CN112862207B - Task scheduling solving method aiming at unknown machine adjustment time and related sequences - Google Patents
Task scheduling solving method aiming at unknown machine adjustment time and related sequences Download PDFInfo
- Publication number
- CN112862207B CN112862207B CN202110238991.7A CN202110238991A CN112862207B CN 112862207 B CN112862207 B CN 112862207B CN 202110238991 A CN202110238991 A CN 202110238991A CN 112862207 B CN112862207 B CN 112862207B
- Authority
- CN
- China
- Prior art keywords
- task
- test
- sequence
- air conditioner
- tasks
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 24
- 238000012545 processing Methods 0.000 claims abstract description 21
- 238000012360 testing method Methods 0.000 claims description 137
- 238000002474 experimental method Methods 0.000 claims description 59
- 230000006870 function Effects 0.000 claims description 10
- 238000003066 decision tree Methods 0.000 claims description 9
- 238000004458 analytical method Methods 0.000 claims description 8
- 230000008569 process Effects 0.000 claims description 8
- 230000005611 electricity Effects 0.000 claims description 7
- 238000013101 initial test Methods 0.000 claims description 7
- 238000012216 screening Methods 0.000 claims description 4
- 238000012163 sequencing technique Methods 0.000 claims description 4
- 238000004140 cleaning Methods 0.000 claims description 3
- 238000013100 final test Methods 0.000 claims description 3
- 238000012935 Averaging Methods 0.000 claims description 2
- 238000004378 air conditioning Methods 0.000 claims description 2
- 238000002922 simulated annealing Methods 0.000 claims description 2
- 238000004519 manufacturing process Methods 0.000 description 16
- 238000005457 optimization Methods 0.000 description 8
- 230000000875 corresponding effect Effects 0.000 description 6
- 238000003754 machining Methods 0.000 description 5
- 230000008859 change Effects 0.000 description 3
- 238000010586 diagram Methods 0.000 description 2
- 238000005265 energy consumption Methods 0.000 description 2
- 230000007613 environmental effect Effects 0.000 description 2
- 238000002360 preparation method Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 1
- 230000002596 correlated effect Effects 0.000 description 1
- 238000007418 data mining Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 238000005315 distribution function Methods 0.000 description 1
- 238000010438 heat treatment Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000009776 industrial production Methods 0.000 description 1
- 238000009434 installation Methods 0.000 description 1
- 238000007726 management method Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000013433 optimization analysis Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
- 239000002699 waste material Substances 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
- G06N20/20—Ensemble learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0631—Resource planning, allocation, distributing or scheduling for enterprises or organisations
- G06Q10/06313—Resource planning in a project environment
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/04—Manufacturing
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Human Resources & Organizations (AREA)
- Theoretical Computer Science (AREA)
- Strategic Management (AREA)
- Economics (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Entrepreneurship & Innovation (AREA)
- General Business, Economics & Management (AREA)
- Tourism & Hospitality (AREA)
- Marketing (AREA)
- Game Theory and Decision Science (AREA)
- Operations Research (AREA)
- Software Systems (AREA)
- Quality & Reliability (AREA)
- Development Economics (AREA)
- Computing Systems (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Primary Health Care (AREA)
- Manufacturing & Machinery (AREA)
- Mathematical Physics (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biodiversity & Conservation Biology (AREA)
- Medical Informatics (AREA)
- Educational Administration (AREA)
- Evolutionary Computation (AREA)
- Data Mining & Analysis (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Artificial Intelligence (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a task scheduling solving method aiming at unknown machine adjusting time and related sequences, which comprises the following steps: constructing a characteristic variable set and a prediction model related to machine adjustment time before task processing; determining the constraint of the problem to be solved, and coding according to the actual characteristics of the problem to obtain a feasible solution of the problem; randomly generating an initial scheduling sequence of tasks, predicting machine adjustment time of each task in the current task sequence, and calculating an objective function value in the initial scheduling sequence; generating a new task scheduling sequence, predicting the machine adjustment time of each task in the current task sequence again, and calculating an objective function value in the new task sequence; and selecting an initial task sequence solution in the next iteration through comparison of the target function increment, and repeating the iteration until a preset termination condition is met to finally obtain the optimal scheduling sequence of the tasks. The method can quickly realize the optimal solution of task scheduling aiming at the task scheduling with unknown machine adjusting time and sequence correlation.
Description
Technical Field
The invention relates to the field of task scheduling solving, in particular to a hybrid algorithm combining an ensemble learning algorithm and a heuristic algorithm, which is suitable for solving task scheduling problems with unknown machine adjustment time and related sequences.
Background
The manufacturing industry is the prop industry of national economy, and most manufacturing industries begin to transform and upgrade towards intelligent manufacturing and intelligent factories. With the continuous improvement of the informatization degree of enterprise workshop equipment, the data which can be used for production optimization analysis is continuously increased, and data support is provided for research and operators to master uncertain factors in the production process.
