment cannot simultaneously process two procedures at a certain time

S _ijskw X _ijskw -E _i'j'sko X _i'j'sko ≥0 (17)

(10) The uniqueness of workers is restricted, one worker can only operate one device and cannot simultaneously operate a plurality of devices

S _ijskw X _ijskw -E _i'j'slw X _i'j'slw ≥0 (18)

(11) Constraint of binary variable

x _ijsk ＝{0,1}，x _ijskw ＝{0,1}，y _wk ＝{0,1} (19)。

3. The integrated learning-ability-based optimization method for skill planning configuration and shop scheduling of claim 2, wherein:

in the step 2, solving the model by adopting a genetic simulated annealing algorithm, which comprises the following specific steps:

s1: generating an initialization population using a genetic algorithm;

s2: calculating the fitness value of the population individuals, and enabling the iteration number g to be 0;

s3: obtaining a temporary population through selection, crossing and variant genetic operations by comparing individual fitness values;

s4: performing simulated annealing algorithm operation in the temporary population, determining the initial temperature of the simulated annealing algorithm, generating an initial solution S, and calculating a target function value corresponding to the solution S;

s5: generating a new solution S' through a disturbance strategy, calculating a target function value corresponding to the new solution, judging whether the new solution is accepted or not by using a Metropolis criterion, and switching to S6 if the new solution is accepted, and switching to S4 if the new solution is not accepted;

s6: detecting whether sufficient search is carried out at the temperature, if the sufficient search is carried out, turning to S7, and if the sufficient search is not carried out, turning to S4;

s7: judging whether the simulated annealing termination condition is met, if so, memorizing the current optimal solution in a memory table to be converted into S8 for continuous iteration; if not, the annealing operation is carried out to S4;

s8: judging whether the iteration process reaches the maximum iteration number, if not, making the iteration number g equal to g +1, and turning to S3 to perform a new iteration process; if yes, go to S9;

s9: and comparing all the optimal solutions in the memory table to obtain a final optimal solution, and outputting an optimal target value and an optimal scheduling scheme.

4. The integrated learning-ability-based optimization method for skill planning configuration and shop scheduling of claim 3, wherein:

step S1 is specifically as follows: initializing population size P, cross probability P _c Probability of mutation P _m Annealing initiation temperature T ₀ Cooling rate λ, maximum number of iterations G _max Threshold temperature T _min Chain length L;

at solution demand of D _i The scheduling time of the work, the lot of each product is determined by the lot size of each lotCounting, scheduling batched products; in genetic algorithms, a computer cannot directly identify parameters in a model, so that the parameters need to be converted into individuals consisting of a plurality of genes, and the process of converting the parameters into the identifiable individuals is called encoding; chromosomes are divided into three parts to code: a lot-based process sequence code, an equipment resource-based allocation code, and a worker resource-based allocation code; workpiece procedure processing sequence information, machine equipment resource allocation information and worker resource allocation information; the length of the chromosome is

N denotes the number of workpieces, N _i Indicates that the workpiece i has N _i The procedure is that three layers of codes are coded, and the length of each layer is

Lot-based process order coding: the code consists of several groups of repeated numbers, the first number represents a workpiece, the second number represents a batch, the number of times of occurrence of the repeated numbers represents the number of processes of each workpiece, and the sequence of occurrence of the numbers represents the processing sequence;

encoding based on the allocation of device resources: coding according to the types and the batches of the workpieces, wherein the codes inside the workpieces are the process priority and the batch sequence of each workpiece from the workpiece 1 to the workpiece n; each procedure is processed on a plurality of devices, and the coding gene number represents the sequential position of the selected processing device in the selectable machine set;

worker resource based allocation coding: the layer of codes corresponds to equipment resource allocation codes, wherein the numbers represent workers, and the equipment resources are matched with the worker resources through allocation of the workers; the layer code comprises two kinds of information, namely, a worker for operating the second layer equipment is selected; secondly, judging whether the selected worker on the layer needs to add new skills, if the selected worker has the capability of operating the equipment on the second layer, namely x _ijkw 1 and x _kw 1, the employee does not need training y _kw 0, if the selected worker does not have to operate the second floor equipmentCapability of (i.e. x) _ijkw 1 and x _kw If 0, the employee needs training, y _kw 1, generating training costs;

the chromosome decoding process is described as: the process sequence and the batch production sequence of the processes, the machine equipment allocated for the work processes of each batch, and the workers are determined step by step starting from the first layer.

5. The integrated learning-ability-based optimization method for skill planning configuration and shop scheduling of claim 3, wherein: step S2 is specifically as follows:

evaluating a skill planning and workshop scheduling optimization scheme by using a fitness function, wherein the larger the fitness value is, the more chance is to be selected as a parent; in the model, the inverse of a weighted objective function of eliminating dimensions among the objective functions by using normalization processing is compared, and a fitness function is as follows

The genetic algorithm selects a proper number of good individuals as parent breeding filial generations through selection operation, offspring is selected by using a roulette method, the higher the fitness value of the individual is, the higher the survival probability is, and the higher the probability of the good individuals selected as parent breeding filial generations is; the selection mode is as follows: assuming that n individuals exist in the population, the fitness value of the ith individual is F _i Then the probability that the individual is selected is:

6. the integrated learning-ability-based skills planning, configuration and shop scheduling optimization method according to any of claims 3-5, wherein: the crossover operation in step S3 is specifically as follows:

the sequence crossing operators are adopted to respectively cross the working procedure sequence of the workpieces; or respectively carrying out intersection on equipment resource allocation and worker resource allocation by using a multipoint intersection operator.

7. The integrated learning ability-based skills planning, configuration and plant scheduling optimization method according to any of claims 3-5, wherein: the mutation operation in step S3 is specifically as follows:

carrying out mutation on the sequence of the work piece working procedures by adopting a reverse mutation operator; alternatively, equipment resource allocation and worker resource allocation are mutated using random single point mutations.

Images