KR102307132B1 - Machine learning automation platform device for decision support in plant engineering - Google Patents

Abstract

It relates to a data analysis device that utilizes the entire engineering cycle data, and the analysis device using the engineering whole cycle data includes a data collection unit that collects project data generated in the entire cycle of the engineering industry, and a data input unit that receives new project data. , a machine learning unit that performs data analysis by applying the new project data received from the data input unit to a learning module learned based on machine learning, and a visualization unit that visualizes and provides the analysis result of the machine learning unit.

Description

Machine learning automation platform device for decision support for each stage of plant engineering work

The present application relates to a data management and data analysis device using data from the entire engineering cycle.

In the engineering industry, various data is generated at each stage from the planning stage corresponding to the entire business cycle to design, purchase, construction, operation, and predictive maintenance of plant facilities in O&M. Various indicators such as MH (man hour) and cost are calculated to minimize the risk.

However, in the current engineering industry, a system for collecting and utilizing data is not systematically equipped. Therefore, when calculating various indicators using data, it can cause various problems including human error, as well as change the situation in the middle of the business. The problem is that it takes a long time to figure out. In addition, these problems may bring a large amount of damage in carrying out the engineering business.

Therefore, it is necessary to study engineering data management and data analysis devices so that risk factors can be analyzed through efficient management of data generated at each stage of the engineering industry, and users can respond in advance and support decision-making.

The background technology of the present application is disclosed in Korean Patent Publication No. 10-1229274.

The present application is to solve the problems of the prior art described above, input project data occurring in the entire cycle of the engineering industry, support the user's desired project data search function, ITB analysis, design cost prediction, design error analysis, design The purpose of this is to provide a data management and data analysis device using data from the entire engineering cycle that can perform change analysis and predictive maintenance of plant facilities using an engineering machine learning platform.

An object of the present application is to provide a data management and data analysis apparatus using data of the entire engineering cycle that can be provided by visualizing the analyzed results in the form of a dashboard in order to solve the problems of the prior art described above.

The present application is intended to solve the problems of the prior art described above, and provides an analysis result to be downloaded in a specific standardized format, and provides an API to provide scalability so that the results can be utilized in other applications. It aims to provide a data management and data analysis device using periodic data.

The present application is intended to solve the problems of the prior art described above, and when a user inputs a new project, it outputs the most similar project compared to the previously input project, and visualizes and provides statistical analysis results for the searched project. It aims to provide a data management and data analysis device using data from the entire engineering cycle that can be used.

However, the technical problems to be achieved by the embodiments of the present application are not limited to the technical problems as described above, and other technical problems may exist.

As a technical means for achieving the above technical task, the data analysis device using the engineering whole cycle data according to an embodiment of the present application is a data collection unit that collects project data generated in the entire cycle of the engineering industry, a new project A data input unit that receives data, a machine learning unit that performs data analysis by applying the new project data received from the data input unit to a learning module learned based on machine learning, and a visualization that visualizes the analysis results of the machine learning unit may include wealth.

Also, the data input unit may receive user input information for selecting any one analysis step item to be analyzed among a plurality of analysis step items of the engineering industry.

In addition, the plurality of analysis step items of the engineering industry may include at least one of a Bidding analysis item, an Engineering analysis item, a Construction & Commissioning analysis item, and an O&M analysis item.

In addition, the learning module learned based on machine learning may include at least one learning module of a regression model, a classification model, a cluster model, and a deep learning model.

In addition, the machine learning unit applies the new project data to the learning module learned based on machine learning to detect the toxin clause included in the ITB document data for the ITB analysis item among the plurality of analysis step items of the engineering industry. The toxin clause analysis can be performed.

In addition, the machine learning unit, predictive analysis of MH (Man Hour) by applying the new project data to the learning module learned based on the machine learning with respect to the design cost prediction analysis item among the plurality of analysis step items of the engineering industry can be done

In addition, the machine learning unit may perform a project delay date analysis by applying the new project data to the learning module learned based on the machine learning with respect to a design error analysis item among a plurality of analysis step items of the engineering industry.

In addition, the machine learning unit may perform a change amount analysis by applying the new project data to the learning module learned based on the machine learning with respect to a design change analysis item among a plurality of analysis step items of the engineering industry.

In addition, the machine learning unit may perform maintenance item detection analysis by applying the new project data to the learning module learned based on the machine learning with respect to the predictive maintenance analysis item among the plurality of analysis step items of the engineering industry.

In addition, the machine learning unit, for each of the plurality of analysis step items of the engineering industry, collects the learning results performed by applying to a plurality of learning models, and collects the learning results of the learning model showing the highest accuracy of the plurality of the engineering industry It can be provided as an analysis result for the items of the analysis stage of

In addition, the data input unit receives user input information for selecting a plurality of search items related to a project to be searched from among the plurality of project data collected by the data collection unit, and the machine learning unit includes the user input information and A similarity analysis between the plurality of project data may be performed, and at least one project list may be provided based on a similarity analysis result.

According to an embodiment of the present application, the data analysis method using the engineering whole cycle data includes the steps of collecting project data occurring in the entire cycle of the engineering industry, receiving new project data, and analyzing the new project data. It may include a step of performing data analysis by applying it to a learning module learned based on machine learning, and a step of providing a visualization result of the analysis.

The above-described problem solving means are merely exemplary, and should not be construed as limiting the present application. In addition to the exemplary embodiments described above, additional embodiments may exist in the drawings and detailed description.

According to the above-described problem solving means of the present application, it is possible to minimize risks through the process of analyzing and visualizing data generated in the entire cycle of the engineering industry in the machine learning unit.

According to the above-described problem solving means of the present application, risk factors can be analyzed through efficient management of data generated at each stage of the engineering industry, and the user can respond in advance and support decision-making.

However, the effects obtainable herein are not limited to the above-described effects, and other effects may exist.