During the production process, one of the larger uncertainties is the machine adjustment time before the task is processed, such as the "on-condition" time in a test task, or the preheating time in heat treatment. Wherein the adjustment time of a number of machines is also highly correlated to the task sequence: if the sequence of the machining tasks is changed, the adjustment time of the machine will vary considerably. In conventional production, this time estimate is based on the experience of the technician, with a large uncertainty factor. This brings a great obstacle to operation optimization such as JIT production, estimation of delivery time, and full use of time-of-use electricity price policy.
As a classical problem in the field of industrial production, task scheduling has been studied for a long time by many scholars and many common problems and solutions have been proposed. With the development of modern society, the traditional solution method applied to the part with personalized requirements in many common problems can not work. For example, in the task scheduling process, the problems that the machine adjustment time is uncertain and the sequence is related are often encountered, in the traditional solution, the adjustment time is often set to obey a certain specific distribution function for processing, and a certain difference exists between the actual condition and the adjustment time, so that the scheduling result cannot meet the actual requirements of enterprises; for example, production scheduling problems mainly aiming at reducing energy consumption cost of enterprises are also concerned by many enterprises, but due to the limitation of personnel experience or the uncertainty of some parameters in tasks, the enterprises cannot obtain an optimal task scheduling sequence, and the energy consumption cost of the enterprises cannot be effectively reduced.
The prior art has the following defects:
1. because the machine adjustment time of the tasks to be processed is unknown (the change curve of the machine adjustment time cannot be described by conventional distribution and needs to be predicted according to historical data) and is related to the task sequence, workers cannot accurately master the task completion time, the task sequencing mostly depends on experience in actual situations, the task scheduling scheme cannot reach the lowest time or cost, and waste and low efficiency in the production process of enterprises are caused.
2. How to effectively apply a data prediction method to solve the problem of parameter uncertainty in mathematical programming is a key field of many experts in recent years. At present, most of research and application is that prediction and optimization are set into two independent modules, the solution idea is that prediction is completed firstly and then optimization is performed, and the two stages are less related.
Disclosure of Invention
Aiming at the problem of task scheduling with unknown machine adjustment time and related sequences, the invention aims to provide a task scheduling solving method with unknown machine adjustment time and related sequences, which can quickly realize the optimization solution of task scheduling.
In order to realize the task, the invention adopts the following technical scheme:
a method for adjusting a time-unknown and sequence-dependent task scheduling solution for a machine, comprising:
step 1, screening out a characteristic variable set related to machine adjustment time before task processing according to enterprise historical data; constructing and storing a prediction model about the machine adjustment time before task processing;
step 2, determining the constraint of the problem to be solved, and coding according to the actual characteristics of the problem to obtain a feasible solution of the problem;
step 3, randomly generating an initial scheduling sequence of the obtained tasks, predicting the machine adjustment time of each task in the current task sequence, and calculating an objective function value in the initial scheduling sequence;
step 4, generating a new task scheduling sequence, predicting the machine adjustment time of each task in the current task sequence again, and calculating an objective function value in the new task sequence;
step 5, comparing the current task sequence with the target function increment in the previous task sequence;
step 6, if the new task scheduling sequence is selected to be accepted as a new solution, the new task scheduling sequence is used as an initial task sequence solution in the next iteration;
otherwise, the original initial scheduling sequence is continuously adopted as the initial task sequence solution in the next iteration;
and 7, repeating the iteration steps 3-6 until a preset termination condition is met, and finally obtaining the optimal scheduling sequence of the tasks.
Further, the prediction model in the step 1 is a prediction model established based on an ensemble learning algorithm; and 3, predicting the machine adjustment time of each task in the current task sequence by adopting an ensemble learning algorithm in the step 4, and generating a new task scheduling sequence by adopting a heuristic algorithm in the step 4.
Further, the step 1 comprises:
step 1.1, after historical data provided by an enterprise are cleaned, analyzed for exploratory property, relevance and significance, a characteristic variable set of machine adjustment time before task processing is screened out by combining with feasibility analysis of a predicted task and actual business of the enterprise;
step 1.2, constructing a sub-prediction model set with Z number of sub-prediction models of the ensemble learning algorithm as a prediction model according to the screened feature variable set;
and 1.3, solving the average value of the predicted values of all the sub-prediction model sets, namely the predicted output value of the final machine adjustment time P.
Further, the step 2 comprises:
step 2.1, setting the task scheduling sequence of the problem to be solved as UwObjective function value of (U) minw)；
Step 2.2, determining the constraint of the problem to be solved, which at least comprises the following steps:
constraint 1:
wherein M ═ {1,2, …, M } represents a set of machines; n ═ {1,2, …, N } represents a set of tasks;
representing a variable of 0-1, if the machine k successively processes the tasks i and j to be 1, or else 0;
constraint 2:
constraint 3:
wherein n +1 represents the end state of the machine;
constraint 4:
constraint 5:
wherein, tciRepresenting the completion time of task i; sijIndicating the adjustment time, t, of the same machine for successive machining tasks i and jjRepresents the processing time of task j;
constraint 6:
wherein, tc0Indicating that each machine is available in its initial state, s0jIndicating that no adjustment is required for the first task of machining.
