JP7462473B2 - Piping, wiring or duct route creation device, piping, wiring or duct route creation method - Google Patents

Description

The present invention relates to an automatic design technology for piping, wiring, or duct routes, and in particular to a technology that is effective when applied to the design of complex routes in power plants, railway-related facilities, etc.

At thermal power plants, nuclear power plants, petrochemical plants, railway-related facilities, and other facilities, huge amounts of equipment, piping, air conditioning ducts, cable trays, and so on are arranged inside the buildings. Furthermore, in these facilities, it is often necessary to install multiple safety systems to ensure safety, and the design of the piping, wiring, and duct routes for this requires the application of a wide variety of design rules and demands highly reliable designs on a par with those of experienced engineers.

As background technology in this technical field, for example, there is technology such as that described in Patent Document 1. Patent Document 1 discloses "an automatic piping routing method using a genetic algorithm that takes into account the ease of piping installation."

Patent Document 2 also discloses a "piping route creation device, creation method, and program that includes a route search unit for searching for a piping route and a route alignment unit for aligning multiple piping routes."

JP 2002-288250 A JP 2019-106141 A

In the above-mentioned Patent Document 1, a genetic algorithm is applied to the creation of piping routes, and the workability of the piping routes is scored. This method generates multiple routes with good piping workability, and a quasi-optimal piping route can be selected from the scores of the multiple candidates.

In addition, the above-mentioned Patent Document 2 has multiple means, including a route search unit that searches for a piping route and a route alignment unit that aligns multiple piping routes. The configuration of these means improves the efficiency of the piping route design process.

However, depending on the type of plant, a large number of design rules (more than several hundred) may be prescribed, and neither Patent Document 1 nor Patent Document 2 describes how to handle these large number and variety of rules.

In the automated design of piping for some plants, the difficulty of applying a large number of design rules has been an obstacle to realizing automated design. In other words, when automatically designing piping that must comply with a large number of design rules, it was necessary to more effectively incorporate the design rules into the route generation and selection process.

However, using algorithms such as genetic algorithms to apply a large number of rules to piping routes requires extremely cumbersome programming work, making it difficult to implement. Even if a large number of rules could be incorporated into a program, route search could take a significant amount of time, and there was a possibility that route search would not even be possible.

In addition, when imitating piping created based on the experience of an expert, it is difficult to systematize all of these piping shapes into specific design rules. Even in these cases, it is not realistic to use algorithms and apply them all to automatic design using a program.

The object of the present invention is to provide a piping, wiring, or duct route creation device and a piping, wiring, or duct route creation method that can create routes with the same reliability as an expert while applying a large number and wide variety of design rules in the automatic design of piping, wiring, or duct routes.

In order to solve the above problems, the present invention provides a cell generation device that generates cells, a route search device that generates routes using a route search algorithm, a route selection device that selects a route using a route selection algorithm, a real coordinate conversion device that converts routes into real coordinates, a dataset that includes route information, learning means that learns the dataset, and learning operation means that utilizes the results of learning by the learning means, and is characterized in that the results of learning by the learning means are utilized for at least one of the route generation by the route search device and the route selection by the route selection device.

The present invention also provides a method for creating a route for piping, wiring, or ducts, comprising the steps of (a) dividing a space into a plurality of cells and distinguishing cells in which a route can be laid from other cells, (b) generating a route using a route search algorithm, and (c) selecting a route using a route selection algorithm, characterized in that the results of learning a data set containing route information by a learning means are used in at least one of the steps (b) and (c).

According to the present invention, it is possible to realize a piping, wiring, or duct route creation device and a piping, wiring, or duct route creation method that can create routes with the same reliability as an expert while applying a large number and wide variety of design rules in the automatic design of piping, wiring, or duct routes.

Problems, configurations, and advantages other than those described above will become clear from the description of the embodiments below.

1 is a diagram showing a schematic configuration and operation (action) of a piping route creation device according to a first embodiment of the present invention; 1 is a diagram showing a piping route creation system according to an embodiment of the present invention. 6 is a diagram showing a schematic configuration and operation (action) of a piping route creation device according to a second embodiment of the present invention; FIG. 11 is a diagram showing a schematic configuration and operation (action) of a piping route creation device according to a third embodiment of the present invention. FIG.