1 is a schematic block diagram of a data analysis apparatus utilizing engineering full-cycle data according to an embodiment of the present application.
2 is a diagram schematically showing a plurality of analysis step item selection screens of the data analysis apparatus according to an embodiment of the present application.
3 is a diagram schematically illustrating an analysis result of an ITB document of the data analysis apparatus according to an embodiment of the present application.
4 is a diagram schematically illustrating a design cost prediction result of the data analysis apparatus according to an embodiment of the present application.
5 is a diagram schematically illustrating a design error analysis result of the data analysis apparatus according to an embodiment of the present application.
6 is a diagram schematically illustrating a design change analysis result of the data analysis apparatus according to an embodiment of the present application.
7 is a diagram schematically illustrating a result of predictive maintenance analysis of the data analysis apparatus according to an embodiment of the present application.
8 is a diagram schematically illustrating a screen providing an API of a data analysis apparatus according to an embodiment of the present application.
9 is a diagram schematically illustrating a data search providing screen and a search result of the data analysis apparatus according to an exemplary embodiment of the present disclosure.
10 is a diagram schematically illustrating a statistical visualization result for a similar project of the data analysis apparatus according to an embodiment of the present application.
11 is an operation flowchart of a data analysis method using engineering full cycle data according to an embodiment of the present application.

Hereinafter, embodiments of the present application will be described in detail with reference to the accompanying drawings so that those of ordinary skill in the art to which the present application pertains can easily implement them. However, the present application may be embodied in several different forms and is not limited to the embodiments described herein. And in order to clearly explain the present application in the drawings, parts irrelevant to the description are omitted, and similar reference numerals are attached to similar parts throughout the specification.

Throughout this specification, when a part is "connected" with another part, it is not only "directly connected" but also "electrically connected" or "indirectly connected" with another element interposed therebetween. "Including cases where

Throughout this specification, when it is said that a member is positioned "on", "on", "on", "under", "under", or "under" another member, this means that a member is positioned on the other member. It includes not only the case where they are in contact, but also the case where another member exists between two members.

Throughout this specification, when a part "includes" a component, it means that other components may be further included, rather than excluding other components, unless otherwise stated.

The machine learning automation platform device for supporting decision-making for each plant engineering task of the present application may be the name of the present application, and may be referred to differently like a data analysis device utilizing data at the end of a claim (data analysis device) and the entire engineering cycle data.

This application relates to an automated machine learning-based engineering platform that supports optimal decision-making by utilizing data generated in the entire cycle of the engineering industry. It is about a method and system to minimize risks that can occur in the entire machine learning cycle by analyzing it, and to search for and utilize projects similar to the input data.

Engineering provides engineering and technical services necessary for planning, design, development, construction, and operation of engineering systems by integrating scientific and technical expertise. The engineering industry is characterized as a knowledge-based service industry in which integrated technology development is essential, such as batch order ordering by a small number of clients, project management, design, equipment procurement, and construction. In addition, plant engineering is a production facility and related system in which a series of mechanical devices are linked to realize continuous manufacturing from raw materials to intermediate goods or final products under normal operating conditions. It is a field that directly affects productivity, performance, and quality throughout the entire life cycle of plant planning, design, construction, operation and disposal.

According to an embodiment of the present application, the data analysis device 10 may collect project data generated in the entire cycle of the engineering industry, and support a project data search function desired by a user. In addition, the data analysis apparatus 10 may analyze and provide ITB analysis, design cost prediction, design error analysis, design change analysis, predictive maintenance, and the like using a machine learning platform. The data analysis apparatus 10 may provide the analyzed result through a user terminal (not shown). In addition, the data analysis apparatus 10 may provide visualization in the form of a dashboard so that a user (administrator) can view the analysis result and make an effective decision with quick judgment.

In addition, the data analysis apparatus 10 may provide the analyzed result to be downloaded in a specific standardized format. In addition, the data analysis apparatus 10 may consider extensibility such as providing an API to utilize the results in other applications.

In addition, in consideration of the characteristics of the continuously changing and added engineering data, the data analysis device 10 allows to add new data by dividing it into stages, and easily converts the input data through an automated machine learning-based learning model. Analyze and provide results.

According to an embodiment of the present application, the data analysis apparatus 10 may provide a data input menu, a data management menu, and a data analysis menu to a user terminal (not shown). For example, a user terminal (not shown) may download and install an application program provided by the data analysis apparatus 10 , and provide a data input menu, a data management menu, and a data analysis menu through the installed application.

The data analysis apparatus 10 may include all kinds of servers, terminals, or devices that transmit and receive data, content, and various communication signals to and from a user terminal (not shown) through a network, and have functions of data storage and processing. .

A user terminal (not shown) is a device that interworks with the horse integrated apparatus 110 through a network, for example, a smartphone, a smart pad, a tablet PC, a wearable device, and the like, a Personal Communication System (PCS). ), GSM (Global System for Mobile communication), PDC (Personal Digital Cellular), PHS (Personal Handyphone System), PDA (Personal Digital Assistant), IMT (International Mobile Telecommunication)-2000, CDMA (Code Division Multiple Access)-2000 , W-CDMA (W-Code Division Multiple Access), and Wibro (Wireless Broadband Internet) terminals of all types of wireless communication devices and desktop computers and fixed terminals such as smart TVs.

An example of a network for sharing information between the data analysis device 10 and the user terminal (not shown) is a 3rd Generation Partnership Project (3GPP) network, a Long Term Evolution (LTE) network, a 5G network, and a World Interoperability for Microwave Access (WIMAX) network. ) network, wired and wireless Internet, LAN (Local Area Network), Wireless LAN (Wireless Local Area Network), WAN (Wide Area Network), PAN (Personal Area Network), Bluetooth (Bluetooth) network, Wifi network, NFC ( A Near Field Communication) network, a satellite broadcasting network, an analog broadcasting network, a Digital Multimedia Broadcasting (DMB) network, etc. may be included, but is not limited thereto.

1 is a schematic block diagram of a data analysis apparatus utilizing engineering full-cycle data according to an embodiment of the present application.

