JPH04321171A - Process design preparing device - Google Patents

Abstract

PURPOSE:To widen the searching range and to increase the possibility to acquire an optimum major solution by providing a feedback means which feeds a possible solution obtained by a synthesizing means back to the executing process of a searching means to improve the searching efficiency. CONSTITUTION:A searching processor 3 searchs the solution of a merging optimization problem with use of a program stored in a searching program store part 1. Then a limit satisfying solution and other data are stored in a data store device 2. Meanwhile a synthesization processor 6 synthesizes a solution with use of a hereditary evolution program stored in a synthesizing program store part 7. This synthesized solution and other data are stored in a data store device 5. A feedback processor 4 feeds the information on the possible solution obtained in the executing process of the processor 3 and the information on the possible solution obtained by the processor 6 back to the processor 3 in a state where a searching point is controlled and the searching efficiency is improved.

Description

[Detailed description of the invention]

0001

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an apparatus for automatically and efficiently creating the best plan while satisfying the constraints that must be met in the preparation of various plant construction process plans.

0002

[Prior Art] A process plan creation support device for the purpose of equalizing human resources in plant installation process planning.
The application has been filed in ``Japanese Patent Application Laid-open No. 19379/1983.'' It uses an iterative method to solve the combinatorial optimization problem to be solved within it. In addition, methods using linear programming and branch-and-bound methods for solving general combinatorial optimization problems are described in Ellis El Johnson et al.'s "Solution of 0-1 Integer Programming Problems for Large-Scale Programming Models," Operation Research. , Volume 33, Issue 4, July-August 1985 (
ELLISL. JONSON et al. ,Solv
sing 0-1 Integer Program
ing Problems Arising from
Large Scale Planning Mode
les, Operations Research, Vo
l. 33. No.4. July-August 1985
), and the application of genetic evolutionary algorithms to combinatorial optimization problems is also discussed in ``Otaku, Sugai, Hirata, Block placement method using an improved annealing method incorporating genetic elements, Journal of the Society of Electronics and Communications Engineers A. , J73-A
-1, 87/97 (1990).

0003

Problem to be Solved by the Invention Since the above-mentioned process planning support system uses a point-by-point method based on an iterative method, it takes a lot of time to create a large-scale plan. In addition, conventional methods for solving combinatorial optimization problems are realized using sequential exploratory search methods for solutions such as branch-and-bound methods, or methods that incorporate stochastic elements such as genetic algorithms and Monte Carlo methods. Since the search range is limited due to time constraints, there is no guarantee that the generally obtained optimal solution is the global optimal solution.

An object of the present invention is to widen the search range by increasing solution search efficiency, increase the possibility of obtaining a globally optimal solution, and apply this method to process planning.

[0005]

[Means for Solving the Problems] In order to achieve the above object, the present invention provides a means for searching for an optimal solution for solving an optimization problem, and a means for searching for another possible solution from the possible solutions obtained in the process of executing the search means. It is equipped with a synthesis means for synthesizing, a feedback means for feeding back the obtained possible solutions to the execution process of the search means to improve the efficiency of the search, and is designed to solve a combinatorial optimization problem when creating a process plan.

[0006]

[Operation] The search means attempts to search for a solution systematically, one by one, or analytically. The synthesis means tries to obtain a possible solution close to the global optimal solution at an early stage, stochastically or approximately. The feedback means feeds back information on possible solutions obtained in the execution process of the search means and information on solutions obtained by the synthesis means to the search means in a manner that controls search points and increases search efficiency.

[0007]

[Example] Hereinafter, an example of the present invention for the case where a process plan for the purpose of leveling human resources is created by a search method using a constraint satisfaction problem solving method and a synthesis method using a genetic evolutionary algorithm. will be explained in detail. FIG. 1 shows an example of a device configuration for realizing this embodiment. Search processing device 3
uses the program stored in the search program storage unit 1 to search for a numerically analytical solution shown in this embodiment, and stores the constraint satisfying solution and other data in the data storage device 2. In parallel, the synthesis processing device 6 synthesizes solutions using the genetic evolution program stored in the synthesis program storage section 7, and stores the synthesis solution and other data in the data storage device 5. Between the search processing device 3 and the synthesis processing device 7,
Information regarding the search solution and information regarding the synthetic solution are exchanged via the feedback processing device 4, and the search means is controlled.

