KR102668289B1 - Neural network-based piping and instrument diagram indexing method and apparatus - Google Patents

Abstract

The present invention discloses a neural network-based process flow indexing method and device. According to the invention, a processor; and a memory connected to the processor, generating a learning data set including images and detailed information for each element including pipes, valves, and instruments, and creating a neural network model using the generated learning data set. A process of learning, inputting a process flow diagram into the learned neural network model, extracting elements included in the process flow diagram, and storing program instructions executable by the processor to generate a control system list corresponding to the extracted elements. A flow chart indexing device is provided.

Description

Neural network-based piping and instrument diagram indexing method and apparatus}

The present invention relates to a neural network-based process flow diagram indexing method and device, and more specifically, to a method and device that can quickly and easily generate a control system list regarding work status from a process flow diagram including complex piping and instruments. .

The design of the control system during the design of a petrochemical plant begins with the task of creating a control system list (Instrument Index) from the process flow diagram (or process piping and instrument diagram, P&ID) to express the instrumentation and control characteristics in an easy-to-understand and concise manner. do.

This work must be performed by a person with specialized knowledge in the relevant field (skilled technician with at least 3 years of experience) to reduce errors and enable smooth follow-up work.

Even with the advancement of science and technology, this part is not easily automated because it is based on human image recognition ability and the basic technology of the control system, and it requires skilled engineers.

For this reason, in the factory control system design practice, a large number of temporary skilled workers are needed when carrying out a project and then are assigned to simple tasks, which reduces the efficiency of manpower operation, or the existing workforce is excessively focused on work (overworked) due to a lack of skilled manpower. ), the reality is that difficulties in supply and demand of manpower, including the problem of a sharp decline in individual life satisfaction, are also occurring.

Republic of Korea Patent Publication No. 10-2185433

In order to solve the problems of the prior art described above, the present invention seeks to propose a neural network-based process flow indexing method and device that can improve the productivity and quality of factory control system design work without skilled technicians.

In order to achieve the above-described object, according to an embodiment of the present invention, a process flow indexing device includes: a processor; and a memory connected to the processor, generating a learning data set including images and detailed information for each element including pipes, valves, and instruments, and creating a neural network model using the generated learning data set. A process of learning, inputting a process flow diagram into the learned neural network model, extracting elements included in the process flow diagram, and storing program instructions executable by the processor to generate a control system list corresponding to the extracted elements. A flow chart indexing device is provided.

The program commands may generate a learning data set by performing a preprocessing process on the device according to the device's properties.

The trigger is divided into a first trigger consisting of one element and a second trigger consisting of a plurality of elements according to the properties of the trigger, and the learning data for the second trigger is an image for each of the plurality of elements. and may be composed of a combination of detailed information.

The program commands may add images and detailed information about elements not extracted from the learned neural network model to the learning data set, and learn the neural network model using the added training data set.

The program commands can set bounding boxes of different colors depending on the type of element extracted from the learned neural network model.

According to another aspect of the present invention, there is a method of indexing a process flow diagram in a device including a processor and memory, generating a learning data set including images and detailed information for each of the elements including piping, valves, and instruments. step; Learning a neural network model using the generated training data set; Inputting a process flow diagram into the learned neural network model to extract elements included in the process flow diagram; And a process flow indexing method is provided, including the step of generating a control system list corresponding to the extracted elements.

According to another aspect of the present invention, a computer readable program for performing the above method is provided.

According to the present invention, there is an advantage in that process flow indexing work, which in the past was only possible for intermediate engineers with more than 3 years of experience, can be done by a non-skilled engineer who is a beginner engineer simply by logging into the project, opening the process flow diagram, and pressing the execution button.

In addition, according to the present invention, the productivity of skilled engineers can be significantly increased by automatically reporting errors by automatically comparing the AI program (AI has already reflected the changes) between the previously prepared output and the changed process flow diagram for the review work. There is an advantage.