And 2.3, coding according to the characteristics of the problem to be solved to obtain a feasible solution of the problem, namely the scheduling sequence of the tasks.
Further, when the solving method is applied to an air conditioner test scheduling task, the step 1 includes:
cleaning, exploratory, correlative and significance analysis are carried out on a plurality of representative actual air conditioner test data provided by an air conditioner test experiment center, and a 9-dimensional characteristic variable set Q (Q) related to the working condition time P of the test task is obtained by combining the feasibility analysis of the prediction task and the actual business of an enterprise1,q2,…,q9Get the working conditions for the current testing task respectivelyTemperature q of dry bulb in front experiment table1And the temperature q of the wet ball in the experiment table before the current test task is worked2And the temperature q of the outer dry bulb of the experiment table before the current test task is worked3Temperature q of outer wet ball of experiment table before working condition of current test task4The current test task requires the temperature q of the internal dry bulb5The current test task requires the internal wet bulb temperature q6The current test task requires the outer dry bulb temperature q7The current test task requires the temperature q of the outer wet ball8Experiment table refrigerating capacity q9；
According to the screened feature variable set Q, a decision tree set { T) with the quantity of Z related to the XGboost ensemble learning algorithm is constructed1(q),T2(q),…,Tz(q) }, the working condition time value predicted by each decision tree is Tz(q)；
The working condition time values predicted by all the decision tree sets are averaged to obtain the predicted output value of the working condition time P of the final test task
Further, when the solving method is applied to an air conditioner test scheduling task, the step 2.2 includes:
determining the scheduling problem constraint of the air conditioner test task, which is shown as the following formula;
constraint 1:
wherein, M ═ {1,2, …, M } represents a set of benches; n ═ {1,2, …, N } represents a set of air conditioning test tasks;
representing a variable of 0-1, if the test tasks i and j of the experiment table k are 1 successively, or else being 0;
constraint 2:
constraint 3:
wherein n +1 represents the termination state of the test stand;
constraint 4:
constraint 5:
wherein, tciRepresenting the completion time, s, of the air conditioner test task iijRepresenting the working condition time t of the same experiment table for successively testing the tasks i and jjRepresenting the testing time of the air conditioner testing task j;
constraint 6:
wherein, tc0Indicates that each experimental bench can be used in its initial state, s0jThe first task test is carried out without adjustment;
constraint 7:
wherein d represents the air conditioners to be tested in the air conditioner set to be tested, ldIndicating the task number of the air conditioner to be tested;
constraint 8:
wherein,
when the d-th air conditioner to be tested is tested on the experiment table k, an enthalpy difference experiment task is firstly carried out, and then an air conditioner assembling task is carried out, HdRepresenting a d-th air conditioner task set to be tested;
constraint 9:
wherein, adRepresenting the assembly task of the d-th air conditioner to be tested;
representing a variable of 0-1, if task i is tested as 1 on a test bench k, otherwise 0; bd,l-1Showing the l-1 th enthalpy difference experimental task of the d-th air conditioner to be tested.
Further, when the solving method is applied to an air conditioner test scheduling task, the step 2.3 includes:
(1) the air conditioner test task scheduling problem comprises 2 sub-problems: allocating a laboratory bench and sequencing air conditioner test tasks; a two-section coding mode is adopted, and 2 subproblems are coded together to represent a feasible solution of the problem;
(2) the encoding mode distributed by the experiment table is plug-in encoding, a plurality of task sets are randomly sequenced and then a number is inserted into any position in the middle of the task sets, so that a plurality of air conditioners to be tested can be distributed to two experiment tables for testing, wherein redundant cost punishment is caused if tasks on the experiment tables are distributed unevenly;
(3) sectional coding is adopted among all tested air conditioners, integer coding is adopted for testing task coding, and each testing task is replaced by a number.
Compared with the prior art, the invention has the following technical characteristics:
1. the invention provides a novel hybrid solving algorithm aiming at the task scheduling problem that the machine adjusting time is unknown and is related to a production sequence in the production process. The algorithm mainly depends on production big data of enterprises instead of manual experience to realize the optimal solution of the task scheduling problem.
2. The invention applies the means of combining data mining and mathematical programming, correspondingly provides a hybrid algorithm combining an integrated learning algorithm and a heuristic algorithm, can accurately predict the machine preparation time before task processing by the data prediction means, and solves the problem by adopting a parallel processing mode through the integrated learning algorithm and the heuristic algorithm in the algorithm iteration process, thereby accelerating the solving speed of the problem and improving the production efficiency and the management level of enterprises.
Drawings
FIG. 1 is a schematic flow diagram of the process of the present invention;
FIG. 2 is a flow chart of the scheduling of the method of the present invention applied to an embodiment of an air conditioner test task;
FIG. 3 is a graph comparing the results of the application of the examples of the present invention.