FIG. 2 shows a processing procedure for creating a process plan for the purpose of equalizing human resources using this method. FIG. 3 shows an integrated process plan diagram used to express a process plan and its human resources pile, and Table 1 shows the types of process plans and various constraints.

[0009]

[Table 1]

[0010] Leveling human resources is a process that suppresses the peak of required human resources under various control conditions as shown in Table 1, and flattens out the pile of human resources as much as possible as shown in Figure 3. It is about creating a plan.

In FIG. 2, first, in process 8, conditional expressions are created using appropriate variables to express various constraints on the work period, work order, and schedule of each process. For example, Figure 4
As shown in Figure 2, considering real variables Xi0 to Xi (M+0) that take values between 0 and 1 corresponding to each day of M days, the i-th process on the required number of days Ni is set on the day when the work is performed. j is expressed by setting the variable Xij to 1 and the other Xij to 0, and also a real variable Yi that takes a value between 0 and 1.
Considering 1~Yi (M+1), Zi1~Zi (M+1),
The start date of the i-th step is day b, and the end date is day e,
Other Yij, with Yib=1, Zi(e+1)=1,
When expressed by setting Zij to 0, constraints on the work period, work order between processes, and constraints on the height of the peak of human resources pile up can be expressed as follows for each process i using these variables: Expressed by the formula.

First, the expression of the work period is expressed as (Equation 1)
~(Math. 5) shows. (Math. 1) represents that only N i of Xij to XiM are 1. (Math. 2) is Yi1~
It represents that only one of Yi(M+1) is 1. (3) represents that only one of Zi1 to Zi(M+1) is 1. (Math. 4) together with (Math. 2) and (Math. 3) represent that Yij and Zij are 0 or 1. (
In Equation 5), Xij becomes 1 continuously for j (01
1011 etc. does not appear).

[0013] Xi0=0, Xi(M+0)=0

001
4]

[Math 1]

[0015]

[Math 2]

[0016]

[Math 3]

[0017]

[Math 4]

[0018]

[Math 5]

Next, expressions of work order constraints and schedule constraints are shown in equations (6) to (10). (Number 6)
represents that the start date of the i-th step is before day p. (Equation 7) represents that the start date of the i-th step is after the q day. (Equation 8) represents that the end date of the i-th step is before r days. (Equation 9) represents that the end date of the i-th step is after the s day. (Equation 10) indicates that step b starts after step a ends. (Equation 11) means that the end of process b is process a.
Indicates that it is after the end of.

[0020]

[Math 6]

[0021]

[Math 7]

[0022]

[Math. 8]

[0023]

[Math. 9]

[0024]

[Math. 10]

[0025]

[Math. 11]

Finally, (Equation 12) is a constraint on the height of the peak of the pile of human resources, and represents that the total number of required personnel on each day is less than or equal to P (the peak of the pile is less than or equal to P).

[0027]

[Math. 12]

Xij that satisfies all of the above conditional expressions
, Yij, and Zij, the solution directly represents a process plan in which the peak of the pile of human resources is P people or less. Therefore, in FIG. 2, first, in process 9, P is set to an appropriately large value, and in process 10, a constraint satisfaction problem is created by combining the constraint on the height of the piled peak (formula 12) and other various constraint condition expressions, In step 11, it is determined whether a satisfying solution exists or not by actually solving a satisfying problem or by using a necessary condition for satisfying the problem. If a solution does not exist, the process moves to step 15, displays that the process plan cannot be created, and ends the process. If a solution exists in process 11,
The process moves to the solution solving process 12, the genetic evolution operation process 13, and the P update process 14. In step 12, a satisfying solution is actually found using an optimization method such as the steepest descent method or conjugate gradient method, or other analytical solution method.

Process 13 is a process in which the solutions previously obtained in process 12 or the solution synthesized by the operation in process 13 are used as parents and another solution is synthesized as a descendant by genetic evolution operation. Processing 13 is performed by a parallel processing device in parallel with the processing of other parts, and outputs the maximum peak value P' of the best solution with the smallest maximum value of the piled peaks among the composite solutions at each time point, and processes 14. pass it on to

In process 14, the maximum value P' of the piled-up peak of the solutions obtained in process 12 and the minimum peak values P'' and P-1 among the composite solutions output from process 13 at the time when process 14 is performed are calculated. The smallest value among them is updated as the next value of P, and the operation returns to process 10 to repeat the operation of solving the constraint satisfaction problem.P is decreased through the loop of processes 10, 11, 12, and 14, and a certain If it becomes impossible to satisfy the constraint at P or less, in step 15, the solution satisfying the constraint at P-1 is displayed and output as a process plan that minimizes the stacked peak, and the process ends.

In the example of the device configuration shown in FIG. 1, for example, processes 10, 11, and 12 in FIG. It may be attached to the device or provided as a separate input/output device to perform the respective operations.

[0032] To operate the process 13 in parallel is to operate the process 13 in parallel.
3, at a certain point a solution with a peak P'' smaller than P-1, P' is synthesized, and in process 14 it is possible to significantly reduce P to P'', which improves the overall processing efficiency. It can be expected that the effect will be improved.

[0033] The device for performing the genetic evolution operation in process 13 is as follows:
A genetic operator that can synthesize a variable sequence representing another process plan from variable sequences representing process plans using variables Xij, Yij, and Zij is appropriately determined, and a solution is thereby synthesized. For example, in the simplest case, Figure 5
As shown in Figure 2, each process is divided into groups that are connected to each other by constraints on the work order, and a cross-breeding rule is used to exchange the variable sequences between mutually independent groups for the variable sequences representing the two process plans. By providing this, it is possible to realize solution synthesis.

[0034]

Effects of the Invention: In addition to the search means, there is a synthesis means that tries to quickly obtain a possible solution close to the global optimal solution, stochastically or approximately, and the search efficiency of the search means is improved by mutual feedback means. By acting in the direction of increasing the search range, the search range can be expanded earlier, or the optimal solution can be approached sooner, thereby achieving the effect of improving the efficiency of the overall processing.

[Brief explanation of the drawing]

FIG. 1 is a block diagram of a device configuration example of a process planning device for the purpose of leveling human resources, which is an embodiment of the present invention.

FIG. 2 is a processing flowchart of FIG. 1;

FIG. 3 is an explanatory diagram showing an integrated process plan diagram used to represent a process plan and a pile of human resources.

FIG. 4 is an explanatory diagram showing how to express the work period of each process using variables.

FIG. 5 is an explanatory diagram showing an example of genetic operators required when the solution synthesis section of the present invention is implemented using a genetic evolutionary algorithm.

[Explanation of symbols]

1... Search program storage unit, 2, 5... Data storage device,
3... Search processing device, 4... Feedback processing device, 6...
Synthesis processing device, 7...Synthesis program storage device.

Claims (8)

[Claims] Claims 1: A means for searching for an optimal solution for solving an optimization problem, a composition means for synthesizing another possible solution from possible solutions obtained in the process of executing the search means, and a process for using the obtained possible solution in the process of executing the search means. A process planning device characterized by comprising a feedback means for providing feedback to improve search efficiency. 2. A process planning device according to claim 1, which uses means for synthesizing another possible solution from a plurality of possible solutions obtained from the searching means or the synthesizing means according to a synthesis rule. 3. A process planning device according to claim 1, which uses means for generating another possible solution by transforming a possible solution obtained from the searching means or the synthesizing means according to a transformation rule. 4. A process planning apparatus according to claim 1, which uses means for performing optimization by giving an objective function as a constraint condition and changing the constraint value. 5. A process planning device according to claim 1, which uses a counting algorithm for pruning combinations. 6. The process planning device according to claim 1, wherein the feedback means compares a possible solution by the searching means with a possible solution by the synthesizing means to control the search points in the searching means. 7. The process planning apparatus according to claim 1, wherein the process planning device uses means for performing the search for a solution by the searching means and the synthesis of solutions by the synthesizing means in parallel processing. 8. The step according to claim 1, comprising a search means for performing optimization by giving an objective function as a constraint condition and changing a constraint value, and a synthesis means for synthesizing a solution using a genetic evolutionary algorithm. Planning device.