Furthermore, according to the present invention, there is an advantage of preventing task mixing, error reexamination, and delay in subsequent processes by automatically storing the latest modified information in the DB and making it immediately available in subsequent processes.

1 is a diagram illustrating the configuration of a neural network-based process flow indexing device according to a preferred embodiment of the present invention.
Figure 2 is a diagram for explaining the learning and indexing process of a neural network model according to an embodiment of the present invention.
Figure 3 is a diagram illustrating a first instrument and its detailed information by way of example.
Figure 4 is a diagram illustrating a second instrument and its detailed information by way of example.
Figure 5 is a diagram for explaining the element extraction process using a neural network model according to this embodiment.

Since the present invention can make various changes and have various embodiments, specific embodiments will be illustrated in the drawings and described in detail.

However, this is not intended to limit the present invention to specific embodiments, and should be understood to include all changes, equivalents, and substitutes included in the spirit and technical scope of the present invention.

Figure 1 is a diagram showing the configuration of a neural network-based process flow indexing device according to a preferred embodiment of the present invention.

As shown in FIG. 1, the neural network-based factory control system indexing device may include a processor 100 and a memory 102.

Here, the processor 100 may include a central processing unit (CPU) capable of executing a computer program or another virtual machine.

Memory 102 may include a non-volatile storage device, such as a non-removable hard drive or a removable storage device. Removable storage devices may include compact flash units, USB memory sticks, etc. Memory 102 may also include volatile memory, such as various types of random access memory.

In this memory 102, program instructions executable by the processor 100 are stored.

Program instructions according to an embodiment of the present invention generate a learning data set containing images and detailed information for each of the elements including pipes, valves, and instruments, and use the generated learning data set to create a neural network model. Learning, inputting the process flow diagram into the learned neural network model, extracting elements included in the process flow diagram, and generating a control system list corresponding to the extracted elements.

The process flow diagram (process piping diagram) expresses the flow of all piping, and displays all process equipment, power machinery, piping, process control, and instruments required during normal operation, emergency operation, trial run, and operation stoppage of the process. It is written so that the interrelationship between them can be seen.

Generally, the process flow diagram includes detailed information on elements including piping, valves, and instruments, as shown in Table 1 below.

One Names, equipment numbers, and main specifications of power machinery, devices, and equipment 2 Nominal diameter of pipe, pipe classification symbol, material, nominal pressure of flange 3 Valves and piping accessories 4 Maintaining heat and keeping pipes and equipment warm and cold 5 Number, type and function of instruments 6 Condition when the control valve stops operating 7 Safety valve size and set pressure 8 Interlock and shutdown systems

In a process flow diagram, power machinery, devices, facilities, piping, valves, and instruments are expressed as images with a preset symbol system.

However, depending on the designer, it may be expressed as an image other than the standardized image.

When the process flow diagram for factory control cannot be shown in one drawing, an interface number is assigned and shown in multiple drawings.

Generally, a process flow diagram includes a large number of elements, and a control system list containing detailed information about each of them must be created, so an experienced engineer checks the element image in the process flow diagram and records detailed information as shown in Table 1. There is a problem that needs to be addressed.

Accordingly, the present invention proposes an efficient indexing method based only on nerves.

According to a preferred embodiment of the present invention, a learning data set containing images and detailed information for each of the elements including pipes, valves, and instruments is created, and a neural network model is learned using the generated learning data set. .

Figure 2 is a diagram for explaining the learning and indexing process of a neural network model according to an embodiment of the present invention.

As shown in FIG. 2, when a process flow diagram is input for learning a neural network model (step 200), a preprocessing process is performed according to the properties of the device (step 202).

The preprocessing process of step 202 can be defined as a process of classifying the image of the device included in the process flow diagram according to the device's properties.

The device according to this embodiment may include a first device composed of one element and a second device composed of a plurality of elements depending on the properties of the device.

A secondary device may be defined as one device combined with another valve or other device, or a valve or other device included within a device.

