CN107609678B - Homotype parallel machine production scheduling method considering moment information uncertainty - Google Patents

Homotype parallel machine production scheduling method considering moment information uncertainty Download PDF

Info

Publication number
CN107609678B
CN107609678B CN201710717959.0A CN201710717959A CN107609678B CN 107609678 B CN107609678 B CN 107609678B CN 201710717959 A CN201710717959 A CN 201710717959A CN 107609678 B CN107609678 B CN 107609678B
Authority
CN
China
Prior art keywords
model
pmsp
formula
workpiece
expression
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
Application number
CN201710717959.0A
Other languages
Chinese (zh)
Other versions
CN107609678A (en
Inventor
宋士吉
常志琦
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Tsinghua University
Original Assignee
Tsinghua University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Tsinghua University filed Critical Tsinghua University
Priority to CN201710717959.0A priority Critical patent/CN107609678B/en
Publication of CN107609678A publication Critical patent/CN107609678A/en
Application granted granted Critical
Publication of CN107609678B publication Critical patent/CN107609678B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • 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

  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention provides a homotype parallel machine production scheduling method considering moment information uncertainty, and belongs to the field of production scheduling and operation research. Firstly, constructing a distribution set robust optimization model DR-PMSP-MU considering moment information uncertainty, and obtaining an expression of an initial model DR-PMSP-MU1 comprising an inner layer problem and an outer layer problem; after the decision variables of the DR-PMSP-MU model are converted, solving the inner-layer problem, and equivalently converting the DR-PMSP-MU1 model into a solvable model DR-PMSP-MU 2; the optimal solution of the DR-PMSP-MU2 model is the optimal solution of the whole model, the optimal solution corresponds to a plurality of optimal production scheduling schemes, and a decision maker can select the optimal solution according to needs; the model established by the invention better accords with the actual production condition, and the decision risk can be reduced by utilizing more information in the production environment under the condition of ensuring the system performance.

