CN105843197A - Teaching-and-learning-algorithm-based static scheduling optimization system for discrete manufacturing shop - Google Patents

Abstract

The invention discloses a teaching-and-learning-algorithm-based static scheduling optimization system for a discrete manufacturing shop. The system comprises a data server, an operation server, and a terminal display screen. A scheduling optimization client is embedded into the operation server and is used for carrying out optimization-algorithm-based reasonable scheduling arrangement on a processing task on the day by analyzing the processing task and consulting various detailed information of the data server, wherein the specific optimization algorithm employs a teaching and learning algorithm. A generation scheduling scheme is decoded; de-compilation processing is carried out according to dual-layer definition on a student during coding; a processing tool and a processing sequence of a workpiece are expressed successively. And then the optimized scheduling scheme is displayed on a terminal display screen, so that the production staff and the management staff in the shop can obtain current specific processing task arrangement information in real time. According to the invention, with the teaching and learning algorithm, the system has advantages of fast convergence speed and high optimization searching capability and adapts to the complex situation in practical production well.

Description

Teaching and learning algorithm-based static scheduling optimization system for discrete manufacturing workshop

Technical Field

The invention relates to the technical field of discrete manufacturing workshop scheduling, in particular to a teaching and learning algorithm-based static scheduling optimization system for a discrete manufacturing workshop.

Background

The manufacturing industry in the 21 st century is increasingly competitive, and enterprises are threatened and challenged while gaining a great development space. The factors that determine the core competitiveness of an enterprise are reflected in the products offered by the enterprise. The product is required to have high quality, low price and short delivery period, and can continuously adapt to changeable markets and individualized customer demands, which puts higher requirements on production operation plans and balanced production of enterprise manufacturing workshops. In the current production process of the discrete manufacturing workshop, the problems of overlong machine tool standby time, unreasonable part production arrangement and the like exist, so that the static scheduling arrangement before the workshop production is necessary, the reasonable static scheduling arrangement is beneficial to improving the energy efficiency of the manufacturing process, and the core benefits of enterprises are met. At present, the academic world has developed extensive research on scheduling arrangement of manufacturing workshops, and various theoretical algorithms are generated, and due to the lack of understanding of actual production environments, most algorithms at present cannot be directly applied to static scheduling of discrete manufacturing workshops, so that reasonable algorithms are applied to the manufacturing workshops, and the aim of reducing working hours and production energy consumption is achieved, and the method is a key point of research.

Disclosure of Invention

Aiming at the defects in the prior art, the invention combines the advantages of teaching and learning algorithms, applies the teaching and learning algorithms to the static scheduling of the manufacturing workshop through partial improvement, and provides a discrete manufacturing workshop static scheduling optimization system, thereby achieving the result of improving the energy efficiency.

According to the technical scheme provided by the invention, the discrete manufacturing workshop static scheduling optimization system based on teaching and learning algorithm comprises a data server, an operation server and a terminal display screen; the data server collects the processing energy consumption, time consumption and standby conditions of the machine tool in the manufacturing workshop, counts the processing conditions of the workshop in real time, tracks the production energy consumption data of the machine tool in the workshop in real time, and counts the processing conditions of the machine tool and the attendance conditions of workers at the same time; the operation server is embedded with a scheduling optimization client, the scheduling optimization client performs reasonable scheduling arrangement on the processing tasks on the same day through an optimization algorithm by analyzing the processing tasks and referring to various detailed information of the data server, the optimization algorithm specifically adopts a teaching and learning optimization algorithm, and the construction method of the scheduling optimization client is as follows:

step one, determining an objective function of scheduling optimization of a discrete manufacturing workshop;

the minimum maximum processing time and the minimum total energy consumption of a machine tool in a manufacturing workshop are taken as objective functions, the two objective functions are fused by adopting a weighting method, and when the fused value is minimum, the scheduling requirement is met, namely

FM＝min(max(FM_i))

(1)

$WM=\underset{i=1}{\overset{n}{\Σ}}{\mathrm{wm}}_i---\left(2\right)$

min(W)＝u₁FM+u₂WM (3)

FM_i-all the work in the machine tool M_iA completion time of the above;

FM-Final finish time of all workpieces;

wm_i-all the work in the machine tool M_iThe consumption of energy of;

WM-Total energy consumption of all machine tools;

i-is a positive integer variable greater than 0;

n is the total amount of the machine tool;

min (W) -the final weighting values of the two objective functions;

u₁、u₂-two weighted values, which are valued according to the actual needs of the enterprise;

during the production, the scheduling optimization algorithm needs to satisfy the following constraint conditions:

1.1, a certain procedure can only use the same machine to process at a certain moment;

1.2, once a process of a certain workpiece starts to process, the process cannot be terminated randomly before the process is completed;

1.3, the processing priorities of all the workpieces are the same at zero time;

1.4, the priority of the processing sequence among the procedures of different workpieces is the same, and the processing of the procedures of the same workpiece must be in accordance with the previously appointed processing sequence;

step two, solving the objective function by adopting an improved teaching and learning optimization algorithm, wherein the specific solving process comprises the following steps:

step 2.1, encoding: each scheduling scheme represents a student, the scores of the students are defined in a double-layer design mode, and the first layer isDefining a process layer which represents each process of the ith part; second layerDefining a machine layer which represents a machine during the process machining of the ith part;

step 2.2, parameter initialization: initializing students in the teaching and learning optimization algorithm according to the number of machine tools and the processing tasks of parts;

step 2.3, solving an objective function based on an improved teaching and learning algorithm, and outputting an optimal scheduling scheme;

the running process of the improved teaching and learning algorithm comprises a teacher teaching process, a student learning process and an updating stage;

a. in the teacher's teaching process, the student is assumed to be X^jThe final weight of the two objective functions is defined as f (X)ⁱ)，

j 1, 2, 3, N, i ∈ {1234.. N }, where N is the total number of students, and min (f (X)ⁱ) Student as teacher X_teacherSelecting the nearestStudent of average value of f (X) as Mean value; the "teach" process formula is as follows:

$X_{new}^j=X_{old}^j+r_j(X_{teacher}-{\mathrm{TF}}_j\·Mean)---\left(4\right)$

wherein,andrespectively representing parameters before and after the optimization of the jth scheduling scheme, namely the processing arrangement and TF of each part_jAs a teaching factor, r_jFor learning step length, the two parameters are used for adjusting the learning speed; at this time, the scheduling scheme obtained by floating point operation may have decimal or over-interval number, and the number needs to be adjustedDiscretizing, wherein the specific discretizing process comprises the following steps:

a.1, firstly, rounding the numbers of the process layer and the machine layer, neglecting decimal, then automatically setting the number exceeding the interval to zero, comparing with the scheduling task list of the current day, marking the workpieces with redundant processes at the process layer, and taking l₁、l₂、...、l_cMarking the workpiece number, wherein c ∈ p and p are workpiece number sets to be marked, and finally randomly picking out the same redundant number from the workpiece number according to the number of redundant processes to be set as zero, and simultaneously setting the corresponding part of the machine layer as zero;

a.2, rearranging the zero-set workpiece numbers, comparing the workpiece numbers with a scheduling task list again, marking the workpiece numbers with insufficient working procedure quantity, randomly generating a group of scheduling schemes according to the specific missing number of the workpieces, sequentially filling the schemes to the positions of the working procedure layer with zero according to the sequence, and simultaneously randomly filling the machine numbers meeting the processing requirements in the positions corresponding to the machine layer;

a.3, taking the scheduling scheme generated for the first time as an initial scheme, carrying out local iteration, setting iteration times, skipping to a.2, if the generated scheduling scheme is superior to the initial scheme, setting the scheduling scheme as the initial scheduling scheme, and outputting the scheduling scheme until the iteration times are met;

b. in the student's learning process, each scheduling scheme X^hRandomly selecting a learning object X from all scheduling schemes^gH 1, 2, 3, ·, N; 1, 2, 3,. N; wherein h is≠g；X^hBy and X^gComparing the difference between the two, X^hAnd carrying out corresponding optimization adjustment, wherein the learning process is represented by the following formula:

when f (X)^h)≤f(X^g) Time of flight

When f (X)^h)＞f(X^g) Time of flight

Wherein r is_hRepresenting the learning step length of the h-th student, and generating a new scheduling schemeThe dispersion is also needed, and the specific dispersion process is the same as that in the teacher's teaching process;

c. the updating stage means that after the learning process, the generated scheduling scheme needs to be updated and optimized, and the updating method is as follows:

when in useTime of flight

When in useTime of flight

Wherein, student d belongs to {1, 2, 3, …, N };

the scheduling optimization client decodes the generated scheduling scheme, performs decompiling according to the definition of the student double-layer during encoding, and sequentially shows the processing machine tool and the processing sequence of the workpiece; and finally, displaying the optimized scheduling scheme on a terminal display screen, so that workshop production personnel and management personnel can know the current specific processing task arrangement condition in real time.

In the teacher's teaching process, the teaching factor TF_j＝round[1+rand(0，1)]Rand (0, 1) denotes a random number randomly generated from 0 to 1, round denotes a rounding function so that the value taken is an integer, and the learning step r is an integer_j＝rand(0，1)。

In the process of student learning, the learning step length r of the h-th student_hRand (0, 1), and rand (0, 1) means that a random number is randomly generated from 0 to 1.

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

1) according to the invention, a database server is built, the production energy consumption data of equipment is directly integrated into the system, the production data is analyzed, and the specific energy consumption of a single process of producing a single workpiece by a machine tool can be detailed, so that the processing differences of different machine tools for producing the same workpiece are determined, and the rationality of scheduling is facilitated; meanwhile, the data server can count the normal working condition of the machine tool in the discrete manufacturing workshop and know the working hour condition of workshop worker processing, so that the workshop can be maintained in a normal production order, and the processing can be rapidly carried out.

2) The teaching and learning optimization algorithm adopted by the invention has the advantages of high convergence rate and strong optimization capability, and compared with other group intelligent algorithms, the algorithm only needs fewer iteration times and has better obtaining effect; meanwhile, the algorithm does not need to set a plurality of optimization parameters, so that the optimization purpose is easier to achieve; in addition, when the algorithm optimizes two objective functions, the optimization degree is determined by actually setting two weight parameters, so that the algorithm is more suitable for the complex situation in actual production.

3) The energy efficiency is directly taken as an optimization target, the energy consumption of a single processing procedure is taken as a unit, the energy-saving requirement can be directly met, the current green manufacturing requirement is better met, and the practicability is better.

4) The invention displays the scheduling scheme on the terminal display, is convenient for processing personnel and management personnel to check, feeds back unreasonable positions in time through checking, and the operation server appropriately adjusts the scheduling scheme according to the feedback condition, thereby ensuring that the scheduling scheme has more rationality.

Drawings

Fig. 1 shows an encoded form of the scheduling algorithm.

FIG. 2 is a discretized block diagram of an improved teaching and learning optimization algorithm.

FIG. 3 is a scheduling flow diagram of an improved teaching and learning optimization algorithm.

FIG. 4 is a diagram of a discrete manufacturing shop optimization system operating.

Detailed Description

The present invention will be described in detail below with reference to the accompanying drawings and examples.

The scheduling optimization system set up by the invention comprises a data server, an operation server and a terminal display screen. As shown in fig. 4, a scheduling algorithm is embedded in the operation server, and the scheduling is displayed on the terminal display screen by analyzing and calculating the data of the data server. The specific implementation steps are as follows:

firstly, the method comprises the following steps: the method comprises the steps of building a data server, collecting the machining energy consumption, time consumption and standby conditions of a machine tool in a manufacturing workshop, counting the machining conditions of the workshop in real time, tracking the production energy consumption data of the machine tool in the workshop in real time, and counting the machining conditions of the machine tool and the attendance conditions of workers at the same time.

II, secondly: an operation server is set up, a scheduling optimization client is compiled in the server, the scheduling optimization client reasonably schedules the processing tasks on the day through an optimization algorithm by analyzing the processing tasks and referring to various detailed information of a data server, the optimization algorithm adopts a teaching-learning-based optimization algorithm (teaching-learning-optimization algorithm), the algorithm is provided for a continuous system, the application of the discrete manufacturing workshop needs to be properly improved, and the specific construction scheme of the scheduling optimization client is as follows:

1) an objective function for the discrete manufacturing shop scheduling optimization is determined.

According to the comprehensive consideration of energy consumption and benefits of enterprises, the minimum maximum processing time and the minimum total energy consumption of machine tools in a manufacturing workshop are taken as objective functions. In order to meet different requirements, a weighting method is adopted to fuse the two objective functions, and when the fused value is minimum, the scheduling requirement is met, namely

FM＝min(max(FM_i)) (1)

$WM=\underset{i=1}{\overset{n}{\Σ}}{\mathrm{wm}}_i---\left(2\right)$

min(W)＝u₁FM+u₂WM (3)

FM_i-all the work in the machine tool M_iA completion time of the above;

FM-Final finish time of all workpieces;

wm_i-all the work in the machine tool M_iThe consumption of energy of;

WM-Total good energy consumption of all machine tools;

i-is a positive integer variable greater than 0;

n is the total amount of the machine tool;

min (W) -the final weighting values of the two objective functions;

u₁、u₂two weighted values, the actual value is according to the actual needs of the enterprise;

during the production, the scheduling optimization algorithm needs to satisfy the following constraint conditions:

1) a certain procedure can only be processed by the same machine at a certain time;

2) a process of a workpiece cannot be terminated randomly (i.e., without regard to occurrence of machine failure) once it begins to be machined, until the process is completed;

3) the processing priorities of all the workpieces are the same at zero time;

4) the priority of the processing sequence among the processes of different workpieces is the same, and the processing of the processes of the same workpiece must be in accordance with the previously appointed processing sequence.

2) And solving the objective function by adopting an improved teaching and learning optimization algorithm, wherein the concrete solving process is as follows:

2.1) encoding:

each scheduling scheme represents a student, the scores of the students are defined in a double-layer design mode, and the first layer isDefining a process layer which represents each process of the ith part; second layerThe definition is a machine layer, and indicates a machine at the time of process machining of the ith part. For example, student XⁱA3 workpiece 5 machine scheduling scheme is shown, wherein a "1" in the first row and first column indicates a first pass of workpiece 1The work process, the "2" in the first row of the second row indicates that the first process of the workpiece 1 is processed on the machine 2, the first row has three "1" s, indicating that the workpiece 1 has three processes, and the position of the three "1" s indicates the order in which the three processes are arranged to be processed, as shown in fig. 1 in particular.

2.2) parameter initialization: initializing students in the teaching and learning optimization algorithm according to the number of machine tools and the processing tasks of parts;

and 2.3) solving an objective function based on an improved teaching and learning algorithm, and outputting an optimal scheduling scheme.

The operation process of the specific improved teaching and learning algorithm is as follows:

a. the teacher's "teaching" process.

Referring to the "teach" phase of FIG. 3, assume student X^j(j ═ 1, 2, 3.., N where N is the total number of students), the final weighted value of the two objective functions is defined as f (X)ⁱ) (i ∈ {1234.. N }), selecting min (f (X)ⁱ) Student as teacher X_teacherSelecting the nearestThe student of (f) (average of X) was taken as Mean value. The specific "teaching" process formula is as follows:

$X_{new}^j=X_{old}^j+r_j(X_{teacher}-{\mathrm{TF}}_j\·Mean)---\left(4\right)$

wherein,andrespectively representing parameters before and after the optimization of the jth scheduling scheme, namely the processing arrangement and TF of each part_jIs a Teaching Factor (TF)_j＝round[1+rand(0，1)])，r_jFor learning the step length (r)_jRank (0, 1)), where round denotes the rounding function, making its value always an integer, and rank (0, 1) denotes the generation of a random number from 0 to 1, these two key parameters being used to adjust the learning speed. At this time, the scheduling scheme obtained by floating point operation may have decimal or over-interval number, and the number needs to be adjustedDiscretization is carried out, and referring to fig. 2, a specific discretization process is as follows:

step 1: firstly, adopting rounding method to the numbers of the process layer and the machine layer, neglecting decimal, then automatically setting the number exceeding the interval to zero, comparing with the scheduling task list of the same day, marking the workpieces with redundant processes at the process layer, and taking l₁、l₂、...、l_c(c ∈ p, p is the workpiece number set to be marked) marking the workpiece number, and finally randomly picking out the same excess number from the workpiece number according to the number of the excess working procedures to be set as zero, wherein the machine layer is also set as zero;

step 2: rearranging the zero-set workpiece numbers, comparing the workpiece numbers with a scheduling task list again, marking the workpiece numbers with insufficient working procedure quantity, randomly generating a group of scheduling schemes according to the specific missing number of the workpieces, sequentially filling the schemes to the positions with zero working procedure layers according to the sequence, and simultaneously randomly filling machine numbers meeting the processing requirements into the positions corresponding to the machine layers;

step 3: and taking the scheduling scheme generated for the first time as an initial scheme, carrying out local iteration, setting the iteration number as q, skipping step2, if the generated scheduling scheme is superior to the initial scheme, setting the scheduling scheme as the initial scheduling scheme, and if the iteration number meets q, outputting the scheduling scheme.

b. The student's "learning" process.

Referring to the "learning" stage of FIG. 3, in the "learning" process of students, each scheduling scheme X^hRandomly selecting a learning object X from all scheduling schemes^gH 1, 2, 3, ·, N; 1, 2, 3,. N; wherein h is not equal to g; x^hBy and X^gComparing the difference between the two, X^hAnd carrying out corresponding optimization adjustment, wherein the mathematical process is expressed by the following formula:

when f (X)^h)≤f(X^g) Time of flight

When f (X)^h)＞f(X^g) Time of flight

Wherein r is_hRepresenting the learning step length of the h-th student, and generating a new scheduling schemeThe dispersion is also needed, and the specific dispersion process is the same as that in the teacher's teaching process;

c. the updating stage refers to that after the learning process, the generated scheduling scheme needs to be updated and optimized, and the specific process refers to the updating stage of fig. 3, and the updating method is as follows:

when in useTime of flight

When in useTime of flight

Wherein, student d is ∈ {1, 2, 3, …, N }.

Thirdly, the method comprises the following steps: decoding the generated scheduling scheme, performing decompilation according to the definition of student double-layer during encoding, and sequentially showing the processing machine tool and the processing sequence of the workpiece

Fourthly, the method comprises the following steps: and displaying the optimized scheduling scheme on a terminal display screen, so that workshop production personnel and management personnel can know the current specific processing task arrangement condition in real time.

Referring to fig. 4, the data server and the operation server are built in the discrete manufacturing workshop, the machine tool manufacturing information and the workshop processing plan are analyzed, the improved teaching and learning optimization algorithm is adopted to perform offline scheduling arrangement on workshop production, and finally, the specific scheduling task is displayed to the workshop through the display screen, so that the production is conveniently guided.

The following is a specific embodiment:

the method comprises the following steps: a data server is set up and is communicated with a workshop machine tool system to acquire the processing time (including the standby time) of a single procedure of processing a single workpiece by a machine toolDead timeCutting time) And details of energy consumption of the process (including standby energy consumption)No load energy consumptionEnergy consumption for cutting) Wherein i represents a machine tool, j represents a workpiece, k represents a specific process, d represents a standby state, o represents a no-load state, and q represents a cutting state, so that the machine tool i processes the k-th process of the workpiece j for a time period taken for the processAnd energy consumptionAnalyzing and processing the acquired data in the data server, and integrating energy consumption information:

time for machine tool to process workpiece j:

energy consumption of the machine tool for processing the workpiece j:

wherein, t_xyIndicates the time, t, taken by the machine number x to process the y-th pass of the part j_jThe first row of the matrix represents a first procedure, all columns of the first row represent the processing time used by a machine capable of processing the procedure, the first column represents a machine with the machine number 1, and when the machine cannot process the procedure, the corresponding coordinate value is zero, so that the manufacturing characteristics of a discrete manufacturing workshop are better met; in the same way, w_xyThe energy consumption of the y-th process for machining the part j by the machine number x is shown, and the meaning of the matrix coordinate values is the same as that described above, and will not be described in a repeated manner.