Below, an embodiment of the present invention will be described with reference to the drawings. Note that the same components in each drawing will be given the same reference numerals, and detailed descriptions of overlapping parts will be omitted.

The configuration and operation (function) of a piping route creation device according to a first embodiment of the present invention will be described with reference to Figures 1 and 2. Note that the present invention can be applied to route creation for piping, wiring, or ducts, but in the following description, piping route creation will be described as a representative embodiment. In other words, the contents described below for piping can also be applied to route creation for wiring and ducts.

Figure 1 is a diagram showing the schematic configuration and operation (function) of the piping route creation device 1 of this embodiment. Figure 2 is a diagram showing the piping route creation system of this embodiment, showing the relationship between the piping route creation device 1 of Figure 1 and other devices that make up the system.

As shown in FIG. 1, the piping route creation device 1 of this embodiment is composed of a cell generation device 2 which is a means for generating cells (lattices), a route search device 3 which is a means for searching for a route using a route search algorithm, a route selection device 4 which is a means for selecting a route using a route selection algorithm, a real coordinate conversion device 5 which is a means for converting a piping route into real coordinates, a piping data set 6A, a data learning device 7 which is a means for learning the piping data set 6A, a learning operation device (means) 8 which searches for a route from the results learned by the data learning device 7, a piping data set 6B which has contents different from the piping data set 6A, a data learning device 9 which is a means for learning the piping data set 6B, and a learning operation device (means) 10 which operates the learning to select a suboptimal route.

The cell generating device (means) 2, which generates cells, divides the space along which the piping route will be routed into cells based on, for example, 3D CAD information, and distinguishes between areas of obstacles such as equipment, walls, floors, and pillars, and spaces in which the piping route can be laid, using cells. It also searches for the piping route in the center of the generated cell.

The route search device (means) 3 that searches for a route programs a route search algorithm (hereafter abbreviated as algorithm) such as the maze method or line segment search method based on design information such as the start and end points of the route, the size of the pipes, and the location of obstacles, to generate a pipe route. Even with these route search algorithms, simple design rules such as minimum length can be taken into account.

In the present invention, in order to apply a large number of complex design rules, a piping route is generated by using, in addition to the algorithm, a data learning device (means) 7 that learns the piping data set 6A, and a learning operation device (means) 8 that operates the learning in the data learning device (means) 7 for route search.

When searching for a route, not all rules are learned, but only the design rules appropriate for the route search are applied. In addition, based on the design rules used for the route search, a large number of pipes (several dozens or more) that are candidates for the optimal solution are generated. Deep learning is used in the data learning device (means) 7 and the learning operation device (means) 8 that operates the learning in the route search.

The piping data set 6A to be learned also includes data on the shape, characteristics, function, performance, coordinates, etc. of the piping. For example, this is 3D CAD information on piping that complies with design rules, and piping route coordinate information. Data on piping designed by experts in past plants is also used.

The piping data set 6A to be learned can also include information on the piping's thermal stress, pressure loss, vibration, earthquake resistance, stress, wall thinning, two-phase flow issues, water hammer, workability, inspectability, equipment operability, accessibility, quantity, cost, quality, suitability, etc. For example, the input values of the learning data can be data on the shape, characteristics, function, performance, and coordinates of the piping, and the output values can be information (design rules) including the piping's thermal stress, pressure loss, vibration, earthquake resistance, stress, wall thinning, two-phase flow issues, water hammer, workability, inspectability, equipment operability, accessibility, quantity, cost, quality, suitability, etc. By including this information, a more optimal piping route can be generated.

The learning effect is also effective when you want to generate piping with a similar shape to the piping in the dataset. The piping dataset 6A can be changed for each piping to be created. The piping dataset 6A may be processed into weights and biases for deep learning.

The present invention is provided with a route selection device (means) 4 that selects the optimal piping from multiple generated piping. A genetic algorithm or the like (hereinafter abbreviated as algorithm) is used as the selection algorithm.

In addition, in the present invention, in addition to the selection algorithm, a data learning device (means) 9 that learns the piping data set 6B and a learning operation device (means) 10 that operates the contents learned by the data learning device (means) 9 for route selection are used in combination to select the optimal piping.