Referring to FIG. 1 , the data analysis apparatus 10 may include a data collection unit 11 , a data input unit 12 , a machine learning unit 13 , and a visualization unit 14 . However, the configuration of the data analysis apparatus 10 is not limited thereto. For example, the data analysis apparatus 10 may include a database (not shown) for storing a plurality of project data collected by the data collection unit 11 . Also, the data analysis apparatus 10 may include a data providing unit (not shown) for providing a data analysis result to a user terminal (not shown). The database (not shown) may include a toxin clause dictionary, a design error delay severity classification item, a design change project cost severity classification item, a design change key word dictionary, etc.

According to an embodiment of the present application, the data collection unit 11 may collect project data generated in the entire cycle of the engineering industry. For example, the data collection unit 11 includes contracts, specifications, ITB, offshore and land plant data, ERP (Enterprise Resources Management), PMIS (Project Management Information System), commercial data, public data, big data integrated information, public It can collect data, Open API, etc. The data collection unit 11 may collect various data generated in the entire cycle of the engineering industry. In addition, the data collection unit 11 may collect project data generated in the entire cycle of industries in various fields as well as project data of the engineering industry. The project data may be time series data generated in the engineering industry. In addition, the project data may be data stored in a database (not shown). In addition, the data collection unit 11 may collect project data generated in the entire cycle of the engineering industry from an external server through a network. As an example, the external server may be a server of an institution performing an engineering industry.

According to another exemplary embodiment of the present disclosure, the data collection unit 11 may manage data analyzed by the data analysis apparatus 10 . In other words, the data collection unit 11 may store the new project data received from the data input unit 12 in a database (not shown). Also, the data collection unit 11 may store analysis results of various data analyzed by the machine learning unit 13 in a database (not shown). The data collection unit 11 may not only collect various data generated in the engineering industry, but also store and manage analysis data obtained by analyzing data generated in the engineering industry in a database (not shown).

According to an embodiment of the present application, the data input unit 12 may receive new project data. New project data may include documents originating in the engineering industry, such as contracts, specifications, and ITBs. For example, the data input unit 12 may receive new project data from a user terminal (not shown). In addition, the data input unit 12 may receive new project data and data generated in a plurality of stages of the engineering industry, respectively. In addition, the data input unit 12 may receive newly created project data rather than the existing project data of the engineering industry. In other words, the data input unit 12 may receive data related to a plurality of projects from a data set stored in a database (not shown). Also, the data input unit 12 may receive project data collected by the data collection unit 11 .

2 is a diagram schematically showing a plurality of analysis step item selection screens of the data analysis apparatus according to an embodiment of the present application.

According to an embodiment of the present application, the data input unit 12 may receive user input information for selecting any one analysis step item to be analyzed among a plurality of analysis step items in the engineering industry. In addition, the plurality of analysis step items of the engineering industry may include at least one of a Bidding analysis item, an Engineering analysis item, a Construction & Commissioning analysis item, and an O&M analysis item. Referring to FIG. 2 , the bidding analysis item may include an ITB analysis item. In addition, it may include an engineering analysis item, a design cost analysis item, a design error analysis item, and a design change analysis item. In addition, the Construction & Commissioning analysis item may include additional items. In addition, the O&M analysis item may include a predictive maintenance analysis item. Each of the plurality of analysis step items in the engineering industry may include an additional item. The additional item may be an item for entering data of a new subject. The data input unit 12 may receive all of the new projects including a plurality of data, and each project data may also be input separately. In other words, when the user intends to perform only the bidding analysis, the data input unit 12 may receive data for performing the bidding analysis based on user selection information of an additional item included in the bidding analysis item.

For example, the data input unit 12 may receive user input information for selecting at least one of a plurality of analysis items from a user terminal (not shown). Referring to FIG. 2 , the data input unit 12 may receive user input information for selecting a design cost analysis item included in an engineering analysis item among a plurality of analysis items from a user terminal (not shown). The data input unit 12 may provide user input information provided from a user terminal (not shown) to the machine learning unit 13 .

In other words, the visualization unit 14 provides a plurality of analysis step items of the engineering industry to a user terminal (not shown), and the data input unit 12 is a user terminal (not shown) of a plurality of analysis step items of the engineering industry. At least one analysis item to be analyzed is provided, and the machine learning unit 13 may learn new project data in response to the received analysis item.

According to an embodiment of the present application, the machine learning unit 13 may perform data analysis by applying the new project data received from the data input unit 12 to a learning module learned based on machine learning. The machine learning unit 13 may perform data analysis by applying the new data to the learning module learned based on machine learning based on the new project data input from the data input unit 12 and the user's analysis item selection information.

According to another embodiment of the present application, the machine learning unit 13 may perform data analysis by applying the project data collected by the data collection unit 11 to a learning module learned based on machine learning. In other words, the machine learning unit 13 can perform data analysis by applying the existing project data stored in the database (not shown) to the learning module learned based on machine learning, not the data input from the data input unit 12 . have. The machine learning unit 13 may perform learning to detect the poison clause of the ITB document data through the learning module learned based on machine learning. In addition, the machine learning unit 13 may perform learning to predict MH (Man Hour) in design cost analysis through a learning module learned based on machine learning. In addition, the machine learning unit 13 may perform learning to predict the delay analysis in the design error analysis through the learning module learned based on machine learning. In addition, the machine learning unit 13 may perform learning to analyze the change amount in the design change analysis through the learning module learned based on machine learning. In addition, the machine learning unit 13 may perform learning to analyze the detection of maintenance items in the predictive maintenance analysis through the learning module learned based on machine learning.

In addition, the learning module learned based on machine learning may include at least one learning module of a regression model, a classification model, a cluster model, and a deep learning model. In addition, the machine learning unit 13 may include a feature engineering function. Feature engineering functions may include data exploration, data transformation, and feature selection functions.

As an example, the machine learning unit 13 may increase model performance by applying hyperparameters for each algorithm (for each learning model). Hyperparameters are values that influence the performance of the model, such as the number of neurons in each layer, batch size, learning rate when updating parameters, and weight reduction. The machine learning unit 13 may set a range of hyperparameter values to optimize hyperparameters. In addition, the machine learning unit 13 may randomly extract a hyperparameter value from a set range. (Step 1) In addition, the machine learning unit 13 may evaluate the accuracy with the verification data after learning the model using the previously sampled hyperparameter values. (Step 2) The verification data may be data for evaluating the adequacy of hyperparameters. The machine learning unit 13 may repeat steps 1 and 2 described above a preset number of times to determine the accuracy and reset the range of the hyper parameter. The machine learning unit 13 may reset the hyper-parameter in a narrowing direction. The machine learning unit 13 may select the hyperparameter value from the narrowed range. The embodiment in which the model performance is increased by applying the hyperparameters described above is only one embodiment, and is not limited thereto, and more various embodiments may be applied.