Detailed Description
The invention is mainly applied to the task scheduling problem that the machine adjustment time is unknown and is related to the sequence, and the problem can be specifically described as follows: in the production and manufacturing process of an enterprise, one or more machines with all tasks can be processed, all the machines can be processed when in an initial state, and once the processing is started, the operation can not be interrupted halfway; the processing requirements of each task are known and determined, and the time for adjusting the machine is unknown and related to the processing sequence of the tasks when each machine processes different tasks; finally, the task processing sequence on each machine is determined to achieve the optimization goal.
The main factors restricting the problem solving are that the machine adjustment time before task processing is uncertain and related with the sequence, for solving the problem, if the current commonly used solving thought is adopted, all feasible task scheduling sequences need to be listed, and the machine adjustment time of each task under all the sequences needs to be predicted, so that the task calculation amount is increased, the problem becomes more complex, and the problem needs to be further solved through an effective algorithm.
The invention provides a task scheduling solving method aiming at unknown machine adjustment time and relevant sequences, which is a mixed algorithm combining an integrated learning algorithm and a heuristic algorithm, and the algorithm is further explained in detail below.
As the machine adjustment time is unknown and the scheduling problem related to the sequence has higher complexity, aiming at the common problem of the enterprises, the hybrid algorithm combining the integrated learning algorithm and the heuristic algorithm provided by the invention is mainly divided into two parts: the problem is solved by adopting a parallel processing mode between the data prediction part based on the ensemble learning algorithm and the solving part based on the heuristic algorithm, and the solving efficiency and the solving precision are greatly improved.
The specific flow of the algorithm and the symbols of relevant variables appearing in the flow are as follows:
step 1, screening out a characteristic variable set Q related to machine adjustment time P before task processing according to enterprise historical data; and constructing and storing an integrated learning algorithm prediction model of the machine adjustment time P before task processing.
And 2, determining the constraint of the problem to be solved, and coding according to the actual characteristics of the problem to obtain a feasible solution of the problem.
Step 3, randomly generating and obtaining an initial scheduling sequence U of tasks0Predicting the machine regulation time P of each task in the current task sequence by using an ensemble learning algorithm, and obtaining the time P according to the stepsj＝sij(ii) a Calculating an objective function value f (U) in an initial scheduling order0)。
Step 4, generating a new task scheduling sequence U by utilizing a heuristic algorithm1And calling the ensemble learning algorithm again to predict the machine adjustment time P of each task in the current task sequence, and calculating an objective function value f (U) in the new task sequence1)。
And 5, comparing the increment df of the target function between the current task sequence and the previous task sequence to be f (U)1)-f(U0)；
Step 6, if the selection is accepted the new task sequence U1As a new solution, then the new solution U1As the initial task sequential solution U for the next iteration of the algorithm0＝U1(ii) a If no new task sequence U is accepted1As a new solution, then the original initial task sequence U is continued to be adopted0As the initial task sequential solution U for the next iteration of the algorithm0＝U0。
And 7, repeating the iteration steps 3-6 until a preset termination condition of the algorithm is met, and finally obtaining the optimal scheduling sequence of the tasks.
Wherein, the step 1 specifically comprises:
step 1.1, after historical data provided by an enterprise are cleaned, analyzed for exploratory performance, relevance and significance, and combined with the feasibility analysis of a prediction task and the actual business of the enterprise, a characteristic variable set Q (Q) related to machine adjustment time P before task processing is screened out1,q2,…,qg}。
Step 1.2, constructing a sub-prediction model set { T) with Z sub-prediction models of the ensemble learning algorithm according to the screened feature variable set Q1(q),T2(q),…,Tz(q) }, wherein the machine adjustment time prediction value of the z-th sub-prediction model is Tz(q)。
Step 1.3, calculating the average value of all the sub-prediction model set prediction values to obtain the prediction output value of the final machine adjustment time P
In the above technical solution, the step 2 specifically includes:
step 2.1, setting the task scheduling sequence of the problem to be solved as UwObjective function value of (U) minw)。
Step 2.2, determining the constraint of the problem to be solved, wherein the constraint at least comprises the following constraints:
constraint 1:
this formula indicates that each task can only be processed once on one machine, and cannot be interrupted halfway or the machine cannot be replaced.
Constraint 2:
this equation determines the number of participating first processing tasks.
Constraint 3:
this equation determines the number of participating last processing tasks.
Constraint 4:
this equation indicates that the tasks have a deterministic precedence relationship in the corresponding machining sequence.
Constraint 5:
this equation represents the relationship of the completion time between two successive tasks in each task processed by the machine.
Constraint 6:
this formula indicates that each machine is available in its initial state and that no adjustments are required for the first job of machining.
And 2.3, coding according to the characteristics of the problem to be solved to obtain a feasible solution of the problem, namely the scheduling sequence of the tasks.
Example (b):
the algorithm is applied to the scheduling problem of air conditioner test tasks in the air conditioner test industry, and the total power consumption of a laboratory bench is further reduced by reasonably arranging the air conditioner test tasks, so that the effectiveness of the algorithm is demonstrated.