Here, design rules not used in route search are also learned, and these piping data sets 6B are applied by the learning operation device (means) 10. Again, the data learning device (means) 9 and the learning operation device (means) 10 that selectively operates the learning are created by deep learning. Furthermore, the piping data set 6B to be learned is the same data as the piping data set 6A. The learning effect is also effective when it is desired to select piping with a similar shape to piping in the data set.

Finally, a real coordinate conversion device (means) 5 is used to convert the optimal piping route generated on the cell into coordinate values within the 3D CAD.

There are many design rules for generating piping, and the content to be applied is also diverse. In some cases, hundreds of design rules must be applied to a piping route. In addition, the type of fluid flowing through the piping, the environment in which it is used, and the reliability required vary from piping to piping, so the rule set to be applied to each piping is significantly different. In order to incorporate these different design rules into various algorithms and apply them, a huge amount of programming work would be required for each piping, which is not realistic. Therefore, as in the present invention, by separately providing a means for learning the piping shape created taking into account many design rules and a means for operating the learning, automatic design that takes into account many design rules can be realized.

In addition, in order to apply this to a wide variety of piping, it is possible to change the piping data set to be learned for each pipe. On the other hand, if there is insufficient learning data, the means for operating the learning alone may not be able to search for an appropriate piping route. For this reason, in this invention, the route search and route selection algorithms created by the program are stored, and the learning operation means is installed in combination. Note that in the process of generating a single pipe (route), the means for operating the learning and the means for generating a route by an algorithm can be used alternately as appropriate.

For example, if a solution (route) cannot be found using the means for operating the learning, a route search algorithm is operated. In other words, the means for operating the learning has functions separate from the route search algorithm and route selection algorithm, and the learning operation means and these algorithms are used in combination (hybrid) according to the conditions. Also, the means for operating the learning is capable of faster processing than various algorithms. In other words, by having learning and a means for operating the learning, the calculation time for route search and route selection is significantly reduced.

Furthermore, if too many design rules are imported at once, the design rules may become too strict, making it impossible to search for and select piping routes. For this reason, the learning and operation of design rules are each installed in multiple stages. In particular, it is desirable to install multiple learning means and learning operation means before and after the piping generation process.

Different data sets are provided for each means as the learning data set. There are design rules that must be followed (hereafter abbreviated as "must rules") and rules that are preferable to follow (hereafter abbreviated as "better rules"). Must rules are mainly learned for route search, and better rules are mainly learned for route selection. As a result, during route search, many pipes that comply with the rules that must be followed are generated, and during route selection, the rules that are preferable to follow are also taken into consideration, and the optimal pipe is selected from among them. For this reason, in the present invention, it is preferable to learn different data sets before and after route search, and multiple learning means and learning operation means are provided.

Note that if the learning means are not separated for route search and route selection, the following problems arise. If all learning data is applied from the time of route search, the routing conditions will be strict and a lot of time will be spent on route search. On the other hand, if all learning data is provided only at the time of route selection, very random piping will be generated during route search, making it difficult to converge to the optimal piping. Therefore, by appropriately separating learning data and performing and applying learning in multiple stages, as in the present invention, it is possible to improve the feasibility of routing that takes into account multiple design rules and the convergence to the optimal solution.

Next, a piping route creation system to which the piping route creation device 1 described above is applied will be described with reference to FIG. 2. FIG. 2 shows the relationship between the piping route creation device 1 of the present invention and an input/output device. The piping route creation device 1 is installed in a computer, and uses as input CAD model data of a 3D CAD system 11 also installed in the computer. Note that the 3D CAD system 11 does not necessarily have to be installed in the same computer as the piping route creation device 1.

The piping data set 6 is also stored on a hard disk in the computer. The piping data set 6 may be processed into weight and bias data for deep learning.

From these inputs, the piping route creation device 1 of the present invention generates a piping route. The generated piping route is displayed on a display device 12 that displays the route. The display device 12 may also serve as the 3D CAD system 11.

