KR102045626B1 - Apparatus and method for evaluating conditions of equipment - Google Patents

Abstract

An apparatus condition evaluating apparatus according to an embodiment of the present invention includes a data collecting unit which collects data on a condition, operation, and control of a facility in real time; A data mining unit configured to apply at least one index to the data to calculate at least one index; A cluster evaluator configured to cluster the data using at least one of the indices, and evaluate the cluster data by comparing the clustered cluster data with steady state data; And a cluster classification unit classifying the evaluated cluster data according to a state level.

Description

Apparatus and method for evaluating conditions of equipment

The present invention provides an apparatus state evaluation apparatus and method for collecting and interpreting information related to a steel process and diagnosing a state of the plant.

There are various facilities in factories that produce steel products. If a problem occurs in such a facility, a problem occurs in the product to be produced, or the production becomes impossible.

For example, in the current steel market, production of products of high quality, high tensile strength, high value-added properties and properties that did not exist in the past has been required and the load on rolling mills for the rolling process is increasing. Due to such an increase in load, problems such as equipment breakdown of rolling mills are increasing.

Conventional apparatuses for evaluating the condition of rolling mills are configured depending on the sensor technology. Specifically, conventional devices collect sensor signals, control signals, and operation signals and analyze and display only the magnitude or frequency of the signal, and the user subjectively determines the state of the rolling mill based on the displayed signals, thereby making it more accurate. Cannot be guaranteed.

Republic of Korea Patent Publication No. 10-1696105

According to one embodiment of the present invention, there is provided an apparatus condition evaluating apparatus and method for more effectively evaluating the condition of a facility using artificial intelligence technology.

An apparatus condition evaluating apparatus according to an embodiment of the present invention includes a data collecting unit which collects data on a condition, operation, and control of a facility in real time; A data mining unit configured to apply at least one index to the data to calculate at least one index; A cluster evaluator configured to cluster the data using at least one of the indices, and evaluate the cluster data by comparing the clustered cluster data with steady state data; And a cluster classification unit classifying the evaluated cluster data according to a state level.

In addition, the facility condition evaluation method according to an embodiment of the present invention comprises the steps of collecting data on the status, operation, and control of the facility in real time; Calculating at least one index by applying statistical techniques to the data; Clustering the data using at least one of the indices and comparing the clustered cluster data with steady state data to evaluate the cluster data; Classifying the evaluated cluster data according to a state level.

According to an embodiment of the present invention, the information related to the equipment used in the steel process, for example, the information collected by the sensor, the information related to the operation of the equipment, the information related to the characteristics of the equipment itself is collected and artificially Intelligent technology can be used to evaluate anomalies in real time.

In addition, an assessment of the plant's abnormal condition can prevent problems such as faults and breakage of the plant.

1 is a block diagram of a facility state evaluation apparatus according to an embodiment of the present invention.
2 is a diagram illustrating clustering performed by the apparatus condition evaluation apparatus according to an embodiment of the present invention.
3 is a diagram illustrating a neural network algorithm included in an apparatus state evaluation apparatus according to an embodiment of the present invention.
4 is a flowchart of a facility condition evaluation method according to an embodiment of the present invention.

DETAILED DESCRIPTION The following detailed description of the invention refers to the accompanying drawings that show, by way of illustration, specific embodiments in which the invention may be practiced. It should be understood that the various embodiments of the present invention are different but need not be mutually exclusive. The following detailed description, therefore, is not to be taken in a limiting sense, and the scope of the present invention, if properly described, is defined only by the appended claims, along with the full range of equivalents to which such claims are entitled.

DETAILED DESCRIPTION Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings so that those skilled in the art may easily implement the present invention.

1 is a block diagram of a facility state evaluation apparatus according to an embodiment of the present invention.

Equipment condition evaluation apparatus according to an embodiment of the present invention may collect information about a plurality of equipment (for example, rolling mills) for the steel rolling process, and may evaluate the state. To this end, the facility condition evaluation apparatus according to an embodiment of the present invention includes a data collector 110, a data mining unit 120, a cluster evaluation unit 130, and a cluster classification unit 140. In addition, the facility state evaluation apparatus may further include an abnormality determination unit 150.

The term '~ part' used in an embodiment of the present invention refers to software or a hardware component such as a field-programmable gate array (FPGA) or an ASIC, and '~ part' performs certain roles. For example, '~ part' may be implemented as a combination of hardware, such as a microprocessor, and software mounted thereon to perform a predetermined operation.

The data collecting unit 110 collects sensor data s_sen, setting data i_set, control data i_con, and operation data i_work in real time to analyze facility status in a steel process. For example, data for rolling mills in the rolling process can be collected in ms. Hereinafter, the equipment is described assuming a rolling mill disposed in the rolling process, but the equipment is not limited to the rolling mill.

The sensor data s_sen may be a sensing signal output from a sensor arranged to detect state information of the rolling mill. The sensor signal may be a vibration signal that detects vibration of the rolling mill.