Description

Homotype parallel machine production scheduling method considering moment information uncertainty
Technical Field
The invention belongs to the field of production scheduling and operation research, and particularly relates to a homotype parallel machine production scheduling method considering moment information uncertainty.
Background
The scheduling problem of the parallel machines of the same type considers that a batch of workpieces are processed on a series of same machines, each machine can only process one workpiece at the same time, and each workpiece can only be processed by one machine at the same time. Each workpiece has specific processing time, the scheduling problem is to find a scheduling scheme, and all workpieces are distributed to corresponding machines for processing in a certain processing sequence, so that the performance index of a certain system is optimal. In an actual production manufacturing environment, there is often uncertainty in the processing time of a workpiece due to factors such as machine or tool conditions, worker processing level, and processing environment. And the trend of individual customization of small batches in the manufacturing industry is more and more obvious at present, and the processing time of some new products can only be estimated and obtained through some historical processing data, so that the uncertainty of the processing time of workpieces is continuously enhanced. With such a high degree of uncertainty, if the conventional deterministic model is still used to make scheduling decisions, it may lead to a severe deviation from the expectations of the decision-maker, and may even result in a scheduling scheme that is not feasible at all in practice. Therefore, researchers are beginning to gradually study the scheduling problem of parallel machines of the same type with uncertainty.
In the research of scheduling problems of parallel machines of the same type with uncertainty, two methods are mainly used at present, wherein one method is to establish a random scheduling model, and the other method is to establish a robust scheduling model. In the stochastic scheduling model, uncertain workpiece processing time is regarded as a random variable with known distribution, and the optimization target of the model is often the long-term performance expectation of the system. Although the stochastic scheduling model has a good propulsion function on the research of the uncertain parallel machine scheduling problem in theory, some inherent defects of the stochastic scheduling model limit the application of the stochastic scheduling model in the practical large-scale scheduling problem. These disadvantages are mainly manifested in the following aspects: 1) in the stochastic scheduling model, the distribution of uncertain parameters needs to be known accurately. However, in an actual production environment, as the production operation becomes more complicated and the production modes such as small-lot personalized customization gradually change, in many cases, an accurate probability distribution is difficult to obtain, and the interval range of the product can only be estimated according to the processing time of similar products. This problem is particularly acute in connection with new products, especially single-piece or single-process products. In this case, the stochastic scheduling model will no longer apply. 2) In the stochastic scheduling model, some expectation of system performance is typically employed as an optimization objective. Such goals are more suitable for planning the development of an enterprise over a long period of time and are not suitable for solving the problem of maximizing benefits or minimizing risks during each actual operation. 3) Such stochastic models are generally NP-hard, and can generally be solved only by heuristic algorithms or dynamic programming algorithms, and as the scale of the problem gradually expands, the difficulty in solving the stochastic scheduling model will increase exponentially.
Another robust scheduling model was originally proposed by Richard L, Daniels et al, where only the workpiece processing time for which interval information is known is characterized by interval data scenarios (one scenario represents one possible value of workpiece processing time), which is simpler and more realistic than the description of parameter distribution functions in a stochastic scheduling model.
In the optimization problem where the process parameters have uncertainty, there is a new optimization method called distribution set robust optimization. Although the method is rarely applied to the field of production scheduling at present, the method is widely applied to a plurality of fields of electric power, health, securities and the like, and achieves good effect. In the distribution set robust optimization method, an uncertain parameter is represented by a random variable, but the distribution function of the random variable is unknown and belongs to a certain distribution function set. In the optimization process, all possible distribution functions in the set of distribution functions need to be considered. Although the uncertain parameters are still regarded as a random variable in the distribution set robust optimization model, compared with the random optimization model, the specific form of the distribution function is not required to be determined, and only a set of the distribution function is determined. Compared with the robust optimization model based on the uncertainty set, the robust optimization of the distribution set not only utilizes the information of the parameter variation range, but also takes more information such as the mean value, the variance and the like into consideration so as to reduce the conservative degree of decision.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provides a method for producing and scheduling a parallel machine of the same type by considering moment information uncertainty. The model established by the invention better accords with the actual production condition, and by utilizing more information in the production environment, the decision risk can be reduced under the condition of ensuring the system performance, and the optimal production scheduling scheme can be obtained.