As described above, the route creation device for piping, wiring, or ducts in this embodiment includes a cell generation device 2 that generates cells, a route search device 3 that generates routes using a route search algorithm, a route selection device 4 that selects a route using a route selection algorithm, a real coordinate conversion device 5 that converts routes into real coordinates, and a dataset 6 (6A, 6B) that contains route information. When the route search device 3 generates a route, it generates multiple routes based on must rules that should be observed, and when the route selection device 4 selects a route, it selects the optimal route from the multiple routes generated by the route search device 3 based on better rules that should be observed.

As a result, in conventional methods, it becomes difficult to find an optimal route by importing a large number of diverse design rules (must rules and better rules) at once, whereas in this embodiment, must rules and better rules are imported in a hierarchical manner, making it possible to apply a large number of diverse design rules while creating a route with the same reliability as an expert.

A piping route creation device according to a second embodiment of the present invention will be described with reference to FIG. 3. FIG. 3 is a diagram showing the schematic configuration and operation (function) of the piping route creation device 1 of this embodiment.

As shown in Fig. 3, in route search to create a route, the piping route creation device 1 of this embodiment generates a piping route by using a data learning device (means) 7 that learns a piping data set 6A and a learning operation device (means) 8 that operates the learning in the data learning device (means) 7 for route search, in addition to the algorithm of the route search device (means) 3, as in the first embodiment (Fig. 1). On the other hand, in route selection, a piping is selected by a selection algorithm such as a genetic algorithm of the conventional route selection device (means) 4.

As in this embodiment, the effect can be obtained even if the data learning device (means) and the learning operation device (means) are used only when searching for a piping route. In this case, a large number of design rules can be applied in searching for a piping route.

A piping route creation device according to a third embodiment of the present invention will be described with reference to FIG. 4. FIG. 4 is a diagram showing the schematic configuration and operation (function) of the piping route creation device 1 of this embodiment.

As shown in FIG. 4, in route search for creating a route, the piping route creation device 1 of this embodiment generates a piping route using a route search algorithm by a conventional route search device (means) 3. On the other hand, in route selection, the optimal piping route is selected by using a data learning device (means) 9 that learns the piping data set 6B and a learning operation device (means) 10 that operates the contents learned by the data learning device (means) 9 for route selection in addition to the selection algorithm of the route selection device (means) 4.

As in this embodiment, the effect can be obtained even if the data learning device (means) and the learning operation device (means) are used only when selecting a piping route. In this case, a large number of design rules can be applied in searching for a piping route. In this case, a large number of design rules can be applied in selecting a piping route.

As described above, the piping, wiring, or duct route creation device of the present invention is equipped with a learning means (data learning devices 7, 9) that learns the data set 6 (6A, 6B), and a learning operation means (learning operation devices 8, 10) that uses the results learned by the learning means (data learning devices 7, 9), and uses the results learned by the learning means (data learning devices 7, 9) for at least one of route generation and route selection.

Furthermore, the data set 6 (6A, 6B) has a first data set 6A and a second data set 6B different from the first data set 6A, the learning means (data learning devices 7, 9) has a first learning means (data learning device 7) and a second learning means (data learning device 9) different from the first learning means (data learning device 7), and the learning operation means (learning operation devices 8, 10) has a first learning operation means (learning operation device 8) and a second learning operation means (learning operation device 8) different from the first learning operation means (learning operation device 8). The system has an operation means (learning operation device 10), and when the route search device 3 generates a route, the first learning means (data learning device 7) learns a first data set 6A, and the learned result is used to generate the route by the first learning operation means (learning operation device 8), and when the route selection device 4 selects a route, the second learning means (data learning device 9) learns a second data set 6B, and the learned result is used to select the route by the second learning operation means (learning operation device 10).

The present invention is not limited to the above-described embodiments, but includes various modified examples. For example, the above-described embodiments have been described in detail to aid in understanding the present invention, and are not necessarily limited to those having all of the configurations described. It is also possible to replace part of the configuration of one embodiment with the configuration of another embodiment, and it is also possible to add the configuration of another embodiment to the configuration of one embodiment. It is also possible to add, delete, or replace part of the configuration of each embodiment with other configurations.