3 is a diagram schematically illustrating an analysis result of an ITB document of the data analysis apparatus according to an embodiment of the present application.

According to an embodiment of the present application, the machine learning unit 13 is a learning module learned based on machine learning to detect a toxin clause included in the ITB document data for the ITB analysis item among a plurality of analysis step items in the engineering industry. Toxin clause analysis can be performed by applying new project data. As an example, the ITB analysis is to detect the toxin clause within the ITB (Invitation To Bid: bidding guide), and the machine learning unit 13 applies the ITB document included in the new project data to the learning module to analyze the toxin clause. can Here, the learning module may include at least one of a regression model, a classification model, a cluster model, and a deep learning model. In general, the toxin clause refers to a content that cleverly limits the original intention in a law or official document. That is, in the case of a law, it means that the law has an intended purpose, but a phrase that blocks the intention in theory or reality is inserted.

Illustratively, the machine learning unit 13 may perform a toxin clause analysis included in the ITB document data through comparison with a pre-built toxin clause dictionary. The machine learning unit 13 may perform the toxin clause analysis by matching any one of the toxin clause words included in the toxin clause dictionary stored in advance in the received ITB document data. The toxin clause dictionary may include words preset by the user. For example, the machine learning unit 13 may receive the ITB document from the data input unit 12 . The machine learning unit 13 may separate morphemes of terms included in the input ITB document for standardization by using the toxin clause dictionary. In addition, the machine learning unit 13 may perform a similarity calculation using an artificial intelligence-based algorithm to compare a plurality of words included in the separated term with the toxin clause words included in the toxin clause dictionary. Here, the AI-based algorithm may be a fuzzy data matching algorithm, but is not limited thereto. The Fuzzy Data Matching algorithm is an algorithm that performs matching between data using the result calculated based on the editing distance (Levenshtein Distance). However, the AI-based algorithm described above is only an example, and is not limited thereto, and various neural network systems that have been developed or developed in the future may be applied.

According to an embodiment of the present application, referring to FIG. 3A , the machine learning unit 13 displays a plurality of toxin clause analysis selection items divided by category through the visualization unit 14 in a user terminal (not shown). can be provided as The user may select an ITB document to perform ITB analysis through the ITB analysis item provided to the user terminal (not shown). For example, in FIG. 3( a ), the ITB_AA document is selected. The user may select a category to be analyzed from among a plurality of categories. Each category may include toxin clause analysis selection items of different items in order to perform toxin clause analysis. Referring to FIG. 3A , the machine learning unit 13 provides a first toxin clause analysis selection item (Fit for purpose) and a second toxin clause analysis selection item (Open-ended) provided from a user terminal (not shown). cluse), 3rd poison clause analysis selection item (LD execution procedure), 4th poison clause analysis selection item (Payment options: pay-when-paid vs Pay-if-paid), 5th poison clause analysis selection item (Pre -payment: LOC with BL; SB LoC; LoC at-sight; Usance; Document against Payment DA/DP) Based on the information, an ITB analysis can be performed. Referring to (b) of FIG. 3 , the machine learning unit 13 applies the subject, verb, and object to each of the first toxin clause analysis selection item (Fit for purpose) in the sentence of the data included in the ITB document. Corresponding toxin clauses can be extracted. In addition, the machine learning unit 13 may extract a toxin clause corresponding to each of the subject, verb, and object corresponding to the fifth toxin clause analysis selection item (pre-payment) from the sentence of data included in the ITB document. . In addition, the visualization unit 14 may provide the analysis content extracted in response to the toxin clause analysis selection item in the machine learning unit 13 to the user terminal (not shown).

4 is a diagram schematically illustrating a design cost prediction result of the data analysis apparatus according to an embodiment of the present application.

According to an embodiment of the present application, the machine learning unit 13 applies new project data to a learning module learned based on machine learning for a design cost prediction analysis item among a plurality of analysis step items in the engineering industry to apply the new project data to the new project data. It is possible to perform data analysis to predict MH (Man Hour). In addition, the machine learning unit 13 may perform an analysis of predicting the number of design hours based on input information in which an engineering unit price is input for each country (country) based on a result predicted by the MH (Man Hour).

For example, referring to FIG. 4 , the machine learning unit 13 may input data necessary for predicting design cost included in the new project data to the learning module and predict MH (Man Hour). The machine learning unit may predict a man hour (MH) corresponding to each of the classified items (discipline). The visualization unit 14 may provide the predicted Man Hour (MH) to the user terminal (not shown). The data input unit 12 may receive user input information for selecting at least one of a plurality of countries (countries) from a user terminal (not shown). The machine learning unit 13 may predict the number of design hours based on input information of selecting the user's country (country). The number of design hours may be an engineer's working hours.

5 is a diagram schematically illustrating a design error analysis result of the data analysis apparatus according to an embodiment of the present application.

According to an embodiment of the present application, the machine learning unit 13 analyzes the project delay date by applying the new project data to the learning module learned based on machine learning for the design change analysis item among the plurality of analysis step items in the engineering industry. can be done The machine learning unit 13 may perform the design error delay severity analysis of the new project based on the design error delay severity classification item for the project delay date that may occur in the new project. The delay severity classification item may include items quantified by classifying safety, alertness, risk, and severity for each of the first and second construction periods. The learning module learned based on machine learning may be a module built through learning that takes the project period as an input and the severity of the project delay day as an output.