FIG. 3 is a diagram illustrating a first instrument and its detailed information, and FIG. 4 is a diagram illustrating a second instrument and its detailed information by way of example.

As shown in FIG. 3, learning data for the first device may include one image and detailed information about the first device.

On the other hand, as shown in FIG. 4, learning data for a second instrument composed of a plurality of elements may include images of each of the plurality of elements and detailed information about each element.

In this way, preprocessing is performed according to the properties of the instrument to generate a learning data set (step 204), and a neural network model is learned using the generated learning data set (step 206).

The neural network model according to this embodiment is for extracting features from images and may be VGGNet or ResNet.

After learning of the neural network model is completed, the process flow indexing device according to this embodiment extracts elements included in the input process flow diagram using the learned neural network model and generates a control list (step 208).

Figure 5 is a diagram for explaining the element extraction process using a neural network model according to this embodiment.

For user convenience, the device according to this embodiment can set the bounding box to a different color depending on the type of element.

For example, as shown in FIG. 5, the process flow indexing device may set the bounding box for the instrument to green and the piping to blue.

On the other hand, a red boundary box can be set for elements that could not be extracted due to lack of prior training in the neural network model.

In this case, in order to increase extraction accuracy, the process flow indexing device adds images and detailed information about elements that were not extracted from the learned neural network model to the learning data set, and uses the added learning data set to create a neural network model. learn

According to the present invention, it is possible to generate a control system list containing detailed information about a process flow chart composed of complex elements based on a neural network, thereby further improving user convenience.

The above-described embodiments of the present invention have been disclosed for illustrative purposes, and those skilled in the art will be able to make various modifications, changes, and additions within the spirit and scope of the present invention, and such modifications, changes, and additions will be possible. should be regarded as falling within the scope of the patent claims below.

Claims (10)

As a process flow indexing device,
processor; and
Including a memory connected to the processor,
Create a learning data set containing images and detailed information for each of the elements including piping, valves, and instruments,
Learn a neural network model using the generated training data set,
Input the process flow diagram into the learned neural network model to extract elements included in the process flow diagram,
To generate a control system list corresponding to the extracted elements,
Store program instructions executable by the processor,
For the instrument, a preprocessing process is performed according to the properties of the instrument to generate a learning data set,
The trigger is divided into a first trigger consisting of one element and a second trigger consisting of a plurality of elements according to the properties of the trigger,
The learning data for the second instrument consists of a combination of images and detailed information for each of the plurality of elements,
The program commands are:
Bounding boxes of different colors are set according to the type of element extracted from the learned neural network model,
A red boundary box is set for elements that could not be extracted due to lack of prior learning.
Adding images and detailed information about elements that were not extracted from the learned neural network model to the learning data set,
A process flow indexing device that learns the neural network model using the added training data set. delete delete delete delete A method of indexing a process flow diagram in a device including a processor and memory, comprising:
generating a learning data set containing images and detailed information for each of the elements including pipes, valves, and instruments;
Learning a neural network model using the generated training data set;
Inputting a process flow diagram into the learned neural network model to extract elements included in the process flow diagram; and
Including generating a control system table corresponding to the extracted elements,
In the step of generating the learning data set, a preprocessing process is performed on the instrument according to the properties of the instrument to generate a learning data set,
The trigger is divided into a first trigger consisting of one element and a second trigger consisting of a plurality of elements according to the properties of the trigger,
The learning data for the second instrument consists of a combination of images and detailed information for each of the plurality of elements,
Bounding boxes of different colors are set depending on the type of element extracted from the learned neural network model,
A red boundary box is set for elements that could not be extracted due to lack of prior learning.
The step of generating the learning data set includes adding images and detailed information about elements that were not extracted from the learned neural network model to the learning data set, and learning the neural network model using the added learning data set. Process flow indexing method. delete delete delete A computer program stored in a computer-readable recording medium that performs the method according to claim 6.