KR101825809B1 - Plant system, and fault detecting method thereof - Google Patents

Abstract

Disclosed is a plant anomaly detection system and method for diagnosing faults using a diagnosed database and diagnostics logic built with predicted faults in plant systems, and tracking and visualizing root causes.
The disclosed plant anomaly detection system includes: a diagnostic database that stores failures in a plant as diagnostic data; A data collection unit for collecting plant data in real time through each sensor at a predetermined cycle with respect to the plant; A state learning unit for learning a steady state using the collected real time plant data; An alarm unit for comparing the real time plant data with the learned steady state data to alarm the failure early; A modeling unit for modeling the learned steady state data to generate a plant model; A fault diagnosis unit for diagnosing a plant fault using the diagnostic logic based on the diagnosis database and for tracking the root cause; And an output unit for visualizing and outputting diagnosis results of the fault diagnosis unit.
According to the present invention, the failure of the plant can be alertered early, and the root cause of the failure can be diagnosed and visualized.

Description

[0001] The present invention relates to a plant anomaly detection system,

The present invention relates to a plant system and a method for detecting abnormalities. More specifically, the present invention relates to a plant system that acquires plant data in real time, learns a steady state using secured data, and compares real- The present invention relates to a plant system and an abnormality detection method thereof, which can diagnose faults by using predicted faults and built diagnosis databases and diagnostic logic, and can trace and visualize root causes.

Generally, large plants such as power plants or chemical plants are operated in complex operation with hundreds of machines and electrical equipments of various kinds and operated in a central control room called main control room. The number of people working here is 2-10, which is decreasing due to the recent improvement in corporate competitiveness and productivity. In particular, the number of facilities to be managed and controlled per worker has increased dramatically in the case of thermal power plants where dozens of workers have been working in the past. Accordingly, the operation and operation method has been developed from the operation method of the local panel to the operation method of the main control panel in the past. Recently, the operation method using the computer-based MMI (Man-Machine Interface) Respectively.

However, in the case of skilled workers with decades of work experience, they are familiar with various situations and understand the complex internal control logic, but most of the inexperienced workers are often referred to as related reference books (Operating procedures, supplier design data, internal books, etc.). This type of operating environment is extremely vulnerable to the operation of plants that require quick and stable action.

In addition, it also displays basic operation information in a driving operation console (console) for displaying driving information, and does not sufficiently display information necessary for actual operation in various situations.

Therefore, there is a need for a system that can provide appropriate information related to operation outside the operation console of the main control panel providing only basic information, and provide appropriate information according to each situation.

In other words, there is a need for a technique to warn a dangerous state when a plant-related operation-related variable deviates from a normal operation state, and to take prompt action.

Korean Registered Patent No. 1065767 (registered on September 09, 2011)

An object of the present invention to solve the above-mentioned problems is to obtain plant data in real time, to automatically learn steady state by using secured data, to compare steady state obtained by real- The present invention provides a plant system and an abnormality detection method that can diagnose faults using the predicted faults and the constructed diagnosis database and diagnostic logic, and can trace and visualize the root cause.

According to an aspect of the present invention, there is provided a plant anomaly detection system including: a diagnostic database that stores failure cases in a plant as diagnostic data; A data collection unit for collecting plant data in real time through each sensor at a predetermined cycle with respect to the plant; A state learning unit for learning a steady state using the collected real time plant data; An alarm unit for comparing the real time plant data with the learned steady state data to alarm the failure early; A modeling unit for modeling the learned steady state data to generate a plant model; A fault diagnosis unit for diagnosing a plant fault using the diagnostic logic based on the diagnosis database and for tracking the root cause; And an output unit for visualizing and outputting diagnosis results of the fault diagnosis unit.

In addition, the state learning unit deletes the abnormal data through a pre-processing process on the collected real-time plant data, and learns the normal state based on the preprocessed plant data.

In addition, the alarm unit generates, as the actual value, the value obtained by subtracting the predicted value from the actual value, using the learned steady-state data as the actual value, and outputs the alarm logic Logic will output an early warning.

Also, the modeling unit generates a plant model by modeling the learned steady state data into a parametric model or a non-parametric model.

The state learning unit generates the learned steady-state data as a learning data set, extracts the characteristics by combining the steady-state plant data in which the abnormal data is deleted through the preprocessing process in the real-time plant data, And generates a new pattern according to new learning data to update the learning data. The modeling unit receives the updated updated learning data from the state learning unit, applies the new updated learning data to the plant model, and updates the plant model.

According to another aspect of the present invention, there is provided a method for detecting a plant abnormality, comprising the steps of: (a) collecting plant data in real time through each sensor at a predetermined cycle with respect to the plant; (b) learning the steady state using the collected real time plant data by the state learning unit; (c) comparing the learned real time plant data and the learned steady state data in the alarm unit to alarm the fault early; (d) generating a plant model by modeling the steady state data on which the modeling unit has been learned; (e) diagnosing the failure of the plant using the diagnostic logic based on the diagnostic database in the fault diagnosis section and tracking the root cause; And (f) visualizing and outputting a diagnosis result of the plant at the output section.

In step (b), the state learning unit deletes the abnormal data through a pre-processing process on the collected real-time plant data, and learns the normal state based on the preprocessed plant data.

Also, in the step (c), the alarm unit generates, as the residual, a value obtained by subtracting the predicted value from the actual value, using the learned steady-state data as the actual value, and the real time plant data as the actual value. If the generated residual value exceeds the allowable value The alarm signal is output by the alarm logic.

In step (d), the modeling unit generates a plant model by modeling the learned steady state data into a parametric model or a non-parametric model.

In the step (b), the state learning unit generates the learned steady state data as a learning data set, combines the steady-state plant data in which the abnormal data is deleted through the preprocessing process in the real-time plant data, In the step (d), the modeling unit receives the new updated learning data from the state learning unit and applies the new updated learning data to the plant model The plant model will be updated.

According to the present invention, sensor data of various periods generated in a plant can be obtained in real time, and classified into units for creating a model for anomalous symptom prediction, and an automatic learning function can be performed based on steady state data.

In addition, system separation between preprocessing, modeling, residual generation, and diagnosis functions minimizes the effects between functions, ensuring ease of system development and maintenance, and reusing it as an asset.

In addition, we can capitalize diagnostic know-how using diagnostic logic and database, and provide diagnostic guide system.

In addition, it is possible to provide an infrastructure system for capturing plant anomalies, to provide a diagnostic system for anomalous indications regardless of the operator's expertise, and to maximize operational efficiency by reducing the worker's time and effort by hand have.

1 is a block diagram schematically showing functional blocks of a plant anomaly detection system according to an embodiment of the present invention.
FIG. 2 is a flowchart illustrating a method for detecting a plant anomaly according to an embodiment of the present invention. Referring to FIG.
3 is a diagram illustrating an overall process of collecting and automatically diagnosing and collecting plant data according to a plant anomaly detection method according to an embodiment of the present invention.
FIG. 4 is a diagram illustrating an ensemble learning process based on the model of the model and the model of non-parametricity according to the present invention.
FIG. 5 is a diagram illustrating an example of performing ensemble learning by combining an aqueous model and a non-parametric model according to an embodiment of the present invention.
FIG. 6 is a diagram showing a result of calculating a highly accurate prediction value according to a combination of an aqueous model and a non-parametric model according to an embodiment of the present invention.
FIG. 7 is a diagram illustrating a process of outputting an alarm from sensor data of a plant according to an embodiment of the present invention, and an automatic learning process for detecting a plant anomaly.
8 is a diagram illustrating an example of performing data processing on sensor data that is currently measured according to an embodiment of the present invention.
9 is a diagram illustrating an example of generating new learning data by combining existing sensor data and existing learning data according to an embodiment of the present invention.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings, which will be readily apparent to those skilled in the art to which the present invention pertains. The present invention may be embodied in many different forms and is not limited to the embodiments described herein.

In order to clearly illustrate the present invention, parts not related to the description are omitted, and the same or similar components are denoted by the same reference numerals throughout the specification.

Throughout the specification, when a part is referred to as being "connected" to another part, it includes not only "directly connected" but also "electrically connected" with another part in between . Also, when an element is referred to as "comprising ", it means that it can include other elements as well, without departing from the other elements unless specifically stated otherwise.

If any part is referred to as being "on" another part, it may be directly on the other part or may be accompanied by another part therebetween. In contrast, when a section is referred to as being "directly above" another section, no other section is involved.

The terms first, second and third, etc. are used to describe various portions, components, regions, layers and / or sections, but are not limited thereto. These terms are only used to distinguish any moiety, element, region, layer or section from another moiety, moiety, region, layer or section. Thus, a first portion, component, region, layer or section described below may be referred to as a second portion, component, region, layer or section without departing from the scope of the present invention.

The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the invention. The singular forms as used herein include plural forms as long as the phrases do not expressly express the opposite meaning thereto. Means that a particular feature, region, integer, step, operation, element and / or component is specified and that the presence or absence of other features, regions, integers, steps, operations, elements, and / It does not exclude addition.

Terms indicating relative space such as "below "," above ", and the like may be used to more easily describe the relationship to other portions of a portion shown in the figures. These terms are intended to include other meanings or acts of the apparatus in use, as well as intended meanings in the drawings. For example, when inverting a device in the figures, certain portions that are described as being "below" other portions are described as being "above " other portions. Thus, an exemplary term "below" includes both up and down directions. The device can be rotated by 90 degrees or rotated at different angles, and terms indicating relative space are interpreted accordingly.

Unless otherwise defined, all terms including technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. Commonly used predefined terms are further interpreted as having a meaning consistent with the relevant technical literature and the present disclosure, and are not to be construed as ideal or very formal meanings unless defined otherwise.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings so that those skilled in the art can easily carry out the present invention. The present invention may, however, be embodied in many different forms and should not be construed as limited to the embodiments set forth herein.

1 is a block diagram schematically showing functional blocks of a plant anomaly detection system according to an embodiment of the present invention.

1, a plant anomaly detection system 100 according to the present invention includes a data collecting unit 110, a state learning unit 120, an alarm unit 130, a modeling unit 140, a fault diagnosis unit 150, An output unit 160, and a diagnostic database 170. [

The data collecting unit 110 collects plant data in real time through each sensor at a predetermined cycle with respect to the plant. That is, the data collecting unit 110 collects plant data in real time from each measurement sensor of the plant, for example, every five minutes.

The state learning unit 120 learns the steady state using the collected real-time plant data. At this time, the state learning unit 120 deletes the abnormal data through a pre-processing process on the collected real-time plant data, and learns the normal state based on the preprocessed plant data. Here, the pretreatment process will be described in more detail in the following examples.

The alarm unit 130 compares the real-time plant data with the learned steady-state data to alarm the malfunction prematurely. That is, the alarm unit 130 generates, as the actual value, the value obtained by subtracting the predicted value from the actual value, using the learned steady-state data as the actual value, and when the generated residual value exceeds the allowable value, The alarm is output by the logic (Alarm Logic).

The modeling unit 140 models the learned steady state data to generate a plant model as a predictive model.

In addition, the modeling unit 140 generates the plant model by modeling the learned steady state data into a parametric model or a non-parametric model.

The modeling unit 140 generates steady-state data learned in the state learning unit 120 as a learning data set, and generates a steady-state data in which the abnormal data is deleted through preprocessing in the plant data collected in real- The plant data is combined with the learning data set to extract the characteristics to generate new learning data, a new pattern is generated according to the new learning data to update the learning data, and the newly updated learning data is applied to the plant model once a day The plant model will be updated.

The failure diagnosis unit 150 diagnoses the failure of the plant using the diagnosis logic based on the diagnosis database and performs root cause analysis on the root cause.

The output unit 160 visualizes and outputs the diagnosis result of the failure diagnosis unit.

The diagnostic database 170 stores malfunctioning cases as diagnostic data in the plant. Here, the data of the failed cases are generated by using plant data measured in real time as measured values, using the learned steady state data as predicted values, subtracting the predicted value from the measured values as residuals, The abnormal plant data in which the alarm is generated according to the alarm logic in the event of exceeding the normal plant data and the normal plant data in which the abnormal plant data are removed are separately stored and stored.

FIG. 2 is a flowchart illustrating a method for detecting a plant anomaly according to an embodiment of the present invention. Referring to FIG.

Referring to FIG. 2, the plant anomaly detection system 100 according to the present invention first collects plant data in real time through each sensor at a predetermined cycle with respect to the plant in the data collection unit 110 (S210).

That is, the data collecting unit 110 collects the corresponding sensor data every five minutes through each measurement sensor installed in each region of the plant (Data Collecting). As shown in FIG. 3, a batch scheduler ), Sensor data is sequentially received from each sensor every five minutes and collected in the database as raw data. 3 is a diagram illustrating an overall process of collecting and automatically diagnosing and collecting plant data according to a plant anomaly detection method according to an embodiment of the present invention.

Next, the state learning unit 120 learns a normal state using the collected real-time plant data (S220).

That is, as shown in FIG. 3, the state learning unit 120 deletes the abnormal data through a pre-processing process on the plant data collected in real time, and outputs the normal state based on the preprocessed plant data And automatically learns to generate steady state data.

Here, in the pre-processing step, the value obtained by subtracting the predicted value from the actual value is used as the residual, with the real-time plant data as the measured value, the learned steady-state data as the predicted value, and the generated residual value exceeds the allowable value The abnormal plant data in which an alarm is generated is deleted according to the alarm logic and the plant data of the normal state is generated.

At this time, as shown in FIG. 3, the state learning unit 120 generates real-time plant data collected in real time by the data collecting unit 110 every five minutes, for example, every predetermined period through a batch scheduler, And stores the deleted steady state data in the plant database (DB) with reference data (Reference).

Next, the alarm unit 130 compares the real-time plant data with the learned steady-state data, and alerts the user of the failure early (S230).

That is, the alarm unit 130 generates residual data by using the real-time plant data as actual values, using the learned steady-state data as the predicted values, subtracting the predicted value from the actual data as shown in FIG. 3 (Residual Generation) When the residual value exceeds the allowable value, the alarm signal is outputted by the alarm logic.

At this time, the alarm unit 130 generates an early alarm when the residual value exceeds an allowable value (specific value), but may generate an early alarm for a plant failure even when the measured value exceeds an allowable value (specific value).

Next, the modeling unit 140 models the learned steady-state data to generate a plant model (S240).

At this time, the modeling unit 140 may generate a plant model as a predictive model by modeling the learned steady-state data into a parametric model or a non-parametric model.

That is, as shown in FIG. 4, the modeling unit 140 determines a learning model necessary for learning steady-state data by using a parametric model and a non-parametric (non-parametric) model to predict based on a model having different characteristics, -Parametric) model.

At this time, the modeling unit 140 selects a function corresponding to the merit of the specific single models in the model of the model of humor and the model of the non-parametric model, and selects a plurality of prediction models having different characteristics Modeling can be performed in combination.

That is, as shown in FIG. 4, the modeling unit 140 may be configured to select one of the multiple linear regression model (MLRM), the partial least squares (PLS), the dynamic principal component analysis (DPCA), and the nearest neighbor Modeling can be performed by combining a plurality of prediction models having features.

4, the MLRM model includes an ARX (Auto Regressive eXogeneous) model and an MLM (Maximum Likelihood Method) model. The PLS model includes a GLM (Generalized Linear Model) model and an MLM DPCA model includes ARX model and MLM model, and K-NN model can include non-parametric model and RBF (Radial Basis Function) kernel.

Table 1 below shows the advantages and disadvantages of the model of the model and the model of non-parametricity applied to the present invention.

Model Advantages Disadvantages Mother water
Model ARX Model ㅇ Sophisticated model design possible
ㅇ Very high performance
ㅇ Famous model in engineering field
ㅇ High performance
ㅇ Easy model interpretation
ㅇ Nonlinear modeling possible ㅇ Model considering the result, not the process of system state
ㅇ Difficulty modeling MIMO system
ㅇ Very poor expressiveness
ㅇ Use of video and speech recognition Nonmonopoly
Model Non
Parametric
Model ㅇ Excellent performance when applied with technologies such as k-NN ㅇ DB establishment for highly sophisticated learning data
O Predictive performance is highly dependent on the sophistication of the model CNN Model ㅇ Excellent expressiveness
ㅇ Excellent performance compared to general NN model
ㅇ Excellent performance for image processing
ㅇ A useful model when model structure design is difficult
ㅇ Excellent for modeling MIMO system ㅇ Model interpretation is impossible
O Local Optimization
ㅇ There is no guarantee of making a sophisticated model.

Therefore, the modeling unit 140 can perform modeling by combining a plurality of prediction models having different characteristics through review results as shown in the following Table 2, taking into consideration advantages, disadvantages, and performance constraints of the respective models described above .

Model Performance constraints Review the results NN based
HTM O Optimized by model
ㅇ It is difficult to optimize with 50 setting variables.
No formalized design method
ㅇ Not recognized in academia
ㅇ License problem ㅇ Non-conforming (However, it can be used in case of precise detection of specific fault) DPCA
PLS
MLRM ㅇ Performance change due to DPCA time setting. ㅇ Highly suitable for optimization algorithm ㅇ (common) Multivariate Regression method, so when an abnormality is detected in specific sensor, error is detected in other sensor at that point. ㅇ Highly suitable for performance and performance k-NN
(VBM) ㅇ Because the performance of the model depends on the sophistication of the model, there is a possibility that the accuracy of prediction may be lowered when a signal of a pattern that has not existed comes in. ㅇ Fit in terms of implementation and performance. Especially if you can create sophisticated models, you can make very accurate predictions. NN based
DBN O Optimized by model
ㅇ Local Optimization Difficult to Optimize
No formalized design method
O vulnerable to dynamic data ㅇ Not enough due to lack of performance. SVM based
SVDD
SVR ㅇ Data sampling effect is significant
ㅇ Severe performance change due to kernel and variable setting ㅇ Not enough due to lack of performance.

That is, the modeling unit 140 selects the K-NN model when precise detection is required for a specific failure, and when the optimization is needed for each model due to a severe performance change due to time setting, the DPCA model, PLS model, MLRM model Can be selected.

(k)을 얻을 수 있다.For example, as shown in FIG. 5, the modeling unit 140 may use the MLRM model using the ARX model and the LS (Least Square) method as the modeling model and the k-NN model using the RBF kernel as the non-parametric model Can be combined. FIG. 5 is a diagram illustrating an example of performing ensemble learning by combining an aqueous model and a non-parametric model according to an embodiment of the present invention. Therefore, when the data learning unit 130 receives the data (k), the data learning unit 130 obtains the estimated value mlm (k) as x (1) through the MLRM model selected by the learning model selection unit 120 as shown in FIG. ) / Residual_mlrm (k), acquires an estimated value x (2) (Estimated Value_kNN (k) / Residual_kNN (k)) through the k-NN model, and selects the most optimal one among these estimated values The ensemble learning is performed by a bagging method and the ensemble learning is performed to obtain an estimated value and an optimal estimated value as shown in Equation 1 below:

(k).

Accordingly, the modeling unit 140 may perform ensemble learning on the steady state data through the learning model to output the most accurate prediction data (Optimal_Estimated Value) as shown in FIG. Accordingly, the modeling unit 140 transmits the optimal predicted data (Optimal_Estimated Value) calculated according to the ensemble learning to the alarm logic (alarm logic) so that it can be used for alarm of the abnormality of the plant. FIG. 6 is a diagram showing a result of calculating a highly accurate prediction value according to a combination of an aqueous model and a non-parametric model according to an embodiment of the present invention. 6, the data learning unit 130 outputs the Proposed Method ^* having an accuracy of 98.6% and 95.1% as a result of performing ensemble learning using the MLRM, PLS, DPCA, and k-NN models. 6, the modeling unit 140 uses the MLRM, PLS, DPCA, and k-NN models as well as the ensemble learning using the auto-learning algorithm. As a result, %, And the Proposed Method ^** value with an accuracy of 97.9%.

Also, the modeling unit 140 may store the modeled steady state data in the plant database as model data as shown in FIG. 3, so that the alarm unit 130 can refer to it.

The modeling unit 140 generates steady-state data as a learning data set in the state learning unit 120, combines steady-state plant data in which abnormal data is deleted through a preprocessing process in the real-time plant data, The new training data is generated by extracting features and new patterns are applied according to the new learning data, the updated learning data is received once a day, the new updated learning data is applied to the existing plant model, .

First, the state learning unit 120 extracts a feature by combining the currently measured sensor data with an existing learning data set. At this time, the state learning unit 120 sets the magnitude and the angle of the characteristic of the data, and sets the scale of each data set based on the technologies such as N1 Norm, N2 Norm, and N-Infinity Norm. (Magnitude) and Angle (Angle). For example, in the 1 Dimension Method, the magnitude is 1, in the 2 Dimension Method, in the Magnitude and Angle 1, in the 3 Dimension Method, in the Dimension Method, Magnitude and Angle, and Magnitude and Angle N-1 in the N Dimension Method are set as characteristics of data and calculation is performed.

For example, when the number of plant data is 2,000 and alarm information is input from the alarm unit 130 as the alarm unit 130 outputs an alarm, when the sensor data currently measured and input is 288, The learning unit 120 performs data processing on 2,288 pieces of total data (Xm_PoolOn) obtained by adding 288 pieces of sensor data (indexAddData) currently measured to 2,000 existing plant data (Xm_Pool) . Here, the maximum number of additional data indexAddData is 288, and a value of 288 or less per day is input to the state learning unit 120 by automatic learning. FIG. 7 is a diagram illustrating a process of outputting an alarm from sensor data of a plant according to an embodiment of the present invention, and an automatic learning process for detecting a plant anomaly.

As shown in FIG. 8, the state learning unit 120 calculates the number L2morm after the numberXmPoolOn 0 to 288 is input according to the indexAddData value in the number XmPoolOn when i = 1, and performs sorting. 8 is a diagram illustrating an example of performing data processing on sensor data that is currently measured according to an embodiment of the present invention.

Here, the state learning unit 120 divides the total number of data (2,288) by the number of plant data (2,000) and obtains 1.144 with respect to the data interval (interval) according to the following equation (2).

Accordingly, the state learning unit 120 can obtain 1.14, 2.28, 6.864, and 8.008 by sorting the entire data of 2,288 magnitudes at intervals of 1.144, but discarding the decimal points to 1, 2, 6, 8 Is obtained. At this time, the seventh is missing.

In this process, the state learning unit 120 combines the currently measured sensor data Xm_PoolOn (288) and the existing learning data (Xm_Pool) (2,000) as shown in FIG. 9 and deletes the abnormal data 2,000 new learning data Xm are generated. 9 is a diagram illustrating an example of generating new learning data by combining existing sensor data and existing learning data according to an embodiment of the present invention. Accordingly, the state learning unit 120 delivers the new learning data generated in this process to the modeling unit 140. [

Then, the modeling unit 140 applies 2,000 new learning data to the existing plant model after performing auto-learning once a day (once) for a predetermined period, for example, It is a new update.

Here, the plant anomaly detection system 100 according to the present invention may include, in addition to the components shown in Fig. 1, the plant data collected by sensing from the respective sensors as original data, as shown in Fig. 3, The system may further include a plant database for storing the steady state data in which the abnormal data is deleted through the preprocessing process as reference data or modeling the learned steady state data and storing the steady state data as model data .

3, the plant anomaly detection system 100 according to the present invention includes a prediction database (Prediction DB) for storing a plant model generated by modeling learned steady state data as a prediction model, As shown in FIG.

Next, the failure diagnosis unit 150 diagnoses the failure of the plant using the diagnosis logic based on the diagnosis database 170 and tracks the root cause (S250).

Here, the diagnosis logic is an algorithm that is searched based on the diagnostic data based on the root cause of the plant failure if an early alarm occurs due to a difference between the real-time plant data and the steady-state data exceeding the allowable value.

Therefore, the failure diagnosis unit 150 can analyze and trace the root cause of the failure of the plant based on the diagnostic data.

Next, the output unit 160 visualizes and outputs the diagnosis result of the plant (S260).

That is, the output unit 160 may transmit the completed data to each terminal so that the completed data can be confirmed via the web or mobile through the visualization solution.

Also, as shown in FIG. 3, the present invention can secure system development, maintenance easiness, and reusable assets by minimizing the influence between functions by separating system between pre-processing, modeling, residual generation and diagnosis functions.

The present invention provides a system model that reduces the dependency between systems by minimizing the degree of cohesion between systems and minimizing the degree of coupling, thereby making it possible to construct a system that is easy to develop and maintain. It is easily recyclable from other systems.

As described above, according to the present invention, the plant data is acquired in real time, the steady state is automatically learned using the secured data, the steady state obtained by the real-time plant data and the automatic learning is compared with each other, , A plant system for diagnosing faults using a predicted fault, a built-in diagnosis database and diagnostic logic, and tracking and visualizing the root cause can be realized.

It will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the spirit and scope of the present invention as defined by the following claims and their equivalents. Only. It is intended that the present invention covers the modifications and variations of this invention provided they come within the scope of the appended claims and their equivalents. .

100: Plant anomaly detection system 110: Data collection unit
120: state learning unit 130: alarm unit
140: Modeling unit 150: Fault diagnosis unit
160: output unit 170: diagnosis DB

Claims (11)

A diagnostic database that stores failed events in the plant as diagnostic data;
A data collection unit for collecting plant data at a predetermined cycle with respect to the plant;
A state learning unit for learning a steady state using the collected plant data;
An alarm unit for comparing the plant data with the learned steady state data to alarm the failure early;
A modeling unit for modeling the learned steady state data to generate a plant model;
A failure diagnostic unit for diagnosing a failure of the plant based on the diagnostic database and for tracking a root cause; And
An output unit for visualizing and outputting a diagnosis result of the failure diagnosis unit;
Lt; / RTI >
The state learning unit generates the learned steady state data as a learning data set and extracts features by combining the steady state plant data in which the abnormal data is deleted through the preprocessing process in the real time plant data with the learning data set, Generates a new pattern according to the new learning data, updates the learning data,
Wherein the modeling unit receives the updated learning data from the state learning unit and applies the updated learning data to the plant model to update the plant model,
Plant anomaly detection system.
The method according to claim 1,
Wherein the state learning unit deletes the abnormal data through a pre-processing process on the collected real-time plant data, and learns the normal state based on the preprocessed plant data.
The method according to claim 1,
Wherein the warning unit generates the residual data by using the learned real time plant data as the actual value and using the learned steady state data as the predicted value, subtracting the predicted value from the actual value, and when the generated residual value exceeds the allowable value, Alarm Logic) to output an early warning.
The method according to claim 1,
Wherein the modeling unit generates a plant model by modeling the learned steady state data into a parametric model or a non-parametric model.
delete The method according to claim 1,
The diagnostic data for the failed examples are obtained by separately separating the plant data in an abnormal state in which an alarm is generated according to an alarm logic and the plant data in a normal state in which the plant data in an abnormal state is removed, The plant data anomaly detection system.
(a) collecting plant data at a predetermined cycle with respect to the plant in the data collection unit;
(b) learning a steady state by using the collected plant data in a state learning unit;
(c) comparing the plant data and the learned steady state data at an alarm unit to alarm the fault early;
(d) modeling the learned steady state data in a modeling unit to generate a plant model;
(e) diagnosing a failure of the plant based on a diagnostic database in the failure diagnosis unit and tracking the root cause; And
(f) visualizing and outputting the diagnosis result of the plant at the output unit,
In the step (b), the state learning unit may generate the learned steady state data as a learning data set, and combine the steady-state plant data in which the abnormal data is deleted through the preprocessing in the real- Extract new features to generate new learning data, generate new patterns according to new learning data, update learning data,
In the step (d), the modeling unit receives the updated updated learning data from the state learning unit and applies the updated learning data to the plant model to update the plant model.
Plant anomaly detection method.
8. The method of claim 7,
In the step (b), the state learning unit deletes abnormal data through a pre-processing process on the collected real time plant data, and learns a normal state based on the preprocessed plant data. .
8. The method of claim 7,
In the step (c), the warning unit generates, as a residual, a value obtained by subtracting a predicted value from a measured value, using the learned steady-state data as a measured value, and the generated residual value exceeds an allowable value And outputting an early alarm by an alarm logic in the event of a fault.
8. The method of claim 7,
In the step (d), the modeling unit generates a plant model by modeling the learned steady state data into a parametric model or a non-parametric model.
delete

Images