For example, referring to FIG. 5 , the machine learning unit 13 may extract a construction period date from the new project data. In the machine learning unit 13, the extracted construction period date (project period (PJT Months)) corresponds to the first construction period (eg, 26 months or more) or the second construction period (eg, 26 months or less) You can decide whether to do it or not. The machine learning unit 13 may classify at least one error reason among a plurality of error reasons analysis by using the new project data, and may associate the classified error reason with a specific error type. In other words, the machine learning unit 13 may analyze the reason for the error of the new project data and perform the project delay date analysis. For example, when the error reason is the first error reason (valve access to be considered), the machine learning unit 13 may classify the error type as M1. In addition, when the error reason is the second error reason (drain to be provide), the machine learning unit 13 may classify the error type as D1. Also, when the error reason is a third error reason (clear clash between popings), the machine learning unit 13 may classify the error type as H1. In addition, when the error reason is a fourth error reason (low pocket remove on tagged piping to be considered), the machine learning unit 13 may classify the error type as O1. Also, when the error reason is a fifth error reason (duplicated line number with P-101113 suction line to be corrected), the machine learning unit 13 may classify the error type as C1. The machine learning unit 13 may perform a design error delay severity analysis based on an error type, an error reason, a delay date (Delay), and a project period (PJT Months). The machine learning unit 13 considers the error type, error reason, delay date, and project period (PJT Months), and analyzes the design error delay severity analysis as at least one of safety, boundary, risk, and severe. have. As an example, the machine learning unit 13 extracts the project period (PJT Months) from the new project data, applies it to a learning model with the project period as an input, and outputs at least one of safety, alertness, risk, and severe learning. can be performed.

6 is a diagram schematically illustrating a design change analysis result of the data analysis apparatus according to an embodiment of the present application.

According to an embodiment of the present application, the machine learning unit 13 performs a change amount analysis by applying new project data to a learning module learned based on machine learning for a design change analysis item among a plurality of analysis step items in the engineering industry. can do. The machine learning unit 13 may analyze the change amount of the new project based on the design change project cost severity classification item with respect to the project design change amount that may occur as a design change in the new project. In other words, the machine learning unit 13 applies the new project data to the learning module learned based on machine learning in response to the design change analysis item among the plurality of analysis step items in the engineering industry, and analyzes the text included in the new project data. can be performed. The machine learning unit 13 may perform a similarity calculation between the design change key word dictionary and the text included in the new project data, and analyze the change amount of the new projector based on the design change severity classification item.

Referring to FIG. 6 , the machine learning unit 13 may extract a design error word through text analysis included in the new project data. The machine learning unit 13 may extract the design change error type through comparative analysis of the design error word extracted through the text analysis included in the pre-stored design change key word dictionary and the new project data. The machine learning unit 13 may perform the design change severity analysis through the design change key word dictionary and the severity classification for each project cost section due to the design change. The design change key word dictionary may be a stored model classified into construction type, design error type, and design error key word.

For example, the machine learning unit 13 may separate morphemes of terms through text analysis included in the new project data. The machine learning unit 13 may extract a design error key word from the separated terms through a similarity calculation with the design change key word dictionary. The machine learning unit 13 may classify a design error type based on the extracted design error key word. The machine learning unit 13 may construct a data set by classifying it into key words of construction type, design error, type, and design error through text analysis included in the new project data. The machine learning unit 13 may perform an analysis of the change amount generated when the design is changed in response to the design change project cost severity classification item.

7 is a diagram schematically illustrating a result of predictive maintenance analysis of the data analysis apparatus according to an embodiment of the present application.

According to an embodiment of the present application, the machine learning unit 13 performs maintenance item detection analysis by applying new project data to a learning module learned based on machine learning for predictive maintenance analysis items among a plurality of analysis step items in the engineering industry. can be done For example, the machine learning unit 13 may generate a learning model based on at least one of a decision tree algorithm, a random forest algorithm, an SVM algorithm, and a KNN algorithm. For example, the Decision Tree algorithm divides the variable region into two for each branch, and the Random Forest algorithm is an algorithm in which numerous decision trees form a forest and average each prediction result into one result variable. The SVM algorithm is a non-probabilistic algorithm that determines the classification to which data belongs by classifying the boundary of the largest width in the data distribution space. It is an algorithm that Illustratively, the accuracy of the maintenance item detection and analysis using the SVM algorithm in the machine learning unit 13 represents an accuracy of 0.896.

Referring to FIG. 7 , the machine learning unit 13 may perform maintenance item detection and analysis by applying various algorithms. The machine learning unit 13 may perform a maintenance item detection analysis on the predictive maintenance analysis item using the SVM algorithm. The machine learning unit 13 may perform maintenance item detection analysis by applying the predictive maintenance analysis item included in the O&M analysis item to the SVM algorithm, and may represent the analysis result as a confusion matrix. The visualization unit 14 may provide an analysis result represented by a confusion matrix to a user terminal (not shown). The machine learning unit 13 may determine a final model for a plurality of analysis items based on the results of performing various algorithms.

According to an embodiment of the present application, the machine learning unit 13 may collect a learning result performed by applying to a plurality of learning models for each of a plurality of analysis step items in the engineering industry. In addition, the machine learning unit 13 may provide a learning result of the learning model showing the highest accuracy as an analysis result for a plurality of analysis step items in the engineering industry. For example, the machine learning unit 13 performs analysis by applying a plurality of machine learning learning models to each of the ITB analysis item, the design cost analysis item, the design error analysis item, the design change analysis item, and the predictive maintenance analysis item. can The machine learning unit 13 may apply a plurality of machine learning learning models to perform an analysis on each of a plurality of analysis items, and select a learning model showing the highest accuracy as a learning model of the corresponding analysis item. In other words, the machine learning unit 13 may analyze the toxin clause by applying the first learning model to the ITB analysis item. In addition, the machine learning unit 13 may perform predictive analysis of MH (Man Hour) by applying the second learning model to the design cost analysis item. In addition, the machine learning unit 13 may perform project delay analysis by applying the third learning model to the design error analysis item. In addition, the machine learning unit 13 may perform a change amount analysis by applying the fourth learning model to the design error analysis item. In addition, the machine learning unit 13 may perform a maintenance item detection analysis by applying the fifth learning model to the predictive maintenance analysis item.