Introduction of application background:
before a new air conditioner is put into the market, a large number of enthalpy difference experiments are required to be carried out according to different environment working conditions, wherein the preparation ("working condition working") time of the enthalpy difference experiments is difficult to determine and can change along with the change of the test task sequence, an enterprise has a plurality of experiment tables for air conditioner test tasks, and how to determine the optimal task test sequence of the air conditioner on each experiment table and the task working condition working time under the current sequence ensures that the power consumption cost of the experiment tables is minimum, and the specific application flow and relevant variable symbols appearing in the flow are as follows:
step 1, screening out a characteristic variable set Q related to task working condition time P according to enterprise historical data; constructing and storing an XGboost ensemble learning algorithm prediction model related to working condition time P before task testing, wherein the specific process is as follows:
step 1.1, after cleaning, exploratory, correlation and significance analysis are carried out on a plurality of representative actual air conditioner test data provided by the air conditioner test experiment center, and after the feasibility analysis of a prediction task and the actual business of an enterprise is combined, a 9-dimensional characteristic variable set Q (Q) related to the working condition working time P of the test task is finally obtained1,q2,…,q9Get the dry ball temperature q in the experiment table before the working condition is set for the current testing task respectively1And the temperature q of the wet ball in the experiment table before the current test task is worked2And the temperature q of the outer dry bulb of the experiment table before the current test task is worked3Temperature q of outer wet ball of experiment table before working condition of current test task4The current test task requires the temperature q of the internal dry bulb5The current test task requires the internal wet bulb temperature q6The current test task requires the outer dry bulb temperature q7The current test task requires the temperature q of the outer wet ball8Experiment table refrigerating capacity q9。
Step 1.2, according to the screened characteristic variable set Q, a decision tree set (sub-prediction model set) with Z number related to the XGboost ensemble learning algorithm is constructed1(q),T2(q),…,Tz(q) }, the working condition time value predicted by each decision tree is Tz(q)。
Step 1.3, averaging the working condition time values predicted by all decision tree sets to obtain the predicted output value of the working condition time P of the final test task
Step 2, determining the constraint of an enterprise for carrying out an air conditioner test task, and coding the problem to obtain a feasible solution about the scheduling sequence of the air conditioner test task; the specific process is as follows:
step 2.1, setting an objective function value minf (U) of the air conditioner test scheduling problemw) (ii) a The optimal air conditioner test task sequence is found, so that the total power consumption and electricity consumption cost of the experiment table is the lowest;
step 2.2, determining the scheduling problem constraint of the air conditioner test task, as shown in the following formula;
constraint 1:
this formula indicates that each task can only be tested once on a laboratory bench, and the laboratory bench cannot be interrupted halfway or replaced.
Constraint 2:
this formula represents the determination of the number of participation in the first test task.
Constraint 3:
this formula represents the determination of the number of participation in the last test task.
Constraint 4:
this equation indicates that the tasks have a deterministic precedence relationship in the corresponding test sequence.
Constraint 5:
this equation represents the elapsed time relationship between two successive test tasks in each of the tasks tested at the test station.
Constraint 6:
this formula indicates that each bench is usable in its initial state and does not require adjustment for the first task test.
Constraint 7:
the equation represents the equality relationship between the number of air conditioner test task sets, the number of air conditioners to be tested and the total test task.
Constraint 8:
the formula shows that the air conditioner needs to carry out an installation task before carrying out an enthalpy difference experiment task, and the air conditioner to be tested is installed on a corresponding experiment table.
Constraint 9:
the formula indicates that the air conditioner cannot be disassembled once being installed on a corresponding experiment table until all the test tasks of the air conditioner are completed;
and 2.3, coding according to the problem characteristics to obtain a feasible solution of the problem, namely the scheduling sequence of the test tasks, and specifically comprises the following steps:
(1) this embodiment contains 2 sub-problems: and (4) allocating the experiment table and sequencing air conditioner test tasks. Therefore, the present invention uses a two-segment coding scheme to code 2 sub-problems together to represent a feasible solution to the problem. Assuming that 4 air conditioners (1-4) to be tested are distributed to 2 test benches for testing, the number of testing tasks of the 4 air conditioners is respectively 5, 6, 4 and 5 (including the assembly task of installing the air conditioners to the test benches before testing), one possible solution obtained according to the coding mode is as follows:
(2) the encoding mode distributed by the experiment table is plug-in encoding, 4 air conditioners to be tested can be distributed to two experiment tables to be tested by inserting a number (-1) into any position in the middle after 4 task sets are randomly sequenced, and redundant cost punishment is caused if tasks on the experiment tables are unevenly distributed.
