KR20190021560A - Failure prediction system using big data and failure prediction method - Google Patents

Abstract

The present invention relates to a failure prediction system using Big data and a failure prediction method, and more particularly, to a failure prediction system using big data and a failure prediction method, in which a sensor is attached to the system and an environment for exchanging various data generated by the system over the Internet is established, so that it is possible to predict the deformation and determine the failure based on the collected Big data. The system includes: a first system for extracting data; a sensing sensor attached to the first system for monitoring data in real time from the first system; a middleware for allowing the detection sensor to connect to devices of different models to enable smooth communication; a manufacturing execution system (MES) that receives data from the middleware and guides all the work activities of the first system using reliable data; a deletion system interlocked with the manufacturing execution system (MES) to delete unnecessary data; and a failure prediction and maintenance module for analyzing data that has been deleted and remained through the deletion system.

Description

[0001] The present invention relates to a failure prediction system and a failure prediction method using big data,

The present invention relates to a fault prediction and maintenance system and a fault prediction method using big data. More particularly, the present invention relates to a system and method for installing a sensor in a system, The present invention relates to a failure prediction system and a failure prediction method using big data capable of predicting a failure and determining a failure based on collected big data.

The failure prediction system and the failure prediction method using the big data described in the present invention are for predicting a failure and determining a failure using big data. Big data means data generated in a digital environment Large-scale data that includes characters, image data, as well as numerical data in a vast, short generation period, and shape. There is a big data environment around us that produces a lot of information and data. Many of the current information and data are too large in comparison to the past analogue environment, and it is difficult data to collect, store, search, analyze and visualize by the previous methods and tools. This big data environment has a lot of data compared to the past, and the kinds of data are diversified, so that people can analyze and predict opinions and opinions through SNS.

As described above, one of the reasons that the data information is increasing exponentially is the proliferation of reason and intelligent communication. Video content including user-created UCC, characters generated by mobile phone and SNS are not only increasing in data rate but also in form and quality. Twitter only generates an average of 155 million per day, and YouTube's average video playback per day is 4 billion. CCTVs installed on major roads, public buildings, and even apartment elevators are huge enough to exceed the imagination of the amount of video information being shot. Indeed, every single act of everyday life is stored as data.

Therefore, the topic in the field of information communication today is the big data. For example, in 2011, the US Internal Revenue Service (U.S.) combined massive amounts of data and IT technology to create an integrated tax evasion and fraud prevention system. We found anomalies in the anti-fraud solution, and we used predictive modeling to analyze taxpayer's past behavioral information and then to detect behaviors similar to fraud patterns. After that, I analyzed the account, address, telephone number, and relationship between taxpayers related to criminals through Facebook or Twitter and found high tax delinquents. In the process, the US Department of Internal Revenue used low-cost data analysis, such as Hadoop, an open-source, high-volume data processing software, to prevent tax loss of $ 345 billion annually. The above is just one example of the use of Big Data. Now, when you handle and analyze the Big Data properly, you can do things you never thought possible before.

In addition, the equipment fault diagnosis and maintenance technology in the precision machining field utilizing such a big data analysis technology can detect abnormality from various production and facility data sources generated in production activities and collect data in case of failure.

In relation to this, there is a failure type management apparatus and method for the maintenance of the marine resource production equipment, which is Korean Patent No. 10-1518720 (2015.05.01).

However, the conventional technology has a limitation in collecting and analyzing a large amount of training data due to a lack of a real-time monitoring system.

In addition, there is a disadvantage in that the autonomic countermeasure of occurrence faults is slow by predicting a fault element, and the soundness factor and the fault reference value can not be set accurately.

Also, it was a difficult system to get the desired information among big data.

1. Korean Patent No. 10-1518720 (2015.05.01)

SUMMARY OF THE INVENTION Accordingly, the present invention has been made keeping in mind the above problems occurring in the prior art,

The purpose of this study is to overcome the limit of collection of large amount of training data by developing the system by combining big data analysis technique with object internet.

In addition, it extends the remaining useful life of mechanical equipment and production system, develops real-time status check and failure prediction and maintenance system, and improves production efficiency and stabilization by autonomous countermeasure .

In this paper, we propose a data - based method for predicting failures, and propose a method for predicting future failures after mechanically learning the relationship between load and damage using machine learning techniques.

In addition, we propose a model-based method for predicting faults, and propose a method for diagnosing faults based on physical fault models.

The problems to be solved by the present invention are not limited to the above-mentioned problems, and other problems to be solved by the present invention, which are not mentioned here, As will be appreciated by those skilled in the art.

According to a preferred embodiment of the present invention, there is provided a data processing system including a first system for extracting data, a detection sensor attached to the first system for monitoring data from the first system in real time, (MES) for receiving data from the middleware and guiding all the work activities of the first system by using reliable data, middleware for connecting to the equipment and allowing smooth communication, And a failure prediction and maintenance module for analyzing data left and deleted through the deletion system and notifying the failure of the object in advance. Also, a method in which the object Internet and the big data are combined includes a data acquisition step of collecting data into a manufacturing execution system (MES) by detecting an abnormal symptom of the system, and the data obtained in the data acquisition step can be visually confirmed on a monitor A data diagnosis step of determining whether the data is defective after confirming the data in the data acquisition step, a data failure prediction step of predicting future defect progress based on the data analyzed in the data diagnosis step, Presenting a solution to the problem of the future defects predicted in the data failure prediction step and confirming the problem solution through the data presentation and action steps providing the key information for the corrective action and the information confirmed in the data presentation and action steps And post-data analysis steps to diagnose post-conditions It is open configuration. Here, the deletion system can analyze a failure mode of the first system and elements affecting the failure or the like through a data mining technique for obtaining desired information in the big data, And the fault reference value can be accurately set. The data diagnosis step further includes a removal step of discovering a hidden phenomenon while systematically storing and managing data, and obtaining valid data by mining minuses.

According to the present invention, the failure prediction system and the failure prediction system using the big data can effectively collect system operation data in real time and can monitor the system in real time.

In addition, it can be applied not only to large-scale systems such as power plants and large-scale plants, but also to small-sized systems.

In addition, the state information of objects and a lot of information acquired from the sensors are transmitted through the Internet, and they are stored in the cloud to form big data.

In addition, through data mining technique that obtains desired information among big data through analyzing huge amount of information collected through Internet, user's usage pattern, and failure information, And so on.

In addition, the system's health factor and fault reference value can be set accurately, which enables efficient system diagnosis and fault prediction.

In addition, it is possible to fine-tune the system's health degradation model, which is optimized in terms of soundness prediction.

FIG. 1 is a view schematically showing a configuration of a failure prediction system according to an embodiment of the present invention. Referring to FIG.
FIG. 2 is a diagram showing a failure prediction preservation method according to an embodiment of the present invention in each step.
FIG. 3 is a diagram illustrating a process of extracting a failure related message according to an embodiment of the present invention. Referring to FIG.
4 is a diagram illustrating a method for extracting useful data of the present invention according to an embodiment of the present invention.
FIG. 5 is a diagram showing extracted data that can be solved by the present invention according to an embodiment of the present invention.
FIG. 6 is a diagram illustrating an operation flow of the present invention in accordance with an embodiment of the present invention. Referring to FIG.

The above and other objects, features and advantages of the present invention will be more apparent from the following detailed description taken in conjunction with the accompanying drawings, in which: FIG. BRIEF DESCRIPTION OF THE DRAWINGS The advantages and features of the present invention, and the manner of achieving them, will be apparent from and elucidated with reference to the embodiments described hereinafter in conjunction with the accompanying drawings.

The present invention will be described in more detail with reference to the accompanying drawings.

In one embodiment, FIG. 1 is a diagram schematically illustrating a configuration of a failure prediction maintenance system according to an embodiment of the present invention.

A failure prediction system and a failure prediction system using Big Data according to the present invention are a first system for extracting data, a detection sensor for attaching to the first system and capable of monitoring data from the first system in real time, A middleware for allowing the detection sensor to connect to devices of different models to allow smooth communication, a manufacturing execution system (MES) for receiving data from the middleware and guiding all the work activities of the first system using reliable data, ), A deletion system interlocked with the manufacturing execution system (MES), for deleting unnecessary data, and a failure prediction and maintenance module for analyzing data left and deleted through the deletion system and notifying the failure of an object in advance .

First, the present invention provides a first system, such as a smart factory, in which a sensor capable of detecting various abnormalities such as temperature, vibration, signal, and acceleration is installed in a first system, Is a system that is provided to detect the information of machines and equipment in the network.

More specifically, virtualization of the information collected from the sensing sensor 110 of a process instrument enables a variety of sensor-based application services, or a technique capable of exploiting many possibilities in advance, Software is used to direct difficult situations and anticipate possible problems to realize innovative cost reduction.

Here, the environment of the middleware 120, which is software for connecting the devices of different models to the sensing sensor 110 to allow smooth communication, is needed.

Next, the manufacturing execution system (MES) 130, which guides all the work activities of the first system 100 in the middleware 120 that collected various data from the first system 100 such as production and facilities, .

This enables a linkage module with the manufacturing execution system (MES) 130 to be developed and traceable to the work instruction number of the legacy system and the LTO-NO, and can be linked to the basic information and the preventive maintenance information of the system .

In addition, the failure prediction system and the failure prediction system using the Big Data according to the present invention analyze the remaining data deleted through the deletion system 140 for deleting unnecessary data, A failure prediction and maintenance module 150 is provided.

FIG. 2 is a view showing each step of the failure prediction preservation method according to an embodiment of the present invention.

The present invention relates to a method of collecting data from a manufacturing execution system (MES) 130 by detecting an abnormality of a system and combining the object Internet and big data, A data diagnosis step (S300) for checking whether the data is defective or not after checking the data in the data acquisition step (S100), a data diagnosis step (S300) for checking whether the data is defective or not, Based on the data analyzed in step S300, a data fault prediction step S400 for predicting the future defect progress and a problem solution for the future defect predicted in the data fault prediction step S400 are presented.

The data acquisition step S100 may be applied to almost all steps.

In addition, the step of diagnosing the cause of the defect and the defect and classifying the severity of the data acquired in the data acquisition step (S100) is the data diagnosis step (S300).

Here, in the data diagnosis step (S300), there is a process of analyzing the data state of the big data or the machine learning based on whether it is normal data, abnormal data, valid data required by the user, or unnecessary data.

In addition, a data post-analysis for checking the problem solution and diagnosing the post-condition through the data presentation and action step (S500) providing the key information for corrective action and the information confirmed in the data presentation and action step (S500) The method comprising the steps of:

The step of detecting an abnormal symptom from the system and acquiring data is a data acquisition step (S100). In the data acquisition step (S100), an abnormal symptom can be detected, and noise such as vibration and ultrasonic waves can be detected. In the whole process, a step that can be visually confirmed through real-time monitoring is the data acquisition step S100.

Referring to FIG. 6, the data mining technique, the integrity factor setting, the failure threshold setting, and the legacy removal are used in the process of discrimination.

Next, the data prediction step for predicting future faults and defects.

In the data prediction step, probability evaluation and reliability analysis are performed for future defects, progress and prediction.

More specifically, in the data prediction step, a big data or a machine learning based data defect state is predicted.

Next is the data presentation and action phase, which is the step of suggesting a solution to the problem of defects and faults and taking measures. In the data presentation and action step (S500), key information for corrective action is provided.

More specifically, in the data presentation and action step (S500), it is possible to compare the pattern of the message relating to the fault with the existing pattern to judge whether or not the fault is present.

In addition, a message including a negative meaning such as a message informing that the execution of a program, an application, etc. may be delayed or can not be executed, such as a status message related to a failure, is to be mentioned. .

In addition, you can identify the equipment and parts, the types of failures and defects, and take the best recommended action.

Finally, it is the post-data analysis phase, which identifies problem solving and diagnoses the state of the future. In the data post-analysis step, the post-condition, that is, the future state can be predicted and analyzed based on the core information acquired in the data presentation and action step.

FIG. 3 is a diagram illustrating a process of extracting a failure related message according to an embodiment of the present invention. Referring to FIG.

Referring to FIG. 3, the system of the present invention will be described in more detail. The failure prediction system and the failure prediction system using Big Data according to the present invention employ Big Data or Machine Learning Analyzer, , And collects a large amount of data.

Also, production data such as product information and quality are acquired from the existing legacy system together with the data-based method.

In the case of fault diagnosis through preventive maintenance technology, it is detected when sufficient degradation has already occurred and a sufficient state change has occurred, thereby complementing the limit of effectiveness of preventive maintenance technology.

It thus extends the residual useful life of mechanical equipment and production systems and compensates for the lack of existing real-time status checks and fault diagnosis systems.

Here, there are data-based methods and model-based methods for failure prediction and health management.

The data-based method is a method for predicting a future failure by mechanically learning the relationship between load-to-force damage using a machine learning technique, and can be applied to a multivariate system that is difficult to implement. But it takes a lot of data to collect the data.

The model-based method diagnoses and predicts faults based on physical fault models. Although the physical model varies depending on the application target, there is an advantage that a more accurate prediction can be made on the behavior of the damage over a longer period of time.

Fault diagnosis can be performed with a small amount of fault data, and it is possible to apply it to various operating environments by changing the parameters of the model.

FIG. 4 is a diagram illustrating a method for extracting useful data according to an embodiment of the present invention. FIG. 5 is a flowchart illustrating a method of extracting data that can be solved by the present invention according to an embodiment of the present invention And FIG. 6 is a view showing the entire operation flow of the present invention according to an embodiment of the present invention.

Referring to FIGS. 4 to 6, the advantages of the Internet and the Big Data Analysis technology can be obtained by combining the Internet of Things and the Big Data Analysis technology.

Advantages of applying the Internet technology include that real-time data of the system operation can be collected easily and the system can be monitored in real time.

Further, the advantages of the present invention are applicable not only to large systems such as power plants and large plants but also to small systems

In addition, through the Internet application of objects, various devices and sensors can be freed from communication restriction, and real-time status monitoring of almost all systems is possible.

In addition, much information such as the status information of the object, sensor data acquired from the sensor, usage information of the user, and the like are transmitted through the Internet and stored in the cloud to form big data.

The advantage of applying the Big Data Analysis technology is that it analyzes a huge amount of object operation information, failure information, usage patterns of users, and the like, Through the mining technique, the failure mode of the system and the factors affecting the failure are analyzed.

In addition, efficient system diagnosis and prediction that can accurately set the health factor and the failure reference value indicating the health state of the system are possible.

In addition, since the system's health degradation model can be finely set, it can be said that it is optimized for the health prediction.

Accordingly, the fault prediction and maintenance method using the Big Data according to the present invention supports a customized interface, applies a real-time big data analysis algorithm, presents a real-time data processing screen, and variously visualizes analysis data.

It also provides user-defined tools and is easy to integrate with manufacturing equipment.

It supports real-time analysis and batch analysis through real-time notification service, and links Big Data System and MES system.

Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings, but the present invention is not limited by the scope of the present invention.

While the present invention has been described with reference to exemplary embodiments, it is to be understood that the invention is not limited to the disclosed exemplary embodiments.

100: First system
110: detection sensor
120: Middleware
130: Manufacturing Execution System (MES)
140: Delete system
150: Fault prevention module
S100: Data acquisition step
S200: Data acquisition step
S300: Data diagnosis step
S400: Data failure prediction step
S500: Data presentation and action steps
S600: Data post-analysis phase

Claims (5)

A first system for extracting data;
A detection sensor attached to the first system and capable of monitoring data from the second system in real time;
A middleware for allowing the detection sensor to connect to devices of different models to enable smooth communication;
A manufacturing execution system (MES) that receives data from the middleware and guides all the work activities of the first system using reliable data;
A deletion system interfaced with the manufacturing execution system (MES), for deleting unnecessary data; And
And a failure prediction and maintenance module for analyzing data left over through the deletion system and informing a failure of an object in advance. The method according to claim 1,
The deletion system comprises:
Through a data mining technique that obtains the desired information in big data,
The failure mode of the system and the factors affecting the failure can be analyzed. 3. The method of claim 2,
The deletion system comprises:
The system is able to accurately set the health factor and the fault reference value. In the method of combining the object Internet and the big data,
A data acquisition step of detecting abnormality signs of the first system and collecting data into a manufacturing execution system (MES);
A real-time monitoring step of visually observing the data obtained in the data acquisition step on a monitor in real time;
A data diagnosis step of determining whether the data is defective after confirming the data in the data acquisition step;
A data failure prediction step of predicting future defect progress based on the data analyzed in the data diagnosis step;
A data presenting and correcting step of presenting a problem solution to the future defect predicted in the data failure prediction step and providing key information for corrective action; And
And a data post-analysis step of confirming a problem solution and diagnosing a post-condition through information confirmed at the data presentation and action step. 5. The method of claim 4,
The data diagnosis step may include:
Further comprising a removal step of systematically storing and managing data to discover a hidden phenomenon and obtaining valid data by mining minus.

Images