For example, the machine learning unit 13 may calculate an evaluation index of each of the plurality of machine learning algorithms based on the analysis result data. The machine learning unit 13 may calculate an evaluation index by using a root mean squared error (RMSE) value for each result value of a plurality of machine learning algorithms corresponding to the regression algorithm type. Since the lower the root mean square error is, the higher the accuracy and reliability of the regression algorithm, the machine learning unit 13 calculates the evaluation index of each of the plurality of machine learning algorithms corresponding to the regression algorithm type, and ranks them in the order of the lowest root mean square error. can be calculated.

For example, a plurality of machine learning learning models may be generated based on a regression algorithm, a classification algorithm, a clustering algorithm, and a deep learning algorithm. Unsupervised learning refers to an algorithm that learns while analyzing or clustering data itself, rather than constructing learning data. The machine learning unit 13 may cluster and calculate the analysis patterns based on the clustering algorithm, and may detect a new analysis pattern based on the degree of separation between clusters of each analysis pattern. For example, a logistic regression algorithm, a random forest algorithm, a support vector machine (SVM) algorithm, a decision-making algorithm, and a clustering algorithm may be used as the clustering algorithm for unsupervised learning. In addition to the above algorithms, the machine learning unit 13 uses the Extra Tree algorithm, the XG Boost algorithm and the Deep Learning algorithm, the K-means clustering algorithm, the SOM (Self-Organizing-Maps) algorithm, and the EM & Canopy algorithm through clustering algorithms. Unsupervised learning can be performed. The Random Forest algorithm is an algorithm that consists of a forest of numerous decision trees and averages each prediction result into a single result variable. It is a non-stochastic algorithm. Extra Tree Algorithm is similar to Random Forest, but it is faster than Random Forest. XGBoost Algorithm is a boost algorithm that applies the result of XGBoost Tree to the next tree if the Tree of Random Forest is independent. The deep learning algorithm is an algorithm that learns by controlling the effect of variable patterns on the results with weights based on a multi-layered neural network. In addition, the K-means clustering algorithm is a traditional classification technique that iteratively subdivides the target group into K clusters based on the average value (similarity) of the distance. It is a learning and clustering technique. In addition, the EM & Canopy algorithm refers to a method of clustering by updating parameter values through an iterative process starting with the maximum possibility with a given initial value.

According to another embodiment of the present application, the machine learning unit 13 may perform pre-processing on the new project data input from the data input unit 12 . For example, the machine learning unit 13 may perform data purification (outlier detection, data correction) and ETL (extraction, transformation, loading) and the like from project data. In addition, the machine learning unit 13 may perform data normalization when the new project data is unstructured data. Data normalization may be to unify variable values of data included in new project data based on a certain standard. In addition, the machine learning unit 13 may build a learning model, evaluate the built learning model, and optimize the model.

For example, the machine learning unit 13 may receive analysis target data from a database (not shown) and the data input unit 12 . The machine learning unit 13 may receive project data stored in advance in a database (not shown) as analysis target data. In addition, the machine learning unit 13 may receive the new project data from the data input unit 12 as analysis target data. The machine learning unit 13 may extract a characteristic variable from the analysis target data based on feature engineering. The data to be analyzed may include text-based unstructured data such as symbols, words, and sentences as well as standardized data including numeric variables such as numbers. The machine learning unit 13 may extract a characteristic variable from structured data or unstructured data included in the analysis target data.

According to an embodiment of the present disclosure, the data input unit 12 may perform a function of registering a word dictionary so that a feature variable optimized for a user can be calculated when a feature variable is calculated from the unstructured data. The machine learning unit 13 may extract a characteristic variable based on an unsupervised learning-based natural language processing algorithm in the case of unstructured data among the data to be analyzed. Specifically, the calculation of the feature variable from the unstructured data will be described. When the analysis target data includes unstructured data, the machine learning unit 13 may extract an optimal vector value from the user-optimized word dictionary. Also, the machine learning unit 13 may extract a noun from the text included in the unstructured data through principal component analysis (PCA) on the optimal vector value. The Soynlp algorithm may be used as an algorithm for principal component analysis, but is not limited thereto. The Soynlp algorithm is an unsupervised learning-based algorithm that can not only extract words from the data to be analyzed without requiring separate learning data, but also decompose sentences into word sequences or discriminate parts of speech. The machine learning unit 13 may calculate the characteristic variable according to a score given according to the frequency of the word including the noun based on the word dictionary. For example, the machine learning unit 13 may give a higher score to the noun extracted from the text based on the word dictionary, that is, the higher the frequency of the word. The scoring method may be achieved through an absolute method of differentially assigning points according to a preset frequency, or a relative method of assigning points according to a relative ratio of words appearing in the data to be analyzed. (For example, a word in the top 10% of occurrences may be given a higher score than a word in the top 30%)

According to an exemplary embodiment of the present application, the visualization unit 14 may provide a visualization result of the analysis of the machine learning unit 13 . The visualization unit 14 may visualize and provide the results analyzed by the machine learning unit 13 by applying different GUIs to the analysis items.

For example, referring to FIG. 2 , the visualization unit 14 may provide a plurality of analysis step items to a user terminal (not shown). In addition, referring to FIG. 3A , the visualization unit 14 may provide an ITB document selection item, a category selection item, and a toxin clause analysis selection item for ITB analysis to a user terminal (not shown). In addition, referring to FIG. 3B , the visualization unit 14 may provide the ITB analysis result performed by the machine learning unit 13 to the user terminal (not shown). Also, referring to FIG. 4 , the visualization unit 14 may provide a design cost prediction result to a user terminal (not shown). Also, referring to FIG. 5 , the visualization unit 14 may provide a design error analysis result to a user terminal (not shown). Also, referring to FIG. 6 , the visualization unit 14 may provide a design change analysis result to a user terminal (not shown). Also, referring to FIG. 7 , the visualization unit 14 may provide a predictive maintenance analysis result to a user terminal (not shown).