The invention provides a homotype parallel machine production scheduling method considering moment information uncertainty, which is characterized by comprising the following steps of:
1) aiming at the scheduling problem of a same-type parallel machine, a distribution set robust optimization model DR-PMSP-MU considering moment information uncertainty is constructed, and an expression of an initial model DR-PMSP-MU1 is obtained;
in a DR-PMSP-MU model, the performance index of the system is selected as total flow passing time TFT; assuming that all workpieces are released at the moment when the machining starts, namely the release time is 0, the machining time of the workpieces has random uncertainty, the distribution of the random machining time is unknown, but belongs to a distribution set determined by a support set, a sampling mean vector and a sampling covariance matrix; the DR-PMSP-MU model aims to find an optimal robust scheduling scheme, so that the scheduling scheme has the minimum expected TFT under the condition that the workpiece processing time obeys worst distribution;
1-1) determining model decision variables;
decision variables of the DR-PMSP-MU model are feasible scheduling schemes, and include machine selection of all workpieces and processing sequences on each machine; let J workpieces and M machines in the model, the set of workpieces and machines is JS1,2, …, J and MS1,2, … M, a feasible scheduling scheme is represented by a three-dimensional matrix X ∈ {0,1}J×M×J＝{xjml∈{0,1}|j∈JS,m∈MS,l∈LS＝JSRepresents; wherein, if the workpiece j is assigned to the m-th machine,and processed in order of the last but one, then xjml1, otherwise xjml＝0；
1-2) random vector representation of processing time;
the processing time of all the workpieces is a random vector p, the distribution obeyed by the vector is unknown, but belongs to a distribution set determined by a support set, a sampling mean vector and a sampling covariance matrix, and the expression of the distribution set is shown as formula (2):
Figure GDA0002425944090000031
wherein the content of the first and second substances,
Figure GDA0002425944090000032
indicating that the processing time for each workpiece is non-negative,
Figure GDA0002425944090000033
true mean vector, μ, representing the processing time of all workpiecessAnd sigmasRespectively representing the sampling mean vector and the sampling covariance matrix of all the workpiece processing time, wherein gamma is a constraint parameter;
1-3) determining an objective function of a DR-PMSP-MU model;
given a schedule X and all workpiece processing time vectors p, the TFT is calculated by equation (3):
Figure GDA0002425944090000034
in the formula, pjRepresents the processing time of the workpiece j;
the total flow time TFT is a random variable, and the expectation is taken as the measurement of the random TFT to obtain the target function of the DR-PMSP-MU model as
Figure GDA0002425944090000035
Representing the total flow through time expected value obtained when all the workpiece processing time vectors p obey a certain distribution function F;
1-4) determining a constraint condition of a DR-PMSP-MU model;
1-4-1) random processing time constraint, wherein the expression is shown as a formula (2);
1-4-2) feasible scheduling scheme constraints;
each element in the feasible scheduling scheme X is 0 or 1, and the expression is shown in formula (4):
Figure GDA0002425944090000041
1-4-3) workpiece occupation position constraint;
each workpiece only occupies one position on one machine, and the expression is shown in formula (5):
Figure GDA0002425944090000042
1-4-4) the position is occupied and restricted by the workpiece;
each position on each machine is occupied by at most one workpiece, and the expression is shown in the formula (6):
Figure GDA0002425944090000043
1-4-5) ordering compact constraints;
the occupied positions on each machine are consecutive and starting from 1, the expression is given by the equation (7):
Figure GDA0002425944090000044
the latter four kinds of constraints shown as formula (4) -formula (7) are all feasible for restricting the scheduling scheme, and are integrated together to form a feasible domain of the scheduling scheme
Figure GDA0002425944090000045
As shown in formula (8):
Figure GDA0002425944090000046
1-5) establishing an expression of a same-type parallel machine scheduling distribution set robust initial model DR-PMSP-MU1 considering moment information uncertainty, as shown in formula (9):
Figure GDA0002425944090000047
in the formula, max is represented in the distribution set
Figure GDA0002425944090000048
In finding the objective function
Figure GDA0002425944090000049
The maximum distribution function F, min represents the feasible region in the scheduling scheme
Figure GDA00024259440900000410
Middle search causes inner layer problems
Figure GDA00024259440900000411
The scheduling scheme X with the minimum optimal value;
2) performing equivalent transformation on the DR-PMSP-MU1 model established in the step 1);
2-1) converting decision variables;
equivalently converting decision variables of the DR-PMSP-MU model from a three-dimensional matrix X into a two-dimensional matrix Y, wherein the conversion relation is shown as a formula (10):
Figure GDA0002425944090000051
the expression of the feasible domain of Y is shown as formula (11):
Figure GDA0002425944090000052
the two-dimensional matrix Y is expressed as a vector pi, and the expression is shown as the formula (12):
Figure GDA0002425944090000053
pi represents the reverse order of the processing sequence of the workpiece after the machine serial number is ignored;
the expression of the feasible domain of pi is shown as formula (13):
Figure GDA0002425944090000054
the TFT is expressed as the inner product of pi and p, and the expression is shown as the formula (14):
f(π,p)＝f(X,p)＝πTp； (14)
2-2) solving an inner layer problem;
after the decision variable X is converted into pi through the step 2-1), the decision variable X is determined according to
Figure GDA0002425944090000055
The inner layer problem in the DR-PMSP-MU1 model shown in equation (9) is simplified as shown in equation (15):
Figure GDA0002425944090000056
in the formula (I), the compound is shown in the specification,
Figure GDA0002425944090000057
representing the real mean vector of all the workpiece processing time by s.t. representing constraint conditions;
solving equation (15) to obtain the optimal solution of