1...piping route creation device, 2...cell generation device (means), 3...route search device (means), 4...route selection device (means), 5...actual coordinate conversion device (means), 6, 6A, 6B...piping data set, 7...data learning device (means), 8...learning operation device (means), 9...data learning device (means), 10...learning operation device (means), 11...3D CAD system, 12...display device.

Claims (14)

A cell generating device that generates cells;
A route search device that generates a route by a route search algorithm;
a route selection device for selecting a route according to a route selection algorithm;
a real coordinate conversion device for converting the route into real coordinates;
A dataset containing information about the route;
learning means for learning said data set;
A learning operation means for utilizing the results learned by the learning means,
A route creation device for piping, wiring or ducts, characterized in that the results learned by said learning means are utilized for at least one of route generation by said route search device and route selection by said route selection device. The piping, wiring, or duct route creation device according to claim 1,
When generating a route by the route search device, the route search algorithm and the learning operation means are used in combination to generate a route;
A piping, wiring or duct route creation device, characterized in that when a route is selected by said route selection device, an optimum route is selected by using said route selection algorithm in combination with said learning operation means. The piping, wiring, or duct route creation device according to claim 1,
the data sets include a first data set and a second data set different from the first data set;
the learning means includes a first learning means and a second learning means different from the first learning means,
the learning management means includes a first learning management means and a second learning management means different from the first learning management means,
When generating a route by the route search device, the first learning means learns the first data set, and the first learning operation means uses the learned result to generate the route;
A piping, wiring or duct route creation device, characterized in that when a route is selected by the route selection device, the second learning means learns the second data set, and the learned result is used to select a route by the second learning operation means. The piping, wiring, or duct route creation device according to claim 1,
A piping, wiring or duct route creation device, characterized in that deep learning is used for the learning means and the learning operation means. The piping, wiring, or duct route creation device according to claim 1,
The data set includes any one of information on the shape, characteristics, function, performance, and coordinates of the route. The piping, wiring, or duct route creation device according to claim 1,
The device for creating routes for piping, wiring, or ducts, characterized in that the dataset includes any of information on piping thermal stress, pressure loss, vibration, earthquake resistance, stress, wall thinning, two-phase flow issues, water hammer, workability, inspectability, accessibility, equipment operability, quantity, and cost. The piping, wiring, or duct route creation device according to claim 1,
A piping, wiring or duct route creation device, characterized in that when a route is created by the route search device, the route search algorithm and the learning operation means are alternately used to create a route. (a) dividing a space into a plurality of cells and distinguishing a cell in which a route can be laid from other cells;
(b) generating a route using a route search algorithm;
(c) selecting a route according to a route selection algorithm;
A method for creating a route for a pipe, wiring, or duct, comprising:
A method for creating a route for piping, wiring, or ducts, comprising using a result of learning a data set including route information by a learning means in at least one of the steps (b) and (c). The method for creating a route for piping, wiring, or duct according to claim 8,
In the step (b), a route is generated by using the route search algorithm and the result learned by the learning means in combination;
A method for creating a route for piping, wiring, or ducts, wherein in the step (c), an optimum route is selected by using the route selection algorithm in combination with the result learned by the learning means. The method for creating a route for piping, wiring, or duct according to claim 8,
A method for creating a route of piping, wiring, or duct, characterized in that different learning means and data sets are used in the steps (b) and (c), respectively. The method for creating a route for piping, wiring, or duct according to claim 8,
A method for creating routes for piping, wiring, or ducts, characterized by using deep learning. The method for creating a route for piping, wiring, or duct according to claim 8,
A method for creating a route for piping, wiring, or duct, characterized in that the data set includes information on any one of the shape, characteristics, function, performance, and coordinates of the route. The method for creating a route for piping, wiring, or duct according to claim 8,
The method for creating a route for piping, wiring, or duct, characterized in that the dataset includes any of information on piping thermal stress, pressure loss, vibration, earthquake resistance, stress, wall thinning, two-phase flow issues, water hammer, workability, inspectability, accessibility, equipment operability, quantity, and cost. The method for creating a route for piping, wiring, or duct according to claim 8,
A method for creating a route for piping, wiring, or ducts, characterized in that in the step (b), a route is created by alternately utilizing the route search algorithm and the results learned by the learning means.

Images