Also, the visualization unit 14 may visualize and provide the variable properties and variable value distribution of the extracted characteristic variable in consideration of the characteristics of the analysis target data extracted by the machine learning unit 13 . Variable properties and variable value distribution may be statistically calculated and may be provided in the form of key-valuew. When the variable according to the analysis target data is a numeric variable, the machine learning unit 13 may calculate a statistic according to a characteristic. In addition, when the variable according to the analysis target data is a categorical variable, the machine learning unit 13 may calculate the count number and count ratio of the characteristic variable for each category of the variable. The visualization unit 14 may provide a variable attribute, a variable value distribution, and a variable description of the analysis target data to a user terminal (not shown). Also, the visualization unit 14 may provide statistics, the number of counts for each category, and a count ratio to a user terminal (not shown). The user terminal (not shown) may output the data provided by the visualization unit 14 .

According to an embodiment of the present application, understanding of variables can be improved by calculating the statistics of structured data, that is, numerical variables, and counts and ratios for each category of unstructured data, and by utilizing this for calculating characteristic variables, a machine learning algorithm A more reliable feature variable can be calculated as input data of . In addition, the user can easily grasp the characteristics of the data to be analyzed in the process of calculating the characteristic variable.

8 is a diagram schematically illustrating a screen providing an API of a data analysis apparatus according to an embodiment of the present application. Referring to FIG. 8 , the data analysis device 10 may indicate the result of calling the API in consideration of scalability for each analysis result of the bidding analysis step, the engineering analysis step, the construction & commissioning analysis step, and the O&M analysis step. .

9 is a view schematically showing a data search providing screen and a search result of the data analysis apparatus according to an embodiment of the present application, and FIG. 10 is a statistical visualization result for a similar project of the data analysis apparatus according to an embodiment of the present application It is a schematic drawing.

According to an embodiment of the present disclosure, the data input unit 12 may receive user input information for selecting a plurality of search items related to a project to be searched among a plurality of project data collected by the data collection unit 11 . . In addition, the machine learning unit 13 may perform a similarity analysis between the user input information and a plurality of project data, and provide at least one project list based on the similarity analysis result.

For example, referring to FIG. 9 , the visualization unit 14 may provide a plurality of search items for performing a project search to a user terminal (not shown). The data input unit 12 may receive user input information for selecting at least one of a plurality of search items from a user terminal (not shown). The plurality of search items may include a plant type, a project code, a project name, a project type, a site location, a business field, a size (capacity), a project period, an ordering party, and the like. The machine learning unit 13 may perform similarity analysis with project data stored in a database (not shown) based on user input information. As an example, the similarity analysis may be performed based on a fuzzy matching algorithm. The fuzzy matching algorithm may be an algorithm for calculating the similarity between data using a result calculated based on the editing distance (Levenshtein Distance). The machine learning unit 13 may provide a similar project list based on the similarity analysis result. Also, the data input unit 12 may receive user input information for selecting at least one of the provided similar project list based on the similarity analysis result. The visualization unit 14 may provide a similar project based on user input information.

Also, referring to FIG. 10 , the visualization unit 14 may provide statistical visualization results for similar projects. The visualization unit 14 may derive the project delay date, MM, and the like with respect to the search result using a statistical technique, and visualize it and provide it.

Hereinafter, an operation flow of the present application will be briefly reviewed based on the details described above.

11 is an operation flowchart of a data analysis method using engineering full cycle data according to an embodiment of the present application.

The data analysis method using the engineering whole cycle data shown in FIG. 11 may be performed by the data analysis apparatus 10 described above. Therefore, even if omitted below, the description of the data analysis apparatus 10 may be equally applied to the description of the data analysis method utilizing the data of the entire engineering cycle.

In step S101, the data analysis device 10 may collect project data generated in the entire cycle of the engineering industry.

In step S102, the data analysis device 10 may receive new project data.

In step S103, the data analysis apparatus 10 may perform data analysis by applying the received new project data to a learning module learned based on machine learning.

In step S104 , the data analysis apparatus 10 may provide a visualization result of data analysis.

In the above description, steps S101 to S104 may be further divided into additional steps or combined into fewer steps, according to an embodiment of the present application. In addition, some steps may be omitted as necessary, and the order between steps may be changed.

The data analysis method using the engineering whole cycle data according to an embodiment of the present application may be implemented in the form of a program command that can be executed through various computer means and recorded in a computer readable medium. The computer-readable medium may include program instructions, data files, data structures, etc. alone or in combination. The program instructions recorded on the medium may be specially designed and configured for the present invention, or may be known and available to those skilled in the art of computer software. Examples of the computer readable recording medium include magnetic media such as hard disks, floppy disks and magnetic tapes, optical media such as CD-ROMs and DVDs, and magnetic media such as floppy disks. - includes magneto-optical media, and hardware devices specially configured to store and execute program instructions, such as ROM, RAM, flash memory, and the like. Examples of program instructions include not only machine language codes such as those generated by a compiler, but also high-level language codes that can be executed by a computer using an interpreter or the like. The hardware devices described above may be configured to operate as one or more software modules to perform the operations of the present invention, and vice versa.

In addition, the data analysis method utilizing the above-described engineering full cycle data may be implemented in the form of a computer program or application executed by a computer stored in a recording medium.

The above description of the present application is for illustration, and those of ordinary skill in the art to which the present application pertains will understand that it can be easily modified into other specific forms without changing the technical spirit or essential features of the present application. Therefore, it should be understood that the embodiments described above are illustrative in all respects and not restrictive. For example, each component described as a single type may be implemented in a dispersed form, and likewise components described as distributed may be implemented in a combined form.

The scope of the present application is indicated by the following claims rather than the above detailed description, and all changes or modifications derived from the meaning and scope of the claims and their equivalents should be construed as being included in the scope of the present application.