The control data i_con is data input from a rolling control PLC (Programmable Logic Controller) for controlling a rolling mill, and represents mechanical and physical characteristics for each rolling mill. The control data (i_con) includes speeds per mill, motor current per mill, roll force per mill, roll gap per mill, tension information of entry and exit per mill, push up cylinder position per mill, work roll bending rate, and IMR (Intermediate). Rolle) may include data on at least one of the bending rate.

The setting data i_set and the operation data i_work are data input from a computer controlling the rolling process, and represent information on rolling operation and rolling mill setup information. The set data (i_set) and the operation data (i_work) include at least one of a coil product number currently produced, steel class class information, strip thickness and width existing information, thickness setup information, product strain resistance information, forward slip information, and yield point information of the product. May contain data for one or more.

The data mining unit 120 receives the collected data Data collected by the data collecting unit 110 and calculates at least one index DMi by applying a statistical technique to the data. In detail, the data collector 110 stacks data collected in real time at a predetermined time interval (for example, 1 second) and processes the data primarily through a data mining processing technique. The data mining processing technique may be a technique for calculating the index with respect to the accumulated data. The indices are average, variance, standard deviation, skewness, kurtosis, peak-to-peak, crest factor, k-factor, and mahal. It may include at least one of a Mahalnobis distance, Coefficient of Variance. Hereinafter, since the index may be plural as described above with respect to the data, it is referred to as a plurality of indexes, but is used as a meaning including one index.

The cluster evaluation unit 130 receives a plurality of indexes DMi and the collection data Data from the data mining unit 120. In addition, the collection data Data is clustered using at least one of a plurality of indices DMi, and the clustered data is evaluated by comparing the clustered cluster data with the steady state data.

The cluster evaluator 130 clusters the collection data Data by using a clustering technique among artificial intelligence techniques using one or more of the plurality of indices. In this case, the number of clusters may be defined by the user, and the clustering may be performed using one or more of the plurality of indexes DMi calculated by the mining unit 120. For example, a large amount of data about a rolling mill collected in ms can be accumulated, and when the plurality of indices DMi is used, the data is more easily made according to the regularity of the data represented by the plurality of indices DMi. That is, the plurality of indices DMi may be used as a criterion for clustering. In addition, the influence of noise included in the data on the clustering may be minimized. For example, the clustering can be performed using the K-Means algorithm. The clustering will be described again with reference to FIG. 2.

In addition, the cluster evaluator 130 may compare the clustered cluster data with the steady state data and evaluate the cluster data using a standard deviation based distance from which the cluster data is separated from the steady state data. Here, the steady state data D_std may be input to the cluster evaluator 130 from the standard database 135. Through such evaluation, the cluster evaluation unit 130 may evaluate the degree of normal and abnormal states with respect to the cluster data.

Thereafter, the cluster evaluator 130 outputs the evaluated cluster data D_eva to the cluster classifier 140. In addition, the cluster evaluator 130 may output the evaluated cluster data D_eva to the evaluation database 145.

The cluster classification unit 140 classifies the evaluated cluster data according to the state level. In addition, the cluster classification unit 140 may use a neural network algorithm in classifying the evaluated cluster data. For example, the cluster classification unit 140 may classify the cluster data using a deep neural network (DNN) technique, which is an artificial intelligence technique.

The status level consists of a plurality of grades representing normal and abnormal levels. Such a plurality of grades may be set from the outside by a user using a setting language (eg, Extensible Markup Language (XML)). In addition, the cluster classification unit may learn a plurality of preset grades and based on the plurality of grades. Normal and abnormal levels can be classified.

Meanwhile, the evaluated cluster data D_eva may be stored in the evaluation database 145, and the connection weight between the plurality of nodes included in the neural network algorithm may be updated according to the evaluated cluster data D_eva. Through such an update process, the neural network algorithm of the cluster classification unit 140 may learn the previous evaluation result and reflect it in the current classification process. The neural network algorithm will be described again with reference to FIG. 3.

The abnormality determining unit 150 outputs a probability value of the abnormality of the rolling mill based on the classification distribution ratio of the classified cluster data. The user can recognize the current state of the rolling mill on the basis of the probabilistic value, so that the reliable state can be determined.

As such, the apparatus condition evaluating apparatus according to an embodiment of the present invention may provide an environment capable of processing and analyzing data collected in real time from a steel process in various ways. In addition, this process and analysis can be used to evaluate the abnormal condition of the equipment in real time.

2 is a diagram illustrating clustering performed by the apparatus condition evaluation apparatus according to an embodiment of the present invention.

The cluster evaluation unit 130 (FIG. 1) may receive the collected data (Data) and cluster it. To this end, the cluster evaluation unit may use the K-Means algorithm for clustering. The K-Means algorithm begins by dividing the data into k clusters (A, B, C) and setting initial centroids (ca, cb, cc) for each divided cluster. In addition, the data is classified by comparing spatial distances (eg, Euclidean distance) between the data and the initial center value and matching the data to the initial center value close to the distance. Then, the center value of each cluster is adjusted using the classified data. Next, the process of reclassifying the data based on the adjusted center value and adjusting the center value is repeated until there is no change in the classification of the data. The K-Means algorithm is an embodiment of clustering that the cluster evaluator of the present invention can perform, but is not limited thereto.