Figure GDA0002425944090000058
The optimal value of the inner layer problem in the corresponding DR-PMSP-MU1 model is
Figure GDA0002425944090000059
2-3) equivalently converting the DR-PMSP-MU1 model into a solvable model DR-PMSP-MU 2;
replacing the inner layer problem in the formula (9) with the optimal value obtained in the step 2-2), equivalently converting the DR-PMSP-MU1 model into a DR-PMSP-MU2 model, wherein the expression of the DR-PMSP-MU2 model is shown in the formula (16):
Figure GDA00024259440900000510
3) solving the DR-PMSP-MU model to obtain an optimal production scheduling scheme;
solving the DR-PMSP-MU2 model shown as the formula (16), wherein the obtained optimal solution of the DR-PMSP-MU2 model is the optimal solution of the DR-PMSP-MU model; the optimal solution is an optimal vector pi value, and the vector pi value is equivalently converted into an optimal two-dimensional matrix Y value containing the processing sequence corresponding to each workpiece through the corresponding relation shown in the formula (12), so that an optimal production scheduling scheme is obtained.
The invention has the characteristics and beneficial effects that: :
1) according to the homotype parallel machine production scheduling method considering moment information uncertainty, a distribution set robust model is established on the basis of a traditional certainty model, only a support set, a sampling mean value and a sampling covariance matrix of random processing time are needed, distribution information of the distribution set robust model does not need to be accurately known, and compared with a random model, the homotype parallel machine production scheduling method is more suitable for actual production and has higher practicability.
2) Compared with a robust scheduling model only using an interval change range, the distribution set robust model adopted by the invention has smaller conservative property. By utilizing more information, the obtained optimal robust solution has better system performance on the premise of ensuring the robustness.
3) Since both the mean and the variance of the workpiece processing time are obtained by sampling the historical data in actual production, the obtained sampled data may be biased in the case where the available historical data is insufficient. The invention fully considers the situation, and brings the uncertainty of the sampling data into the model, and the designed homotype parallel machine scheduling method considering the uncertainty of the moment information meets the actual requirement.
4) The model established by the invention can control the trust degree of the sampled data through parameter setting, so that a decision maker can set corresponding parameter values according to the current requirements to obtain the scheduling strategy most suitable for the current production situation.
5) The distributed set robust optimization model adopted by the invention can be combined with the advantages of a random scheduling model and a robust scheduling model based on an uncertainty set, is applied to the production scheduling problem of the parallel machines of the same type, better accords with the actual production condition compared with the existing robust modeling method, and can reduce the risk of decision by utilizing more information in the production environment under the condition of ensuring the system performance.
Detailed Description
The invention provides a homotype parallel machine production scheduling method considering moment information uncertainty, which is further described in detail below by combining specific embodiments.
The invention provides a homotype parallel machine production scheduling method considering moment information uncertainty, which comprises the following steps:
1) aiming at the scheduling problem of a same-type parallel machine, a distribution set robust optimization model DR-PMSP-MU considering moment information uncertainty is constructed, and an expression of an initial model DR-PMSP-MU1 is obtained;
the invention relates to a scheduling problem of a same-type parallel machine with uncertain processing time, and a distributed set robust optimization model (DR-PMSP-MU) considering moment information uncertainty is established aiming at the scheduling problem. In the same type of parallel machine scheduling problem, all workpieces can be processed on any one machine, but each machine can process only one workpiece at the same time, and one workpiece can be allocated to only one machine. Each workpiece has its own specific machining time and cannot be interrupted during machining. The purpose of solving the scheduling problem of the parallel machines of the same type is to find a scheduling scheme, determine the machine allocation conditions of all the workpieces and the processing sequence on each machine, and enable one of the system performance indexes such as total flow time, maximum delay time or delay workpiece number to be optimal.
In the DR-PMSP-MU model, the performance index of the system is selected as Total Flow Time (TFT), and it is assumed that all workpieces are released at the Time when machining starts, i.e., the release Time is 0. The processing time of the workpiece has random uncertainty, and the distribution of the random processing time is unknown but belongs to a distribution set determined by a support set, a sampling mean vector and a sampling covariance matrix. Under this setting, the goal of the DR-PMSP-MU model is to find an optimal robust scheduling scheme that has the smallest expected TFT if the workpiece processing time follows the worst distribution.
1-1) determining model decision variables;
decision variables of the DR-PMSP-MU model are feasible scheduling schemes, including machine selection of all workpieces and machining order on each machine. Let J workpieces and M machines in the model, the set of workpieces and machines is JS1,2, …, J and MSA feasible scheduling scheme is represented by a three-dimensional matrix X ∈ {0,1}J×M×J＝{xjml∈{0,1}|j∈JS,m∈MS,l∈LS＝JSRepresents; wherein if a workpiece j is assigned to the m-th machine and processed in order of the last but one, xjml1, otherwise xjml＝0。
For example, for a parallel machine scheduling problem with 5 workpieces, 2 machines, a feasible scheduling scheme SkCan be written as:
the machine 1: workpiece 1-workpiece 2-workpiece 4; the machine 2: workpiece 3-workpiece 5.
Scheduling scheme S according to the corresponding relation between the scheduling scheme and XkThree-dimensional matrix X ofkThe expression is shown in formula (1):
Figure GDA0002425944090000071
in the formula, Xk(m-1) and Xk(m ═ 2) respectively denotes a two-dimensional matrix corresponding to the machine 1 and a two-dimensional matrix corresponding to the machine 2; these two-dimensional matrices together form a scheduling scheme S representing a feasible solutionkThree-dimensional matrix X ofk。
1-2) random vector representation of processing time;
the DR-PMSP-MU model of the invention regards the processing time of all workpieces as a random vector p. The distribution obeyed by the method is unknown, but belongs to a distribution set determined by a support set, a sampling mean vector and a sampling covariance matrix, and the expression of the distribution set is shown as a formula (2):
Figure GDA0002425944090000072
wherein the content of the first and second substances,
Figure GDA0002425944090000073
indicating that the processing time for each workpiece is non-negative,
Figure GDA0002425944090000074
true mean vector, mu, representing the processing time of a workpiecesAnd sigmasAnd the sampling mean vector and the sampling covariance matrix respectively represent the processing time of the workpiece, and gamma is a constraint parameter and reflects the trust degree of the sampling information.
1-3) determining an objective function of a DR-PMSP-MU model;
the system performance index of the DR-PMSP-MU model is Total Flow Time (TFT), and when a scheduling scheme X and all workpiece processing Time vectors p are given, the TFT is calculated by the formula (3):
Figure GDA0002425944090000081
in the formula, pjThe machining time of the workpiece j is shown.
Because the processing time p of all the workpieces is a random vector, the total flow passing time TFT is a random variable, the method adopts expectation as the measurement of the random TFT, and obtains the objective function of the DR-PMSP-MU model as
Figure GDA0002425944090000082
This represents the expected value of the total flow over time, which is determined when all the workpiece processing time vectors p follow a certain distribution function F.
1-4) determining a constraint condition of a DR-PMSP-MU model;
the DR-PMSP-MU model contains 5 constraint conditions, of which 1 is a distribution set to which the constraint random machining time obeys, and the other 4 is feasibility of a constraint scheduling scheme, as shown in detail below:
1-4-1) random processing time constraint, wherein the expression is shown as a formula (2);
1-4-2) feasible scheduling scheme constraints;
each element in the feasible scheduling scheme X is 0 or 1, and the expression is shown in formula (4):
Figure GDA0002425944090000083
1-4-3) workpiece occupation position constraint;
each workpiece can only occupy one position on one machine, and the expression is shown in formula (5):
Figure GDA0002425944090000084
1-4-4) the position is occupied and restricted by the workpiece;
each position on each machine can be occupied by at most one workpiece, and the expression is shown in the formula (6):
Figure GDA0002425944090000085
1-4-5) ordering compact constraints;
the occupied position on each machine must be continuous and starting from 1, the expression is given by the equation (7):
Figure GDA0002425944090000086
the latter four kinds of constraints shown as formula (4) -formula (7) are all feasible for restricting the scheduling scheme, and are integrated together to form a feasible domain of the scheduling scheme
Figure GDA0002425944090000087
As shown in formula (8):
Figure GDA0002425944090000091
1-5) establishing an expression of a same-type parallel machine scheduling distribution set robust initial model DR-PMSP-MU1 considering moment information uncertainty, as shown in formula (9):
Figure GDA0002425944090000092
in the formula, max is represented in the distribution set
Figure GDA0002425944090000093
In finding the objective function
Figure GDA0002425944090000094
The maximum distribution function F, min represents the feasible region in the scheduling scheme
Figure GDA00024259440900000911
Middle search causes inner layer problems
Figure GDA0002425944090000095
The scheduling scheme X having the smallest optimal value.
2) Performing equivalent transformation on the DR-PMSP-MU1 model established in the step 1);
2-1) converting decision variables;
because the machine serial number allocated to the workpiece has no influence on the objective function value, the decision variables of the DR-PMSP-MU model can be equivalently converted into a two-dimensional matrix Y from a three-dimensional matrix X, and the conversion relation is shown as the formula (10):
Figure GDA0002425944090000096
with respect to the feasible domain of Y
Figure GDA0002425944090000097
Written out, the expression is shown as formula (11):
Figure GDA0002425944090000098
to further facilitate the representation and calculation of the model, the two-dimensional matrix Y is represented as a vector pi, and the expression is shown in formula (12):
Figure GDA0002425944090000099
pi denotes the reverse order of the order in which the workpieces are machined after ignoring the machine number, e.g. pijI denotes the last but one sequence of machining of the workpiece j on a certain machine. The expression of the feasible domain of pi is shown as formula (13):
Figure GDA00024259440900000910
the TFT can be expressed as the inner product of pi and p, and the expression is shown as the formula (14):
f(π,p)＝f(X,p)＝πTp； (14)
2-2) solving an inner layer problem;
after the decision variable X is converted into pi through the step 2-1), attention is paid to
Figure GDA0002425944090000101
The inner-layer maximization problem in the DR-PMSP-MU1 model shown in equation (10) can be simplified as shown in equation (15):
Figure GDA0002425944090000102
in the formula (I), the compound is shown in the specification,
Figure GDA0002425944090000103
and representing the constraint condition by s.t. which is a true mean vector of all the workpiece processing time.
Since the maximization problem shown in equation (15) is a convex optimization problem with a linear objective function, the optimal solution thereof must exist. We find the optimum of the problem by Karush-Kuhn-Tucke (KKT) requirementsIs solved as
Figure GDA0002425944090000104
The optimal value of the inner layer problem in the corresponding DR-PMSP-MU1 model is
Figure GDA0002425944090000105
2-3) equivalently converting the DR-PMSP-MU1 model into a solvable model DR-PMSP-MU 2;
replacing the inner layer maximization problem in the formula (9) with the optimal value obtained in the step 2-2), equivalently converting the DR-PMSP-MU1 model into a DR-PMSP-MU2 model, wherein the expression of the DR-PMSP-MU2 model is shown in the formula (16):
Figure GDA0002425944090000106
3) solving the DR-PMSP-MU model to obtain an optimal production scheduling scheme;
the DR-PMSP-MU2 model shown in formula (16) is an integer second-order cone programming model, can be solved through IBM CP L EX and other commercial solvers, and can also be solved through designing a more efficient iterative descent algorithm according to the nature of the problem.
The obtained optimal solution of the DR-PMSP-MU2 model is the optimal solution of the DR-PMSP-MU model, the optimal solution is an optimal vector pi value, and the optimal solution can be equivalently converted back to an optimal two-dimensional matrix Y through the corresponding relation shown in the formula (12), wherein the optimal solution comprises the processing sequence information corresponding to each workpiece; the optimal two-dimensional matrix Y value can correspond to a plurality of optimal production scheduling schemes, and a decision maker can select the optimal two-dimensional matrix Y value according to needs in actual production.
For example, for a parallel machine scheduling problem of the same type with 5 workpieces and 2 machines, if the optimal vector pi value obtained by the DR-PMSP-MU model is pi ═ (1, 3, 2, 2, 1)TThen, the corresponding Y value of the two-dimensional matrix is as shown in equation (17):
Figure GDA0002425944090000107
in the formula, the row number indicates the serial number of the workpiece, the column number indicates the processing order of the reciprocal, and for example, the element "1" on the third column of the second row indicates that the workpiece 2 is processed at the third position of the reciprocal on a certain machine.
Because all machines are the same and the processing effects are equivalent in parallel machines of the same type, when the optimal scheduling scheme is determined, the processing sequence of the workpieces only needs to be ensured to meet the requirements of the optimal solution pi and Y, and for the serial numbers of the machines, the serial numbers of the machines can be arbitrarily selected on the premise of ensuring that each machine only processes one workpiece at the same time.
The optimal two-dimensional matrix Y values shown in equation (17) may correspond to multiple optimal scheduling schemes, such as shown in equations (18) -20:
scheme 1. machine 1: workpiece 2-workpiece 3-workpiece 1; the machine 2: workpiece 4-workpiece 5; (18)
scheme 2. machine 1: workpiece 2-workpiece 4-workpiece 5; the machine 2: workpiece 3 — workpiece 1; (19)
scheme 3. machine 1: workpiece 3 — workpiece 1; the machine 2: workpiece 2-workpiece 4-workpiece 5; (20)
although the scheduling schemes are different in machine allocation, the obtained DR-PMSP-MU models have the same objective function value and belong to equivalent scheduling schemes. In actual production, a decision maker can select the materials according to needs.
In a specific embodiment of the present invention, a robust scheduling scheme obtained by a same-type parallel machine production scheduling method considering moment information uncertainty is compared with a deterministic scheduling scheme obtained only considering mean information, so as to demonstrate that the same-type parallel machine scheduling method considering moment information uncertainty designed by the present invention can greatly reduce interference caused by sampling errors, improve the robustness of a system, and the specific analysis is as follows:
the value of gamma reflects the trust degree of the sampled information and can be obtained by analyzing historical data. When historical data of a batch of workpiece processing time is obtained, the data is divided into two parts on average. One of which is used to calculate the sample mean vector musWith the sampling covariance matrix ∑s(ii) a Another for a given gamma value, such that the mean of this portion of data
Figure GDA0002425944090000111
Located on an ellipsoid
Figure GDA0002425944090000112
Among them.
In this example, the true mean of the random processing time is in the interval [10,60 ]]Is arbitrarily selected, and the variance is influenced by the mean value and is selected as
Figure GDA0002425944090000113
For each pair of mean and variance examples, 10,000 processing time examples were randomly generated as historical data and controlled by sampling rate for calculating μsAnd sigmasThe ratio of sampled data. Table 1 shows the robustness improvement and mean loss obtained for the robust scheduling scheme under different sampling rate values.
Table 1 TFT statistical comparison table for robust scheduling scheme and deterministic scheduling scheme under different sampling rate values
Figure GDA0002425944090000114
The results in table 1 show that:
1) under the same sampling rate, compared with a deterministic scheduling scheme, the robust scheduling scheme obtained by the method of the invention slightly increases the mean value of the TFT and obviously decreases the variance. Namely, the distribution set robust model considering moment information uncertainty is at the cost of small average performance loss, so that the dispersity of the TFT is reduced, the deviation of sampling data can be effectively resisted, and the robustness of a production system is improved.
2) The higher sampling rate causes the value of gamma to become smaller and the two scheduling schemes behave more closely. This shows that the advantage of robust scheduling decision is more significant in the case of more serious historical data loss, i.e. greater deviation of sampling information. The important significance of the homotype parallel machine scheduling method considering moment information uncertainty in the actual production environment with high uncertainty and less historical processing data is fully embodied.

Claims (1)

1. A homotype parallel machine production scheduling method considering moment information uncertainty is characterized by comprising the following steps:
1) aiming at the scheduling problem of a same-type parallel machine, a distribution set robust optimization model DR-PMSP-MU considering moment information uncertainty is constructed, and an expression of an initial model DR-PMSP-MU1 is obtained;
in a DR-PMSP-MU model, the performance index of the system is selected as total flow passing time TFT; assuming that all workpieces are released at the moment when the machining starts, namely the release time is 0, the machining time of the workpieces has random uncertainty, the distribution of the random machining time is unknown, but belongs to a distribution set determined by a support set, a sampling mean vector and a sampling covariance matrix; the DR-PMSP-MU model aims to find an optimal robust scheduling scheme, so that the scheduling scheme has the minimum expected TFT under the condition that the workpiece processing time obeys worst distribution;
1-1) determining model decision variables;
decision variables of the DR-PMSP-MU model are feasible scheduling schemes, and include machine selection of all workpieces and processing sequences on each machine; let J workpieces and M machines in the model, the set of workpieces and machines is JSJ, and M ═ 1,2SA feasible scheduling scheme is represented by a three-dimensional matrix X ∈ {0,1}J×M×J＝{xjml∈{0,1}|j∈JS,m∈MS,l∈LS＝JSRepresents; wherein if a workpiece j is assigned to the m-th machine and processed in order of the last but one, xjml1, otherwise xjml＝0；
1-2) random vector representation of processing time;
the processing time of all the workpieces is a random vector p, the distribution obeyed by the vector is unknown, but belongs to a distribution set determined by a support set, a sampling mean vector and a sampling covariance matrix, and the expression of the distribution set is shown as formula (2):
Figure FDA0002425944080000011
wherein the content of the first and second substances,
Figure FDA0002425944080000012
indicating that the processing time for each workpiece is non-negative,
Figure FDA0002425944080000013
true mean vector, μ, representing the processing time of all workpiecessAnd sigmasRespectively representing the sampling mean vector and the sampling covariance matrix of all the workpiece processing time, wherein gamma is a constraint parameter;
1-3) determining an objective function of a DR-PMSP-MU model;
given a schedule X and all workpiece processing time vectors p, the TFT is calculated by equation (3):
Figure FDA0002425944080000014
in the formula, pjRepresents the processing time of the workpiece j;
the total flow time TFT is a random variable, and the expectation is taken as the measurement of the random TFT to obtain the target function of the DR-PMSP-MU model as
Figure FDA0002425944080000029
Representing the total flow through time expected value obtained when all the workpiece processing time vectors p obey a certain distribution function F;
1-4) determining a constraint condition of a DR-PMSP-MU model;
1-4-1) random processing time constraint, wherein the expression is shown as a formula (2);
1-4-2) feasible scheduling scheme constraints;
each element in the feasible scheduling scheme X is 0 or 1, and the expression is shown in formula (4):
Figure FDA0002425944080000021
1-4-3) workpiece occupation position constraint;
each workpiece only occupies one position on one machine, and the expression is shown in formula (5):
Figure FDA0002425944080000022
1-4-4) the position is occupied and restricted by the workpiece;
each position on each machine is occupied by at most one workpiece, and the expression is shown in the formula (6):
Figure FDA0002425944080000023
1-4-5) ordering compact constraints;
the occupied positions on each machine are consecutive and starting from 1, the expression is given by the equation (7):
Figure FDA0002425944080000024
the latter four kinds of constraints shown as formula (4) -formula (7) are all feasible for restricting the scheduling scheme, and are integrated together to form a feasible domain of the scheduling scheme
Figure FDA00024259440800000210
As shown in formula (8):
Figure FDA0002425944080000025
1-5) establishing an expression of a same-type parallel machine scheduling distribution set robust initial model DR-PMSP-MU1 considering moment information uncertainty, as shown in formula (9):
Figure FDA0002425944080000026
in the formula, max is represented in the distribution set
Figure FDA0002425944080000027
In finding the objective function
Figure FDA0002425944080000028
The maximum distribution function F, min represents the feasible region in the scheduling scheme
Figure FDA0002425944080000031
Middle search causes inner layer problems
Figure FDA0002425944080000032
The scheduling scheme X with the minimum optimal value;
2) performing equivalent transformation on the DR-PMSP-MU1 model established in the step 1);
2-1) converting decision variables;
equivalently converting decision variables of the DR-PMSP-MU model from a three-dimensional matrix X into a two-dimensional matrix Y, wherein the conversion relation is shown as a formula (10):
Figure FDA0002425944080000033
the expression of the feasible domain of Y is shown as formula (11):
Figure FDA0002425944080000034
the two-dimensional matrix Y is expressed as a vector pi, and the expression is shown as the formula (12):
Figure FDA0002425944080000035
pi represents the reverse order of the processing sequence of the workpiece after the machine serial number is ignored;
the expression of the feasible domain of pi is shown as formula (13):
Figure FDA0002425944080000036
the TFT is expressed as the inner product of pi and p, and the expression is shown as the formula (14):
f(π,p)＝f(X,p)＝πTp； (14)
2-2) solving an inner layer problem;
after the decision variable X is converted into pi through the step 2-1), the decision variable X is determined according to
Figure FDA0002425944080000037
The inner layer problem in the DR-PMSP-MU1 model shown in equation (9) is simplified as shown in equation (15):
Figure FDA0002425944080000038
in the formula (I), the compound is shown in the specification,
Figure FDA0002425944080000039
representing the real mean vector of all the workpiece processing time by s.t. representing constraint conditions;
solving equation (15) to obtain the optimal solution of
Figure FDA00024259440800000310
The optimal value of the inner layer problem in the corresponding DR-PMSP-MU1 model is
Figure FDA00024259440800000311
2-3) equivalently converting the DR-PMSP-MU1 model into a solvable model DR-PMSP-MU 2;
replacing the inner layer problem in the formula (9) with the optimal value obtained in the step 2-2), equivalently converting the DR-PMSP-MU1 model into a DR-PMSP-MU2 model, wherein the expression of the DR-PMSP-MU2 model is shown in the formula (16):
Figure FDA0002425944080000041
3) solving the DR-PMSP-MU model to obtain an optimal production scheduling scheme;
solving the DR-PMSP-MU2 model shown as the formula (16), wherein the obtained optimal solution of the DR-PMSP-MU2 model is the optimal solution of the DR-PMSP-MU model; the optimal solution is an optimal vector pi value, and the vector pi value is equivalently converted into an optimal two-dimensional matrix Y value containing the processing sequence corresponding to each workpiece through the corresponding relation shown in the formula (12), so that an optimal production scheduling scheme is obtained.
CN201710717959.0A 2017-08-21 2017-08-21 Homotype parallel machine production scheduling method considering moment information uncertainty Active CN107609678B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201710717959.0A CN107609678B (en) 2017-08-21 2017-08-21 Homotype parallel machine production scheduling method considering moment information uncertainty

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201710717959.0A CN107609678B (en) 2017-08-21 2017-08-21 Homotype parallel machine production scheduling method considering moment information uncertainty

Publications (2)

Publication Number Publication Date
CN107609678A CN107609678A (en) 2018-01-19
CN107609678B true CN107609678B (en) 2020-07-31

Family

ID=61065509

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201710717959.0A Active CN107609678B (en) 2017-08-21 2017-08-21 Homotype parallel machine production scheduling method considering moment information uncertainty

Country Status (1)

Country Link
CN (1) CN107609678B (en)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115729198A (en) * 2022-12-02 2023-03-03 福州大学 Robust optimized group production method considering uncertainty of material-to-material time

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104809327A (en) * 2014-09-02 2015-07-29 长沙理工大学 New energy contained electric power dispatching moment uncertainty distribution robustness optimization method
CN106651089A (en) * 2016-09-19 2017-05-10 清华大学 Modeling and optimization method for production scheduling problem distribution set robust model

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104809327A (en) * 2014-09-02 2015-07-29 长沙理工大学 New energy contained electric power dispatching moment uncertainty distribution robustness optimization method
CN106651089A (en) * 2016-09-19 2017-05-10 清华大学 Modeling and optimization method for production scheduling problem distribution set robust model

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
Distributionally robust single machine scheduling with risk aversion;Zhiqi Chang 等;《European Journal of Operational Research》;20160616;第256卷;第261-274页 *
基于DE和EDA的智能算法求解复杂车间调度问题;李作成;《中国优秀硕士学位论文全文数据库工程科技II辑》;20150131(第1期);全文 *
遗传-分布估计算法求解化工生产中一类带多工序的异构并行机调度问题;李作成 等;《化工学报》;20140331;第65卷(第3期);第981-992页 *

Also Published As

Publication number Publication date
CN107609678A (en) 2018-01-19

Similar Documents

Publication Publication Date Title
CN106651089B (en) Modeling and optimization solving method for distribution set robust model of production scheduling problem
Aich et al. Modeling of EDM responses by support vector machine regression with parameters selected by particle swarm optimization
CN107544251B (en) Single machine scheduling method for minimizing total stall period based on distributed robust model
CN101551884B (en) A fast CVR electric load forecast method for large samples
Chauhan et al. Parameter optimization of multi-pass turning using chaotic PSO
CN111427750B (en) GPU power consumption estimation method, system and medium of computer platform
CN111069328A (en) Isothermal extrusion process parameter optimization method based on particle swarm optimization
CN104636871B (en) A kind of control method of the single phase multi-product batch processing based on data
CN113544599A (en) Method for executing a process and optimizing a control signal used in the process
CN107292391B (en) Flexible workshop task scheduling optimization method based on DE and L-BFGS-B hybrid algorithm
CN110045613B (en) Mixed integer optimal control numerical solution method based on quantum annealing
CN111210125B (en) Multi-target workpiece batch scheduling method and device based on historical information guidance
Gadakh et al. Selection of cutting parameters in side milling operation using graph theory and matrix approach
CN112308298A (en) Multi-scenario performance index prediction method and system for semiconductor production line
CN107609678B (en) Homotype parallel machine production scheduling method considering moment information uncertainty
CN108920900A (en) The unsupervised extreme learning machine Feature Extraction System and method of gene expression profile data
CN108808657B (en) Short-term prediction method for power load
Dizaji et al. PARTICLE SWARM OPTIMIZATION AND CHAOS THEORY BASED APPROACH FOR SOFTWARE COST ESTIMATION.
CN111259340A (en) Saturated load prediction method based on logistic regression
CN113485278B (en) Flexible job shop scheduling multi-target distribution estimation method for optimizing two production indexes
CN111091242B (en) Optimal nonparametric interval prediction method for power load
Gregor et al. SIMULATION META-MODEL OF ASSEMBLY LINE WITH CONWIP CONTROL.
CN113656707A (en) Financing product recommendation method, system, storage medium and equipment
CN112667394A (en) Computer resource utilization rate optimization method
CN112308340A (en) Power data processing method and device

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