10: data analysis device
11: data collection unit
12: data input
13: Machine Learning Department
14: visualization unit

Claims (12)

In the data analysis device using the engineering whole cycle data,
a data collection unit that collects project data occurring in the entire cycle of the engineering industry;
a data input unit for receiving new project data;
a machine learning unit for performing data analysis by applying the new project data received from the data input unit to a learning module learned based on machine learning; and
a visualization unit that visualizes and provides the analysis result of the machine learning unit;
including,
The machine learning unit,
Toxin clause analysis is performed by applying the new project data to the first learning module learned based on machine learning to detect the toxin clause included in the ITB document data for the ITB analysis item among the plurality of analysis step items of the engineering industry However, the toxin clause analysis uses the toxin clause dictionary to separate the morphemes of the terms included in the input ITB document to perform standardization, and includes a plurality of words included in the separated term in the toxin clause dictionary stored in advance Comparison is performed with the words of the toxin clause, but the comparison is performed by an artificial intelligence-based algorithm Fuzzy Data Matching algorithm,
Predictive analysis of MH (Man Hour) is performed by applying the new project data to the second learning module learned based on machine learning for the design cost prediction analysis item among the plurality of analysis step items of the engineering industry,
Performing project delay date analysis by applying the new project data to the third learning module learned based on machine learning for the design error analysis item among the plurality of analysis step items of the engineering industry,
A change amount analysis is performed by applying the new project data to a fourth learning module learned based on machine learning for a design change analysis item among a plurality of analysis step items of the engineering industry,
A maintenance item detection analysis is performed by applying the new project data to a fifth learning module learned based on machine learning for a predictive maintenance analysis item among a plurality of analysis step items of the engineering industry,
Reset the value of the hyper parameter to increase the performance of the first to fifth learning modules,
Randomly extract the value of the hyperparameter from a set range, perform module learning using the sampled hyperparameter value, evaluate the accuracy with validation data, which is data for evaluating the adequacy of the hyperparameter,
The process of randomly extracting the value of the hyper parameter, learning the module, and repeating the process of evaluating the accuracy with the verification data a preset number of times, determining the accuracy and resetting the range of the hyper parameter, The reset is reset in the direction of narrowing the range of the hyperparameter,
For each of the plurality of analysis step items of the engineering industry, the learning results performed by applying to a plurality of learning models are collected, and the learning results of the learning model representing the highest accuracy learning results are applied to the plurality of analysis step items of the engineering industry. provided as an analysis result for
When the plurality of learning models are of the regression algorithm type, the evaluation index of each of the plurality of machine learning algorithms corresponding to the regression algorithm type is calculated, and the ranking is calculated in the order of the lowest root mean square error, and a plurality of analysis step items of the engineering industry A data analysis device that is to be provided as an analysis result for the. According to claim 1,
The data input unit,
Of the plurality of analysis step items of the engineering industry, which receives user input information for selecting any one analysis step item to be analyzed, the data analysis device. 3. The method of claim 2,
A plurality of analysis step items of the engineering industry,
Bidding analysis items, Engineering analysis items, Construction & Commissioning analysis items, O & M analysis items, the data analysis device comprising at least one of the items. 4. The method of claim 3,
The learning module learned based on the machine learning,
A data analysis device comprising at least one learning module of a regression model, a classification model, a cluster model, and a deep learning model. delete delete delete delete delete 5. The method of claim 4,
The machine learning unit,
For each of the plurality of analysis step items of the engineering industry, the learning results performed by applying to a plurality of learning models are collected, and the learning results of the learning model representing the highest accuracy learning results are applied to the plurality of analysis step items of the engineering industry. A data analysis device that is to be provided as an analysis result for. 5. The method of claim 4,
The data input unit,
Receive user input information for selecting a plurality of search items related to a project to be searched among a plurality of project data collected by the data collection unit,
The machine learning unit,
A data analysis apparatus that performs a similarity analysis between the user input information and the plurality of project data, and provides at least one project list based on a similarity analysis result. In a data analysis method using data of the entire engineering cycle in which each step is performed by a data analysis device implemented by a computer,
collecting project data occurring in the entire cycle of the engineering industry;
receiving new project data;
performing data analysis by applying the received new project data to a learning module learned based on machine learning; and
Visualizing and providing analysis results;
including,
The step of performing the data analysis is,
Toxin clause analysis is performed by applying the new project data to the first learning module learned based on machine learning to detect the toxin clause included in the ITB document data for the ITB analysis item among the plurality of analysis step items of the engineering industry However, the toxin clause analysis uses the toxin clause dictionary to separate the morphemes of the terms included in the input ITB document to perform standardization, and includes a plurality of words included in the separated term in the toxin clause dictionary stored in advance Comparison is performed with the words of the toxin clause, but the comparison is performed by an artificial intelligence-based algorithm Fuzzy Data Matching algorithm,
Predictive analysis of MH (Man Hour) is performed by applying the new project data to the second learning module learned based on machine learning for the design cost prediction analysis item among the plurality of analysis step items of the engineering industry,
Performing project delay date analysis by applying the new project data to the third learning module learned based on machine learning for the design error analysis item among the plurality of analysis step items of the engineering industry,
A change amount analysis is performed by applying the new project data to a fourth learning module learned based on machine learning for a design change analysis item among a plurality of analysis step items of the engineering industry,
A maintenance item detection analysis is performed by applying the new project data to a fifth learning module learned based on machine learning for a predictive maintenance analysis item among a plurality of analysis step items of the engineering industry,
Reset the value of the hyper parameter to increase the performance of the first to fifth learning modules,
Randomly extract the value of the hyperparameter from a set range, perform module learning using the sampled hyperparameter value, evaluate the accuracy with validation data, which is data for evaluating the adequacy of the hyperparameter,
The process of randomly extracting the value of the hyper parameter, learning the module, and repeating the process of evaluating the accuracy with the verification data a preset number of times, determining the accuracy and resetting the range of the hyper parameter, resetting is reset in the direction of narrowing the range of the hyperparameter,
For each of the plurality of analysis step items of the engineering industry, the learning results performed by applying to a plurality of learning models are collected, and the learning results of the learning model representing the highest accuracy learning results are applied to the plurality of analysis step items of the engineering industry. provided as an analysis result for
When the plurality of learning models are of the regression algorithm type, the evaluation index of each of the plurality of machine learning algorithms corresponding to the regression algorithm type is calculated, and the ranking is calculated in the order of the lowest root mean square error, and a plurality of analysis step items of the engineering industry To provide as an analysis result for the, data analysis method.

Images