(3) The number of testing tasks performed by each tested air conditioner is not equal, according to actual conditions, each air conditioner to be tested can not be disassembled after being installed on a corresponding experiment table until all testing tasks are completed, and all testing tasks can only be completed on the current experiment table, so that sectional coding is adopted among the tested air conditioners, integer coding is adopted for testing task coding, and each testing task is replaced by a number. In the above possible solution 5 test tasks are to be performed in the air conditioner 1, and the task to be tested for the air conditioner is then coded as a digital combination of 1-5.
Step 3, solving the problem by using a Simulated Annealing algorithm (SA), and taking the initial temperature T0Is large enough to make T ═ T0Randomly generating and obtaining an initial scheduling sequence U of tasks0Calling XGboost algorithm to predict working condition time P of each task in current test task sequence, and obtaining P according to the stepsj＝sij(ii) a Calculating the total power consumption cost value f (U) of the experiment table under the initial test task sequence0) And repeating the steps 4-7 for the current temperature T.
Step 4, processing sequence U of current task0Random disturbance generates a new air conditioner test task sequence U1And calling the XGboost prediction model again to predict the working condition time P of each test task in the current test task sequence, and calculating the total power consumption f (U) of the experiment table in the new test task sequence1)。
Step 5, comparing the increment df of the power consumption expense of the experiment table under the two task sequences with f (U)1)-f(U0)。
Step 6, if df is less than or equal to 0, selecting to acceptNew test task sequence U1As a new solution, then the new solution U1Sequential solution U of initial test task as next iteration of algorithm0＝U1(ii) a Otherwise, calculating U1Is the probability of acceptance of exp (-df/T), T is the current temperature. Randomly generating a random number rand evenly distributed over the (0,1) interval if exp (-df/T)>rand, also receiving U1As a new current solution U0＝U1Otherwise, keeping the current solution U0I.e. continue to adopt the original initial testing task sequence U0Sequential solution U of initial test task as next iteration of algorithm0＝U0。
Step 7, if the set termination condition is met, outputting the current solution U0As the optimal solution, the routine is ended.
And 8, gradually reducing T, and then turning to the step 3.
And 9, finally obtaining the air conditioner test task scheduling sequence with the minimum total power consumption of the experiment table through the steps.
Experimental results for this example:
fig. 3 is a cost comparison diagram of the daily arrangement scheme of the test tasks of the experiment center and the application scheme of the embodiment, and the invention randomly selects 10 actual cases to respectively calculate the corresponding electric charges of the experiment table for verification. In the arrangement scheme for the daily air conditioner test of the enterprise, the sequence of the test tasks is judged according to the experience of testers, the test sequence cannot reach the optimum, and the difference value of the environmental parameters of the test tasks between two adjacent items is relatively large, so that the working condition time of each test task is generally long, and the power consumption cost of the whole experiment table is at a high level.
According to the scheme, after the proposed mixed algorithm combining the integrated learning algorithm and the heuristic algorithm is applied, the difference value of the environmental parameters of the test tasks between two adjacent items is relatively reduced, the working condition working time of the test tasks is shortened, and the electricity charge of the experiment table is optimized, wherein the maximum electricity charge optimization proportion is 34.3%, the minimum electricity charge optimization proportion is 13.3%, and the average electricity charge optimization proportion is 23%.
The above embodiments are only used to illustrate the technical solutions of the present application, and not to limit the same; although the present application has been described in detail with reference to the foregoing embodiments, it should be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; such modifications and substitutions do not substantially depart from the spirit and scope of the embodiments of the present application and are intended to be included within the scope of the present application.
Claims (1)
1. A method for solving scheduling of tasks with unknown time and sequence correlation aiming at a machine adjustment is characterized by comprising the following steps:
step 1, screening out a characteristic variable set related to machine adjustment time before task processing according to enterprise historical data; constructing and storing an XGboost ensemble learning algorithm prediction model related to the working condition time P before task testing; the prediction model is established based on an integrated learning algorithm; when the solving method is applied to an air conditioner test scheduling task, the step 1 comprises the following steps:
step 1.1, cleaning, exploratory, correlation and significance analysis are carried out on a plurality of representative actual air conditioner test data provided by an air conditioner test experiment center, and a 9-dimensional characteristic variable set Q (Q) related to the working condition time P of a test task is obtained after a prediction task and enterprise actual business feasibility analysis are combined1,q2,…,q9Get the dry ball temperature q in the experiment table before the working condition is set for the current testing task respectively1And the temperature q of the wet ball in the experiment table before the current test task is worked2And the temperature q of the outer dry ball of the experiment table before the current test task is worked out3Temperature q of outer wet ball of experiment table before working condition of current test task4The current test task requires the temperature q of the internal dry bulb5The current test task requires the internal wet bulb temperature q6The current test task requires the outer dry bulb temperature q7The current test task requires the temperature q of the outer wet ball8Experiment table refrigerating capacity q9；