3 is a diagram illustrating a neural network algorithm included in an apparatus state evaluation apparatus according to an embodiment of the present invention.

The neural network algorithm may have variable connection weights according to input and output, and may generate current output information from current input information and past output information by using the connection weight. The past output information may have characteristics similar to the short-term memory of living things. In other words, the neural network algorithm can perform the recognition and determination of the equipment state on a principle similar to that of living beings using long-term and short-term memory.

While the neural network algorithm generates current output information, the information for determining the connection weight may be updated by current input information and past output information. That is, the neural network algorithm may learn a method for recognizing and determining a facility on a principle similar to that a living organism learns through state information whenever output information is output.

The neural network algorithm may include an input layer, a hidden layer, and an output layer. In addition, the hidden layer may be composed of a plurality of layers as shown in FIG.

The input layer includes a plurality of input nodes for receiving cluster data. The hidden layer includes a plurality of hidden nodes that receive output information of the input layer and output a result value according to an activation function. In addition, the output layer receives the result value and outputs state level information.

In addition, according to the design, the facility evaluation apparatus may store the current and past evaluated cluster data in the evaluation database and update the connection weights included in the plurality of nodes of the neural network algorithm according to the evaluated cluster data. Through this updating process, the neural network algorithm can learn previous evaluation results and reflect them in the current classification process.

Accordingly, the equipment condition evaluation apparatus according to an embodiment of the present invention can continuously recognize, determine and learn in the rolling process.

In addition, by the subjective judgment of the evaluator, the problem of inconsistent evaluation can be solved and the accuracy of the facility condition evaluation can be improved.

4 is a flowchart of a facility condition evaluation method according to an embodiment of the present invention.

According to an embodiment of the present invention, the method for evaluating facility status starts with collecting data on status, operation, and control of a facility in real time (S110). Next, at least one index is calculated by applying a statistical technique to the data (S120). Thereafter, the data is clustered using one or more of the indices, and the clustered cluster data is evaluated after being compared with the steady state data (S130). In addition, the evaluated cluster data is classified according to the state level using a neural network algorithm (S140).

Finally, a probabilistic numerical value may be output to the user based on the classification ratio of the cluster data and may be provided to the user (S150).

Equipment condition evaluation method according to an embodiment of the present invention can be understood in more detail from the above description with reference to FIGS.

Although the present invention has been described by specific embodiments such as specific components and the like, but the embodiments and the drawings are provided to assist in a more general understanding of the present invention, the present invention is not limited to the above embodiments. For those skilled in the art, various modifications and variations can be made from these descriptions.

100: equipment condition evaluation device
110: data collector
120: data mining unit
130: cluster evaluation unit
140: cluster classification
150: abnormality determination unit

Claims (9)

A data collector for collecting data on the status, operation, and control of the facility in real time;
A data mining unit configured to apply at least one index to the data to calculate at least one index;
A cluster evaluator configured to cluster the data using at least one of the indices, and compare the clustered cluster data with steady state data input from a standard database to evaluate the cluster data;
A cluster classification unit classifying the evaluated cluster data according to a state level using a neural network algorithm; And
An abnormality discrimination unit outputting a probability of abnormality of the facility based on a classification distribution ratio of the classified cluster data as a probabilistic value
Equipment condition evaluation apparatus comprising a.
delete delete The method of claim 1
Wherein the index includes at least one of average, variance, skewness, kurtosis, peak-to-peak value, crest factor, Mahalanobis distance, and coefficient of variation for the data.
The method of claim 1
And the cluster evaluator evaluates the cluster data using a standard deviation based distance between the cluster data and the steady state data.
delete The neural network algorithm of claim 1, wherein
An input layer including a plurality of input nodes for receiving the cluster data;
A hidden layer including a plurality of hidden nodes that receive output information of the input layer and output a result value according to an activation function; And
An output layer that receives the result value and outputs state level information
Equipment condition evaluation apparatus comprising a.
According to claim 1,
The apparatus state evaluation device of the connection weight between a plurality of nodes included in the neural network algorithm is updated according to the result of the cluster evaluation unit.
An apparatus condition evaluation method performed in an apparatus condition evaluation apparatus including at least one microprocessor,
The at least one microprocessor collecting data in real time on the status, operation, and control of the facility;
Calculating, by the at least one microprocessor, at least one index by applying statistical techniques to the data;
The at least one microprocessor clustering the data using one or more of the indices, and evaluating the cluster data by comparing the clustered cluster data with steady state data input from a standard database;
The at least one microprocessor classifying the evaluated cluster data according to a state level using a neural network algorithm; And
Outputting, by the at least one microprocessor, a probability value of whether the facility is abnormal based on a classification distribution ratio of the classified cluster data;
Facility status evaluation method comprising a.

Images