Meanwhile, the data server receives the current production task list from the workshop management system at regular time, additionally summarizes the existing workload of the workshop machine tool and the normal operation condition of the machine tool, counts the on-duty condition of workshop operation workers and ensures that the scheduling task can be normally executed.

Step two: the operation server requests a current production task list from the data server and simultaneously acquires energy consumption integration information;

step three: the optimization objective is determined from the obtained information and requirements, as follows:

FM＝min(max(FM_i)) (11)

$WM=\underset{i=1}{\overset{n}{\Σ}}{\mathrm{wm}}_i---\left(12\right)$

min(W)＝u₁FM+u₂WM(13)

the symbols are defined as follows:

FM_i-all the work in the machine tool M_iA completion time of the above;

FM-Final finish time of all workpieces;

wm_i-all the work in the machine tool M_iThe consumption of energy of;

WM-Total good energy consumption of all machine tools;

i-is a positive integer variable greater than 0;

n is the total amount of the machine tool;

min (W) -the final weighting values of the two objective functions;

u₁、u₂-two weighting values, actuallyThe value is according to the actual needs of the enterprise;

step four: initializing parameters of an improved teaching and learning algorithm: number of students z, number of iterations B, and optimization objective weight u₁、u₂(must satisfy u₁+u₂1); the number of students is set to be 15, so that the operation speed of the algorithm is not influenced, and the optimization precision of the algorithm is ensured; the iteration number B is set to 500; the optimized target weight value is dynamically adjusted according to the processing list and the customer urgency;

step five: coding;

determining the number m of workpieces to be machined and the number n of workpieces which can be used for machining according to the task list and the machine tool running condition, numbering the workpieces according to 1, 2, … and n, numbering the machine tool according to 1, 2, … and m in the same way, and counting the number c of the machining processes required by each workpiece_i(i ∈ {1, 2, 3, …, n }) Each student adopts a two-layer structure, the first layer being a process layerThe second layer is a machine layerNumber of elements of first layerThe number of times each workpiece i appears is equal to the number of processes c of the workpiece_iAnd all the occurrence times are represented by i, the process of the workpiece i is determined by the sequence of the numbers i appearing in the process layer, if the number i appears for the first time, the process is the first process of the workpiece i, and the workpiece i is randomly inserted into the first layer according to the process number of the workpiece to finish the random arrangement of the process. And the second layer randomly selects the serial number of one machine tool from the machine tools meeting the processing conditions according to the working procedures of the workpiece corresponding to the first layer and puts the serial number into the position.

Step six: performing teacher 'teaching' stage of the algorithm;

general student X^j(j＝1，2，3，...，15) Substituting into an objective function formula, respectively solving W_j(W_jFor time-optimized and energy-optimized integrated values, (j ═ 1, 2, 3.., 15)), when the student is X_iWhen min (W) ═ W_iTaking X_iAs X_teacher(ii) a When the student is X_hWhen W is_hNearest mean (W), take X_hAs the Mean.

Wherein:

W_j＝u₁(FM)_j+u₂(WM)_j(14)

$mean\left(W\right)=\underset{i=1}{\overset{15}{\Σ}}{u}_1{\left(FM\right)}_j+u_2{\left(WM\right)}_j/15---\left(15\right)$

the specific "teaching" process formula is as follows:

$X_{new}^j=X_{old}^j+r_j(X_{teacher}-{\mathrm{TF}}_j\·Mean)---\left(16\right)$

andrespectively representing parameters before and after the optimization of the jth scheduling scheme, namely the processing arrangement and TF of each part_jIs a Teaching Factor (TF)_j＝round[1+rand(0，1)])，r_jFor learning the step length (r)_jRand (0, 1)), these two more critical parameters are used to adjust the speed of learning.

Respectively generating students after the teaching processPerforming discretization processing and outputtingInstead of the former

Step seven: performing a learning process of the algorithm;

in the process of student learning each other, each scheduling scheme X^h(h ═ 1, 2, 3.., 15.) a learning object X was randomly selected from all scheduling schemes^g(g ≠ 1, 2, 3.., 15 wherein h ≠ g), X^hBy and X^gComparing the difference between the two, X^hAnd carrying out corresponding optimization adjustment, wherein the specific chemical process is expressed by the following formula:

when f (X)^h)≤f(X^g) Time of flight

When f (X)^h)＞f(X^g) Time of flight

Wherein r is_hIndicates the learning step size, r, of the h-th student_hRan (0, 1), the generated new scheduling schemeThe discretization is also needed, and the specific discretization process is the same as the teacher's teaching' process.

Step eight: updating the algorithm if W 'produced by "science" process'_newIs less than W_newW'_newReplace W_newAnd simultaneously replacing X with the scheduling scheme X'. Adding 1 to B, judging whether B is more than 500, if so, terminating the algorithm, and carrying out the ninth step; otherwise, jumping to the step six.

Step nine: and decoding the generated scheduling scheme, performing decompiling according to the definition of the student double-layer during encoding, and sequentially showing the processing machine tool and the processing sequence of the workpiece.

Step ten: outputting the current static scheduling scheme X, uploading the scheme to a workshop display through an intranet, verifying workshop production personnel and management personnel, if the scheme has a problem, manually turning to the step three to adjust the parameters, regenerating the scheduling scheme, ensuring the rationality of the scheduling scheme, and if the scheme has no problem, processing according to the scheme.

The off-line scheduling optimization system built by the invention is used for off-line scheduling manufacturing workshop machine tool processing tasks, can realize static scheduling and scheduling management of manufacturing workshop production, and can find the best feasible scheduling scheme meeting the requirements in a shorter specified time.

Claims (3)

1. A teaching and learning algorithm-based static scheduling optimization system for a discrete manufacturing workshop is characterized by comprising a data server, an arithmetic server and a terminal display screen,

the data server collects the processing energy consumption, time consumption and standby conditions of the machine tool in the manufacturing workshop, counts the processing conditions of the workshop in real time, tracks the production energy consumption data of the machine tool in the workshop in real time, and counts the processing conditions of the machine tool and the attendance conditions of workers at the same time;

the operation server is embedded with a scheduling optimization client, the scheduling optimization client performs reasonable scheduling arrangement on the processing tasks on the same day through an optimization algorithm by analyzing the processing tasks and referring to various detailed information of the data server, the optimization algorithm specifically adopts a teaching and learning optimization algorithm, and the construction method of the scheduling optimization client is as follows:

step one, determining an objective function of scheduling optimization of a discrete manufacturing workshop;

the minimum maximum processing time and the minimum total energy consumption of a machine tool in a manufacturing workshop are taken as objective functions, the two objective functions are fused by adopting a weighting method, and when the fused value is minimum, the scheduling requirement is met, namely

FM＝min(max(FM_i)) (1)

$WM=\underset{i=1}{\overset{n}{\Σ}}{\mathrm{wm}}_i---\left(2\right)$

min(W)＝u₁FM+u₂WM (3)

FM_i-all the work in the machine tool M_iA completion time of the above;

FM-Final finish time of all workpieces;

wm_i-all the work in the machine tool M_iThe consumption of energy of;

WM-Total energy consumption of all machine tools;

i-is a positive integer variable greater than 0;

n is the total amount of the machine tool;

min (W) -the final weighting values of the two objective functions;

u₁、u₂-two weighted values, which are valued according to the actual needs of the enterprise;

during the production, the scheduling optimization algorithm needs to satisfy the following constraint conditions:

1.1, a certain procedure can only use the same machine to process at a certain moment;

1.2, once a process of a certain workpiece starts to process, the process cannot be terminated randomly before the process is completed;

1.3, the processing priorities of all the workpieces are the same at zero time;

1.4, the priority of the processing sequence among the procedures of different workpieces is the same, and the processing of the procedures of the same workpiece must be in accordance with the previously appointed processing sequence;

step two, solving the objective function by adopting an improved teaching and learning optimization algorithm, wherein the specific solving process comprises the following steps:

step 2.1, encoding: each scheduling scheme represents a student, the scores of the students are defined in a double-layer design mode, and the first layer isDefining a process layer which represents each process of the ith part; second layerDefining a machine layer which represents a machine during the process machining of the ith part;

step 2.2, parameter initialization: initializing students in the teaching and learning optimization algorithm according to the number of machine tools and the processing tasks of parts;

step 2.3, solving an objective function based on an improved teaching and learning algorithm, and outputting an optimal scheduling scheme;

the running process of the improved teaching and learning algorithm comprises a teacher teaching process, a student learning process and an updating stage;

a. in the teacher's teaching process, the student is assumed to be X^jThe final weight of the two objective functions is defined as f (X)ⁱ)，

j 1, 2, 3,., N, i ∈ {1234 … N }, where N is the total number of students, and min (f (X) is selectedⁱ) Student as teacher X_teacherSelecting the nearestStudent of average value of f (X) as Mean value; the "teach" process formula is as follows:

$X_{new}^j-X_{old}^j+r_j(X_{teacher}-{\mathrm{TF}}_j\·Mean)---\left(4\right)$

wherein,andrespectively representing parameters before and after the optimization of the jth scheduling scheme, namely the processing arrangement and TF of each part_jAs a teaching factor, r_jFor learning step length, the two parameters are used for adjusting the learning speed; at this time, the scheduling scheme obtained by floating point operation may have decimal or over-interval number, and the number needs to be adjustedDiscretizing, wherein the specific discretizing process comprises the following steps:

a.1, firstly, rounding the numbers of the process layer and the machine layer, neglecting decimal, then automatically setting the number exceeding the interval to zero, comparing with the scheduling task list of the current day, marking the workpieces with redundant processes at the process layer, and taking l₁、l₂、…、l_cMarking the workpiece number, wherein c ∈ p and p are workpiece number sets to be marked, and finally randomly picking out the same redundant number from the workpiece number according to the number of redundant processes to be set as zero, and simultaneously setting the corresponding part of the machine layer as zero;

a.2, rearranging the zero-set workpiece numbers, comparing the workpiece numbers with a scheduling task list again, marking the workpiece numbers with insufficient working procedure quantity, randomly generating a group of scheduling schemes according to the specific missing number of the workpieces, sequentially filling the schemes to the positions of the working procedure layer with zero according to the sequence, and simultaneously randomly filling the machine numbers meeting the processing requirements in the positions corresponding to the machine layer;

a.3, taking the scheduling scheme generated for the first time as an initial scheme, carrying out local iteration, setting iteration times, skipping to a.2, if the generated scheduling scheme is superior to the initial scheme, setting the scheduling scheme as the initial scheduling scheme, and outputting the scheduling scheme until the iteration times are met;

b. in the student's learning process, each scheduling scheme X^hRandomly selecting a learning object X from all scheduling schemes^gH 1, 2, 3, ·, N; 1, 2, 3,. N; wherein h is not equal to g; x^hBy and X^gComparing the difference between the two, X^hAnd carrying out corresponding optimization adjustment, wherein the learning process is represented by the following formula:

when f (X)^h)≤f(X^g) Time of flight

When f (X)^h)＞f(X^g) Time of flight

Wherein r is_hRepresenting the learning step length of the h-th student, and generating a new scheduling schemeThe dispersion is also needed, and the specific dispersion process is the same as that in the teacher's teaching process;

c. the updating stage means that after the learning process, the generated scheduling scheme needs to be updated and optimized, and the updating method is as follows:

when in useTime of flight

When in useTime of flight

Wherein, student d belongs to {1, 2, 3, …, N };

the scheduling optimization client decodes the generated scheduling scheme, performs decompiling according to the definition of the student double-layer during encoding, and sequentially shows the processing machine tool and the processing sequence of the workpiece; and finally, displaying the optimized scheduling scheme on a terminal display screen, so that workshop production personnel and management personnel can know the current specific processing task arrangement condition in real time.

2. The system of claim 1, wherein the teaching factor TF is used in the teacher's teaching process_j＝round[1+rand(0，1)]Rand (0, 1) denotes a random number randomly generated from 0 to 1, round denotes a rounding function so that the value taken is an integer, and the learning step r is an integer_i＝rand(0，1)。

3. The system of claim 1, wherein the learning step length r of the h-th student is determined by learning step length r of the h-th student during learning of the h-th student_hRand (0, 1), and rand (0, 1) means that a random number is randomly generated from 0 to 1.