Step 1.2, according to the screeningConstructing a decision tree set { T) with the quantity of Z about the XGboost ensemble learning algorithm by using the obtained characteristic variable set Q1(q),T2(q),…,Tz(q) }, the working condition time value predicted by each decision tree is Tz(q)；
Step 1.3, averaging the working condition time values predicted by all decision tree sets to obtain a predicted output value of the working condition time P of the final test task;
step 2, determining the constraint of an enterprise for carrying out an air conditioner test task, and coding the problem to obtain a feasible solution about the scheduling sequence of the air conditioner test task; the specific process is as follows:
step 2.1, setting an objective function value minf (U) related to the air conditioner test scheduling problemw) (ii) a The optimal air conditioner test task sequence is found, so that the total power consumption and electricity consumption cost of the experiment table is the lowest;
step 2.2, determining the scheduling problem constraint of the air conditioner test task, as shown in the following formula;
constraint 1:
wherein, M ═ {1,2, …, M } represents a set of benches; n ═ {1,2, …, N } represents a set of air conditioning test tasks;
representing a variable of 0-1, if the test tasks i and j of the experiment table k are 1 successively and if not, 0;
constraint 2:
wherein,
representing a variable of 0-1, if the test tasks 0 and j of the experiment table k are 1 successively and if not, 0;
constraint 3:
wherein n +1 represents the termination state test task of the test bed;
representing a variable of 0-1, if the test task j and the n +1 of the experiment table k are 1 successively and otherwise, the test task j and the n +1 are 0;
constraint 4:
wherein,
representing a variable of 0-1, if the test tasks j and i of the experiment table k are 1 successively, otherwise, the test tasks are 0;
constraint 5:
wherein, tciRepresents the completion time of the air conditioner test task i, sijRepresenting the working condition time t of the same experiment table for successively testing the tasks i and jjRepresenting the testing time of the air conditioner testing task j;
constraint 6:
wherein, tc0Indicates that each experimental bench can be used in its initial state, s0jThe method is characterized in that adjustment is not needed when the first task test is carried out;
constraint 7:
d represents the D-th air conditioner to be tested in the set of air conditioners to be tested, and D is {1,2, …, D }; ldThe number of tasks of the air conditioner to be tested is represented;
constraint 8:
wherein,
when the d-th air conditioner to be tested is tested on the experiment table k, an enthalpy difference experiment task is firstly carried out, and then an air conditioner assembling task is carried out, HdRepresenting a task set of the d-th air conditioner to be tested;
constraint 9:
wherein, adShowing the assembly task of the d-th air conditioner to be tested,
represents a variable of 0 to 1;
representing a variable of 0-1, if task o is tested as 1 on the test bench k, otherwise 0; o ═ bd,1、bd,2…bd,l-1；bd,l-1Showing the (l-1) th enthalpy difference experiment task of the d-th air conditioner to be tested;
and 2.3, coding according to the problem characteristics to obtain a feasible solution of the problem, namely the scheduling sequence of the test tasks, and specifically comprises the following steps:
(1) the air conditioner test task scheduling problem comprises 2 sub-problems: allocating a laboratory bench and sequencing air conditioner test tasks; a two-section coding mode is adopted, and 2 subproblems are coded together to represent a feasible solution of the problem;
(2) the encoding mode distributed by the experiment table is plug-in encoding, a plurality of task sets are randomly ordered, and then a number is inserted into any position in the middle of the task sets, so that a plurality of air conditioners to be tested can be distributed to two experiment tables for testing, wherein redundant cost punishment is caused if tasks on the experiment tables are unevenly distributed;
(3) sectional coding is adopted among all tested air conditioners, integer coding is adopted for testing task coding, and each testing task is replaced by a number;
and 3, solving the problem by using a simulated annealing algorithm, and enabling T to be equal to T0，T0Is the initial temperature; randomly generating and obtaining an initial scheduling sequence U of tasks0Calling an XGboost algorithm to predict working condition time P of each task in the current test task sequence; calculating the total power consumption cost value f (U) of the experiment table under the initial test task sequence0) Repeating steps 4-7 for the current temperature T;
step 4, the initial scheduling sequence U is processed0Random disturbance generates a new air conditioner test task sequence U1And calling the XGboost prediction model again to predict the working condition time P of each test task in the current test task sequence, and calculating the total power consumption f (U) of the experiment table in the new test task sequence1)；
Step 5, comparing the increment df of the power consumption expense of the experiment table under the two task sequences with f (U)1)-f(U0)；
Step 6, if df is less than or equal to 0, selecting to receive a new air conditioner test task sequence U1As a new solution, then U will be1As the initial scheduling sequence for the next iteration of the algorithm; otherwise, calculating a new air conditioner test task sequence U1(ii) an acceptance probability of exp (-df/T), T being the current temperature;
