US-12626031-B2 - Clustering method

US12626031B2US 12626031 B2US12626031 B2US 12626031B2US-12626031-B2

Abstract

A clustering method includes the steps of: creating a plant model in which connection relationships between a plurality of devices ( 31 ) forming a processing plant ( 1 ) and pipes ( 4 ) are represented by a graph showing a linking relationship between a node ( 51 ) and an edge ( 52 ); setting, regarding the plant model, a cluster division condition; and performing, by a computer, the cluster division of the plant model by searching for an edge between the clusters in which modularity is further increased while the cluster division condition is satisfied.

Inventors

Yoshinori Yamada
Kaoru SAWAYANAGI

Assignees

JGC CORPORATION

Dates

Publication Date: 20260512
Application Date: 20200518

Claims (7)

1 . A clustering method, comprising the steps of: creating a plant model in which connection relationships between a plurality of devices for constructing a processing plant configured to process a fluid and pipes connecting the plurality of devices to each other are represented by a graph showing a linking relationship between a node corresponding to each of the plurality of devices and an edge corresponding to each of the pipes; setting, regarding the plant model, a cluster division condition which is at least one of the number of clusters into which cluster division is to be performed and a range of a grain size that is an addition value of a weight set to each of the nodes comprised in each of the clusters; and performing, by a computer, the cluster division of the plant model by searching for an edge between the clusters in which modularity is further increased while the cluster division condition is satisfied.
2 . The clustering method according to claim 1 , wherein the step of creating the plant model comprises creating the plant model in which, when cost per unit length of the corresponding pipe is higher, the edge is weighted with a larger weight.
3 . The clustering method according to claim 1 , wherein the step of creating the plant model comprises creating the plant model in which, when an area of occupation of the corresponding device is larger or when evaluation values set in advance to the device are higher, the node is weighted with a larger weight.
4 . The clustering method according to claim 1 , wherein the step of performing the cluster division is performed based on a Girvan-Newman algorithm.
5 . The clustering method according to claim 4 , wherein the step of creating the plant model comprises creating the plant model in which, when cost per unit length of the pipe is higher, the number of unweighted edges that link two nodes corresponding to two devices, to which the pipe is connected, to each other is increased.
6 . The clustering method according to claim 4 , wherein the nodes corresponding to the plurality of devices comprised in the plant model are unweighted nodes, and wherein the step of creating the plant model comprises creating the plant model in which, when an area of occupation of the device is larger or when evaluation values set in advance to the device are higher, the number of connections of dummy nodes is increased, the dummy nodes each being a node which is linked to one node corresponding to the device via one edge and which is prevented from being linked to the nodes other than the one node.
7 . The clustering method according to claim 1 , wherein the step of performing the cluster division comprises the step of assigning a count value to each of edges in a shortest path between two nodes comprised in the plant model by counting the edges, and wherein, when the processing plant comprises a pipe branched at a branch point, the step of creating the plant model comprises creating the plant model in which a branch node prevented from being selected as the two nodes when the step of assigning the count value is performed is provided so as to correspond to the branch point, and the branch node and a node corresponding to the device connected via the branched pipe are linked to each other via an edge to represent the branched pipe.

Description

CROSS-REFERENCE TO RELATED APPLICATION This application is a 371 application of the International PCT application serial no. PCT/JP2020/019684, filed on May 18, 2020. The entirety of the above-mentioned patent application is hereby incorporated by reference herein and made a part of this specification. TECHNICAL FIELD The present invention relates to a technology of grouping devices for constructing a processing plant for a fluid. BACKGROUND ART Processing plants configured to process a fluid include, for example, natural gas plants configured to liquefy natural gas or separate and recover natural gas liquid, petroleum refining plants configured to distill or desulfurize crude oil or various intermediates, and chemical plants configured to produce petrochemical products, intermediate chemicals, or polymers. In this application, the “fluid” includes, in addition to gas and liquid, granular materials (powder, grains, pellets, and the like) having flowability. Those processing plants each have a configuration in which a large number of devices are arranged in a predetermined plot and those devices are connected to each other by pipes. Examples of the large number of devices include static devices such as columns, tanks, and heat exchangers, and dynamic devices such as pumps and compressors. At this time, the arrangement of the large number of devices included in the processing plant is determined with comprehensive consideration of, for example, restrictions on a landform and area of the plot, restrictions in terms of safety, and use amounts of materials for forming the devices and pipes (economic efficiency). In this case, in arrangement of the devices for constructing the processing plant, when a large number of devices are disorderly arranged, the pipes connecting the devices to each other become long, and the total use amount of pipe materials is increased. For this reason, a large number of devices may be divided into a plurality of groups, and arrangement positions may be determined so that the devices included in the same group are arranged in a collective region. As grouping of the devices, in many cases, there has hitherto been adopted, for example, a method for sorting out a series of processing to be performed in the processing plant into a plurality of sections from the viewpoint of process design and including the devices to be used in each sorted processing in one group (for example, module division corresponding to each processing block of Patent Literature 1). Meanwhile, when focus is given on the connection relationships between the devices and the pipes while the content of processing of a fluid is disregarded, whether or not the grouping by the related-art method reflects the degree of relationships between the devices is not necessarily clear. Accordingly, there is a demand for a grouping method capable of quantitatively grasping the results of determined grouping of devices with reference to the connection relationships between the devices and the pipes. CITATION LIST Patent Literature Patent Literature 1: WO 2018/220703 A1 SUMMARY OF INVENTION Technical Problem The present technology provides a technology of grouping devices for constructing a processing plant for a fluid while enabling quantitative evaluation of the results of grouping. Solution to Problem The present method includes the steps of: creating a plant model in which connection relationships between a plurality of devices for constructing a processing plant configured to process a fluid and pipes connecting the devices to each other are represented by a graph showing a linking relationship between a node corresponding to each of the devices and an edge corresponding to each of the pipes; setting, regarding the plant model, a cluster division condition which is at least one of the number of clusters into which cluster division is to be performed and a range of a grain size that is an addition value of a weight set to each of the nodes included in each of the clusters; and performing, by a computer, the cluster division of the plant model by searching for an edge between the clusters in which modularity is further increased while the cluster division condition is satisfied. The clustering method may include the following features. (a) The step of creating the plant model includes creating the plant model in which, when cost per unit length of the corresponding pipe is higher, the edge is weighted with a larger weight.(b) The step of creating the plant model includes creating the plant model in which, when an area of occupation of the corresponding device is larger or when evaluation values set in advance to the device are higher, the node is weighted with a larger weight.(c) The step of performing the cluster division is performed based on a Girvan-Newman algorithm.(d) In Item (c), the step of creating the plant model includes creating the plant model in which, when cost per unit length of the pipe is higher, the number of unwe