randomly generating a random number rand evenly distributed over the (0,1) interval if exp (-df/T)>rand, accept U1As the new current initial scheduling sequence, otherwise, the current initial scheduling sequence U is kept0I.e. continue to adopt the original initial testing task sequence U0As the initial test task sequential solution during the next algorithm iteration;
step 7, if the set termination condition is met, outputting the current initial scheduling sequence U0And finally obtaining the optimal scheduling sequence of the tasks as the optimal solution.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110238991.7A CN112862207B (en) | 2021-03-04 | 2021-03-04 | Task scheduling solving method aiming at unknown machine adjustment time and related sequences |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110238991.7A CN112862207B (en) | 2021-03-04 | 2021-03-04 | Task scheduling solving method aiming at unknown machine adjustment time and related sequences |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112862207A CN112862207A (en) | 2021-05-28 |
| CN112862207B true CN112862207B (en) | 2022-05-13 |
Family
ID=75991647
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110238991.7A Active CN112862207B (en) | 2021-03-04 | 2021-03-04 | Task scheduling solving method aiming at unknown machine adjustment time and related sequences |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112862207B (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104808636A (en) * | 2015-04-28 | 2015-07-29 | 广东工业大学 | Flexible flow shop energy consumption optimization scheduling method |
| CN107862411A (en) * | 2017-11-09 | 2018-03-30 | 西南交通大学 | A kind of extensive flexible job shop scheduling optimization method |
| CN110533301A (en) * | 2019-08-09 | 2019-12-03 | 大连理工大学 | A kind of population dispatching method based on dynamic constrained matrix |
| CN111626496A (en) * | 2020-05-25 | 2020-09-04 | 北京化工大学 | Hybrid optimization method for flexible assembly job shop scheduling |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP2172887A3 (en) * | 2008-09-30 | 2011-11-09 | Rockwell Automation Technologies, Inc. | System and method for dynamic multi-objective optimization of machine selection, integration and utilization |
| CN104391488A (en) * | 2014-11-18 | 2015-03-04 | 广东工业大学 | Optimizing and dispatching method of energy consumption of flexible flow shop with associated adjustment time and sequence |
-
2021
- 2021-03-04 CN CN202110238991.7A patent/CN112862207B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104808636A (en) * | 2015-04-28 | 2015-07-29 | 广东工业大学 | Flexible flow shop energy consumption optimization scheduling method |
| CN107862411A (en) * | 2017-11-09 | 2018-03-30 | 西南交通大学 | A kind of extensive flexible job shop scheduling optimization method |
| CN110533301A (en) * | 2019-08-09 | 2019-12-03 | 大连理工大学 | A kind of population dispatching method based on dynamic constrained matrix |
| CN111626496A (en) * | 2020-05-25 | 2020-09-04 | 北京化工大学 | Hybrid optimization method for flexible assembly job shop scheduling |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112862207A (en) | 2021-05-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Zhou et al. | A multi-objective differential evolution algorithm for parallel batch processing machine scheduling considering electricity consumption cost | |
| CN110543151B (en) | Method for solving workshop energy-saving scheduling problem based on improved NSGA-II | |
| CN109946965B (en) | Discrete manufacturing workshop scheduling method based on improved multi-target Jaya algorithm | |
| CN104808636A (en) | Flexible flow shop energy consumption optimization scheduling method | |
| CN115600774B (en) | Multi-target production scheduling optimization method for assembly type building component production line | |
| CN110458326B (en) | Mixed group intelligent optimization method for distributed blocking type pipeline scheduling | |
| CN105023066A (en) | Business expansion analytical prediction system and method based on seasonal adjustment | |
| Kim et al. | Scheduling a single machine with multiple preventive maintenance activities and position-based deteriorations using genetic algorithms | |
| CN109034479B (en) | Multi-target scheduling method and device based on differential evolution algorithm | |
| CN113313360B (en) | Collaborative task allocation method based on simulated annealing-scattering point hybrid algorithm | |
| CN114912346A (en) | Skill planning configuration and workshop scheduling integrated optimization method based on learning ability | |
| CN111859242A (en) | Household power energy efficiency optimization method and system | |
| CN114707748B (en) | Mixed flow production line intelligent scheduling method based on group immunity-genetic algorithm | |
| CN111291969B (en) | Automobile reordering method based on genetic algorithm | |
| CN112862207B (en) | Task scheduling solving method aiming at unknown machine adjustment time and related sequences | |
| CN113642988B (en) | Multi-station multi-type energy storage power station cost benefit analysis method and shaping system | |
| CN113283755B (en) | Intelligent scheduling decision-making method for workshops | |
| Chamnanlor et al. | Hybrid genetic algorithms for solving reentrant flow-shop scheduling with time windows | |
| CN113592064A (en) | Ring polishing machine process parameter prediction method, system, application, terminal and medium | |
| CN111549193B (en) | Furnace changing method, furnace changing device and control equipment for multiple blast furnace hot blast stoves | |
| CN117540990A (en) | Production line scheduling method based on deep reinforcement learning and multi-objective optimization | |
| CN111290360A (en) | Multi-objective optimization method for casting production line | |
| Staggemeier et al. | A hybrid genetic algorithm to solve a lot-sizing and scheduling problem | |
| CN110705844A (en) | Robust optimization method of job shop scheduling scheme based on non-forced idle time | |
| CN115766475A (en) | Semi-asynchronous power federal learning network based on communication efficiency and communication method thereof |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |