Abnormal detection device, plant system, feature information generation device, abnormal detection method, feature information generation method, and program

Abstract

The abnormality detection device includes a process data acquisition unit that acquires process data of a process having a plurality of process characteristics, and an abnormality detection unit that detects an abnormal state of the process based on the acquired process data and the degree of abnormality indicating a difference from the normal state of the process represented by feature information representing the characteristics of the normal process data for each process characteristic. The feature information representing the normal state of the process is the optimal feature information for each process characteristic, which is searched by evaluating candidates for feature information generated for each process characteristic from a plurality of normal process data acquired when the process is in the normal state based on the error between the estimated process data calculated using the candidates for feature information and the normal process data.

Description

Technical Field

[0001] The present disclosure relates to an abnormality detection device for detecting an abnormality in a process, a plant system including the abnormality detection device, a feature information generation device for generating feature information used for abnormality detection, an abnormality detection method, a feature information generation method, and a program.

Background Art

[0002] Conventionally, a technique for estimating the value of process data acquired in the operation of a plant process has been proposed. In particular, in a process operated based on the experience or intuition of an operator, by estimating the value of a physical quantity that has not been measured or the value of a physical quantity acquired under planned operating conditions, the state of the process can be grasped more accurately. Based on the estimated state of the process, it is possible to improve the operation of the process, detect abnormalities in the process, and the like.

[0003] For example, Patent Document 1 discloses an abnormality prediction detection system that detects an abnormality prediction based on vibration waveform data of a target facility. In the abnormality prediction detection system of Patent Document 1, in order to identify the cause of the abnormality together with the detection of the prediction of the facility abnormality, non-negative matrix factorization (hereinafter, also referred to as "NMF") is used to decompose the vibration spectrogram obtained from the observed vibration waveform data into the frequency components of the original signal source, and feature amount extraction and abnormality detection of the vibration spectrum are performed.

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0005] Here, in a process where there are various operating states, process data having different tendencies can be obtained from various sensors and command signals. In particular, even in operations using the same equipment, such as in the processes of a plant (e.g., a steel plant, a chemical plant, a power generation plant, a power plant), there are also processes in which the measured values of temperature, pressure, etc. vary greatly due to differences in product types, raw materials, compositions, changes in demand, etc. In such a process, if the same estimation model is used to estimate process data, the estimation error may become large depending on the operating state, and it may not be possible to correctly detect an abnormality occurring in the process.

[0006] Also, in the data of a steel process in which a wide variety of operations coexist, it is necessary to appropriately associate a plurality of operating states with the data. If the data is simply segmented along the operating state, the estimation model will be dominated by the tendency of local data, and the estimation error will increase. On the other hand, if the data is roughly classified ignoring the operating state, the estimation model will be dominated by the data with lost feature information, and the estimation error will increase. From this, in the estimation model of the steel process, it is necessary to optimize the classification of data that appropriately captures the tendency and characteristics of the operating state, but conventionally, the classification has been performed by trial and error based on human judgment.

[0007] Therefore, the present disclosure has been made in view of the above problems, and an object of the present disclosure is to provide an abnormality detection device, a plant system, a feature information generation device, an abnormality detection method, a feature information generation method, and a program capable of accurately detecting an abnormality occurring in a process having a plurality of process characteristics.

Means for Solving the Problems

[0008] In order to solve the above problems, according to the present disclosure, there is provided a process data acquisition unit that acquires process data of a process having a plurality of process characteristics, an abnormal detection unit that detects an abnormal state of the process based on the acquired process data and the degree of abnormality indicating a difference from the normal state of the process represented by feature information representing the characteristics of the normal process data for each process characteristic, and the feature information representing the normal state of the process is obtained by evaluating candidates for feature information generated for each process characteristic from a plurality of normal process data acquired when the process is in the normal state based on the error between the estimated process data calculated using the candidates for feature information and the normal process data, and searching for the optimal feature information for each process characteristic.

[0009] The abnormal detection unit may factorize a matrix representing the acquired process data with a feature matrix representing the feature information representing the characteristics of the normal process data of the process to obtain a coefficient matrix representing the feature information of the process data, and calculate the degree of abnormality based on the matrix representing the process data, the feature matrix, and the coefficient matrix.

[0010] The degree of abnormality may be the distance between the value of the process data and the hyperplane representing the normal state of the process by the feature information representing the characteristics of the normal process data for each process characteristic.

[0011] The abnormal detection unit may calculate the degree of deviation representing the degree of abnormality for each operating condition of the process data.

[0012] Furthermore, according to the present disclosure, there is provided a plant system including a plant having a process having a plurality of process characteristics and the above-described abnormal detection device, wherein the abnormal detection device calculates the degree of abnormality or the degree of deviation representing the degree of abnormality for each operating condition of the process data, and controls the plant based on the degree of abnormality or the degree of deviation.

[0013] Also, in order to solve the above problems, according to the present disclosure, from a plurality of normal process data obtained when a process having a plurality of process characteristics is in a normal state, for each process characteristic, a feature information generation unit that generates candidates for feature information representing the characteristics of the normal process data, and a search unit that evaluates the candidates for feature information based on the error between the estimated process data calculated using the candidates for feature information and the normal process data, and searches for the optimal feature information for each process characteristic, a feature information generation device is provided.

[0014] The search unit may calculate an evaluation threshold from the quartiles of the error, and evaluate the feature information based on the number of normal process data exceeding the evaluation threshold.

[0015] The search unit may search for the feature information that minimizes the number of normal process data exceeding the evaluation threshold as the optimal feature information.

[0016] Furthermore, in order to solve the above problems, according to the present disclosure, a process data acquisition step of acquiring process data of a process having a plurality of process characteristics, and an abnormality detection step of detecting an abnormal state of the process based on the abnormality degree indicating the difference between the acquired process data and the normal state of the process represented by the feature information representing the characteristics of the normal process data for each process characteristic, are included. The feature information representing the normal state of the process is the optimal feature information for each process characteristic, searched by evaluating candidates for feature information generated for each process characteristic from a plurality of normal process data obtained when the process is in a normal state, based on the error between the estimated process data calculated using the candidates for feature information and the normal process data. An abnormality detection method is provided.

[0017] Further, in order to solve the above problems, according to the present disclosure, from a plurality of normal process data obtained when a process having a plurality of process characteristics is in a normal state, for each process characteristic, a feature information generation step of generating candidates for feature information representing the characteristics of the normal process data; and a search step of evaluating candidates for feature information based on the error between the estimated process data calculated using the candidates for feature information and the normal process data, and searching for the optimal feature information for each process characteristic, are provided.

[0018] Furthermore, in order to solve the above problems, according to the present disclosure, a computer is caused to function as an abnormality detection device including: a process data acquisition unit that acquires process data of a process having a plurality of process characteristics; and an abnormality detection unit that detects an abnormal state of the process based on the acquired process data and the degree of abnormality indicating the difference from the normal state of the process represented by feature information representing the characteristics of the normal process data for each process characteristic, wherein the feature information representing the normal state of the process is the optimal feature information for each process characteristic searched by evaluating candidates for feature information generated for each process characteristic from a plurality of normal process data obtained when the process is in the normal state, based on the error between the estimated process data calculated using the candidates for feature information and the normal process data.

[0019] Further, in order to solve the above problems, according to the present disclosure, a computer is caused to function as a feature information generation device including: a feature information generation unit that generates candidates for feature information representing the characteristics of normal process data for each process characteristic from a plurality of normal process data obtained when a process having a plurality of process characteristics is in a normal state; and a search unit that evaluates candidates for feature information based on the error between the estimated process data calculated using the candidates for feature information and the normal process data, and searches for the optimal feature information for each process characteristic.

Advantages of the Invention

[0020] As described above, according to the present disclosure, in a process having a plurality of process characteristics, an abnormal state of the process is detected based on the degree of abnormality indicating the difference between the normal state of the process represented by the characteristic information for each process characteristic and the process data. Thereby, an abnormality occurring in the process can be accurately detected.

Brief Description of the Drawings

[0021]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

Figure 11

Figure 12

Figure 13

Figure 14

Figure 15

Embodiments for Carrying Out the Invention

[0022] Hereinafter, preferred embodiments of the present disclosure will be described in detail with reference to the accompanying drawings. In the present specification and drawings, components having substantially the same functional configuration are denoted by the same reference numerals, and redundant description is omitted.

[0023] [1. System Configuration] The abnormality detection system according to the present disclosure is a system for detecting an abnormality occurring in a process based on process data acquired from the process. The process in the present disclosure is a process having a plurality of process characteristics. Process data refers to data obtained in the operation of a process, such as measurement values measured using operating conditions of equipment, product types, sensors, etc. The process characteristic refers to the tendency of the process data acquired in the process. For example, in the process of a steel plant, depending on the process, even if the operation is by the same equipment, the values of the process data may vary greatly depending on differences in product types, etc. For example, in the mold process in a continuous casting facility, the manner in which solidification abnormalities of the solidification shell occur varies depending on the electromagnetic stirring conditions and the type of powder. Such a process is referred to as a process having a plurality of process characteristics.

[0024] In the anomaly detection system according to the present disclosure, as an index for detecting an anomaly in a process, feature information representing the characteristics of normal process data obtained when the process is in a normal state is generated. The feature information serving as the index is generated using normal process data obtained in past operations. For one process characteristic, one optimal piece of feature information representing the characteristics of normal process data that could be obtained in that process characteristic is generated as the feature information serving as the index.

[0025] The anomaly detection system acquires process data from the process while monitoring, for example, whether an anomaly has occurred in the process. The anomaly detection system detects the abnormal state of the process based on the degree of anomaly of the process when the process data is acquired. Here, the degree of anomaly is a value indicating how much the acquired process data deviates from the normal state of the process represented by the generated feature information, and represents the difference between the process data and the normal state of the process represented by the feature information. The anomaly detection system determines how much the process when the process data is acquired deviates from the normal state based on the feature information corresponding to the process characteristic when the process data is acquired, and detects an anomaly in the process. In this way, in a process having a plurality of process characteristics, by detecting an anomaly in the process using the feature information corresponding to the process characteristic of the acquired process data, the accuracy of anomaly detection can be improved.

[0026] FIG. 1 shows a conceptual diagram of the normal state of a process represented by the optimal feature information obtained for each process characteristic. The conceptual diagram shown in FIG. 1 represents the normal state of a process having three process characteristics. Normal process data ND generates clusters c1, c2, and c3 for each process characteristic. The anomaly detection system obtains, for each of the clusters c1, c2, and c3, feature information representing the main characteristics of the normal process data, and obtains a space (normal space) representing the normal state of the entire process. The feature information for each of the clusters c1, c2, and c3 is the direction vector a defining the normal space 1 , a 2 , a 3and the process data is in the normal space spanned by the direction vectors a 1 , a 2 and a 3 the process can be said to be in a normal state if it is within the normal space spanned by. In other words, if the process data deviates from the normal space, the process is likely to be in an abnormal state.

[0027] The degree of abnormality d calculated by the abnormality detection system according to the present disclosure represents how much the process data PD to be determined deviates from the hyperplane indicating the boundary of the normal space, represented by the minimum distance from the normal space. The larger the degree of abnormality d, the more the process data PD to be determined deviates from the normal space, indicating a greater abnormal state of the process. Thus, in the abnormality detection system according to the present disclosure, by expressing the polyhedron of the minimum dimension that reasonably explains the relationship of the process data when the process is normal using the feature information, the abnormality of the process is detected.

[0028] Hereinafter, based on FIG. 2, the configuration of the abnormality detection system 1 according to the present disclosure will be described. FIG. 2 is a block diagram showing an example of the configuration of the abnormality detection system 1 according to the present disclosure. The abnormality detection system 1 according to the present disclosure includes, as shown in FIG. 2, a feature information generation device 100, an abnormality detection device 200, and a feature information storage unit 300.

[0029] [1-1. Feature Information Generation Device] The feature information generation device 100 generates feature information representing the features of the normal process data acquired from the process 10 when the process 10 having a plurality of process characteristics is in a normal state. The feature information generation device 100 includes, as shown in FIG. 2, a feature information generation unit 110 and a search unit 120.

[0030] (Feature Information Generation Unit) The feature information generation unit 110 generates candidates for feature information representing the features of the normal process data from a plurality of normal process data for each process characteristic. The feature information generation unit 110 acquires the normal process data from the process data storage unit 30 that stores the process data acquired in the past operations. The feature information generation unit 110 acquires the normal process data used to generate the candidates for the feature information from the process data storage unit 30. The feature information generation unit 110 may acquire a plurality of process data acquired during a predetermined period, or may acquire a predetermined number of process data.

[0031] The feature information generation unit 110 according to the present disclosure acquires the feature information by using non-negative matrix factorization (hereinafter also referred to as "NMF") for the acquired normal process data. NMF is an algorithm that decomposes one non-negative matrix into the product of two non-negative matrices. The feature information generation unit 110 acquires a matrix (hereinafter also referred to as a "feature matrix") representing the feature information of each process characteristic by decomposing the normal process data into a non-negative matrix using NMF. A detailed description of the generation process of the feature information will be described later. The feature information generation unit 110 outputs the candidates for the feature information of each acquired process characteristic to the search unit 120.

[0032] (Search Unit) The exploration unit 120 evaluates candidates for feature information based on the error between the estimated process data calculated using the candidates for feature information and the normal process data, and explores the optimal feature information for each process characteristic. The exploration by the exploration unit 120 means evaluating a plurality of candidates for feature information to obtain one optimal feature information. The optimal feature information to be explored represents the main features of the normal process data. That is, the exploration unit 120 explores feature information that represents features that are applicable to as much of the process data obtained when the process is in a normal state as possible. The exploration unit 120 collectively evaluates the candidates for feature information of each process characteristic. That is, the exploration unit 120 evaluates a plurality of combination candidates of feature information of each process characteristic, and finally explores the optimal feature information for each process characteristic. Thereby, the feature information of each process characteristic that is optimal for the entire process is obtained.

[0033] Since the estimated process data is calculated based on the candidates for feature information obtained from the normal process data, it is considered to represent the normal process data. Therefore, the error between the estimated process data and the normal process data indicates how much the features represented by the candidates for feature information used in the calculation of the estimated process data deviate from the features of the actual normal process data. The larger the error, the more the features represented by the candidates for feature information deviate from the features of the normal process data.

[0034] Based on the error between such estimated process data and normal process data, the search unit 120 sets, for example, an evaluation index and evaluates candidates for feature information using the evaluation threshold value. The evaluation threshold value can be calculated, for example, from the quartiles of the error between the estimated process data and the normal process data. Quartiles refer to the values at the cutoffs that divide the data into four equal parts when the data is arranged in ascending order of values, and are called the first quartile, the second quartile (i.e., the median), and the third quartile from the smaller value side. The first quartile is the value located at 25% of the whole, and the third quartile is the value located at 75% of the whole. Since the quartiles are values set based on the median, they are useful for setting thresholds based on the main features of the process data. Also, the quartiles are useful values for grasping the variation in the values of the data and have the characteristic of being less affected by outliers. Therefore, the search unit 120 sets an evaluation threshold value for distinguishing between normal process data and process data that deviates from the normal process data using the quartiles of the error.

[0035] For example, the search unit 120 may use the sum of the difference between the third quartile and the median and the third quartile as the evaluation threshold value. Such an evaluation threshold value means a value that is assumed to distinguish most normal process data as normal process data. The smaller the number of normal process data whose error from the estimated process data calculated using the candidate for feature information exceeds the evaluation threshold value, the more it can be said that the evaluation threshold value can correctly evaluate whether the process data is normal process data. The search unit 120 evaluates candidates for feature information using such an evaluation threshold value and searches for the candidate for feature information that minimizes the number of normal process data exceeding the evaluation threshold value as the optimal feature information.

[0036] The search unit 120 records the optimal feature information for each process characteristic in the feature information storage unit 300.

[0037] [1-2. Anomaly Detection Device] The abnormality detection device 200 detects the abnormal state of process 10 from the process data to be determined, using the characteristic information of process 10. As shown in FIG. 2, the abnormality detection device 200 includes a process data acquisition unit 210 and an abnormality detection unit 220.

[0038] (Process Data Acquisition Unit) The process data acquisition unit 210 acquires the process data to be determined. The process data acquisition unit 210 may acquire, for example, the process data obtained from sensors or the like provided in the equipment during the operation of process 10 as the process data to be determined, or may acquire the process data input from the terminal 500 as the process data to be determined. The process data acquisition unit 210 outputs the acquired process data to be determined to the abnormality detection unit 220.

[0039] (Abnormality Detection Unit) The abnormality detection unit 220 detects the abnormal state of process 10 based on the degree of abnormality indicating the difference between the acquired process data to be determined and the normal state of process 10 represented by the characteristic information representing the characteristics of the normal process data for each process characteristic. The degree of abnormality is represented by the distance between the value of the process data to be determined and the hyperplane representing the normal state of process 10 by the characteristic information representing the characteristics of the normal process data for each process characteristic. The greater the degree of abnormality, the more the process 10 deviates from the normal state when the process data to be determined is acquired, and the higher the possibility of an abnormality occurring.

[0040] For example, when the feature information is obtained using NMF, the abnormality detection unit 220 may factorize a matrix representing the process data to be determined using a feature matrix representing the feature information, obtain a coefficient matrix representing the feature information of the process data, and calculate the degree of abnormality, which is the error between the process data to be determined and the feature information, using these matrices. By using NMF, it is possible to automatically determine the degree of abnormality based on the feature information corresponding to the process characteristics of the process data to be determined among the plurality of process characteristics that Process 10 has. Further, the abnormality detection unit 220 may calculate the degree of deviation representing the degree of abnormality for each operating condition of the process data. Detailed explanations of the calculation processes for the degree of abnormality and the degree of deviation will be described later.

[0041] The abnormality detection unit 220 outputs the calculated degree of abnormality to, for example, the terminal 500. When the abnormality detection unit 220 calculates the degree of deviation, it may output the degree of deviation to the terminal 500 together with the degree of abnormality.

[0042] Further, the abnormality detection unit 220 may determine whether or not the calculated degree of abnormality exceeds an evaluation threshold value, and perform abnormality detection on Process 10. When the degree of abnormality exceeds the evaluation threshold value, there may be an abnormality in Process 10. At this time, the abnormality detection unit 220 may notify the terminal 500 that there may be an abnormality in Process 10.

[0043] Note that the terminal 500 is an information processing device for a user to input information and present information to the user. The terminal 500 may be, for example, a personal computer, a tablet terminal, or the like. The user can confirm the value estimated by the abnormality detection device 200 via the terminal 500, and if there is an abnormality in Process 10, the user can take measures to improve it.

[0044] The above described a configuration example of the abnormality detection system 1 according to the present disclosure. In FIG. 2, an example in which the feature information generation device 100, the abnormality detection device 200, and the feature information storage unit 300 are configured by different devices is shown. However, the present disclosure is not limited to such an example. For example, it is also possible to configure at least two of the feature information generation device 100, the abnormality detection device 200, and the feature information storage unit 300 by one device. For example, the abnormality detection device 200 may include the feature information generation unit 110 and the search unit 120 of the feature information generation device 100 as a processing unit for generating feature information representing the normal state of the process. Alternatively, the feature information generation device 100 may be configured by two or more devices, and the abnormality detection device 200 may be configured by two or more devices.

[0045] Also, it is possible to create a program for realizing each function of the feature information generation unit 110, the search unit 120 of the feature information generation device 100, the process data acquisition unit 210, and the abnormality detection unit 220 of the abnormality detection device 200 described above, and install it on a computer or the like. By the computer executing the program installed by the CPU (Central Processing Unit) or the like of the computer, each function of the feature information generation device 100 and the abnormality detection device 200 is realized. Also, a computer-readable recording medium storing such a program can be provided. The recording medium is, for example, a magnetic disk, an optical disk, a magneto-optical disk, a flash memory, or the like. Also, the above program may be distributed via a network, for example, without using a recording medium.

[0046] [2. Abnormality Detection of Process] In the anomaly detection system 1, the feature information generation device 100 executes a feature information generation method for generating feature information representing the features of normal process data acquired from the normal process 10, and the anomaly detection device 200 executes an anomaly detection method for detecting an anomaly in the process 10 from the process data to be determined using the feature information of the process 10. The feature information generation method and the anomaly detection method can each be realized by a computer executing a program including each step described below. Hereinafter, the feature information generation method and the anomaly detection method according to the present disclosure will be described.

[0047] [2-1. Feature Information Generation Method] (2-1-1. Overview) FIG. 3 is a flowchart showing an overview of the feature information generation method according to the present disclosure. In the feature information generation method according to the present disclosure, as shown in FIG. 3, first, the feature information generation unit 110 generates candidates for feature information representing the features of the normal process data from a plurality of normal process data for each process characteristic (S11: feature information generation step). Next, the search unit 120 evaluates the candidates for feature information based on the error between the estimated process data calculated using the candidates for feature information generated in step S11 and the normal process data, and searches for the optimal feature information for each process characteristic (S13: search step).

[0048] More specifically, in the feature information generation method according to the present disclosure, as an example, non-negative matrix factorization (NMF) is used to generate feature information. NMF is an algorithm that decomposes one non-negative matrix into the product of two non-negative matrices. By decomposing the normal process data used for feature information generation into a non-negative matrix using NMF, it is possible to automatically obtain feature information on a plurality of process characteristics possessed by the process. Hereinafter, an example of the feature information generation method using NMF will be described in detail.

[0049] (2-1-2. Feature Information Generation Method Using NMF) FIG. 4 is a flowchart showing an example of a feature information generation method using NMF. In using NMF, the past process data recorded in the process data storage unit 30 is represented by a matrix Y as shown in FIG. 5. The matrix Y is a T-row M-column matrix representing time in rows and data items of past process data in columns (Y ∈ R (T,M) ). The value of each data item in each row of the matrix Y indicates the value of the process data acquired at each time and is associated with the time. Hereinafter, the matrix Y is also referred to as the "past data matrix Y". Note that the process data at each time of the past data matrix Y does not have to be arranged in time series. Also, the process data of the past data matrix Y may be process data acquired at equal intervals or data acquired at unequal intervals. Furthermore, the process data of the past data matrix Y may be continuous data or discontinuous data.

[0050] (S101: Initial value setting) In the feature information generation method using NMF, as shown in FIG. 4, first, the feature information generation unit 110 sets initial values of the classification number N of process data, the coefficient matrix Φ, and the operation state matrix X (S101).

[0051] The classification number N is the number of classes for classifying process data. The initial value and the upper limit value of the classification number N can be arbitrarily set. For example, the initial value N 0 of the classification number N may be set to 2, and the upper limit value N max of the classification number may be set to 10.

[0052] The coefficient matrix Φ and the operation state matrix X are non-negative matrices generated by decomposing the past data matrix Y using NMF as shown in FIG. 6. The coefficient matrix Φ is a T-row N-column matrix representing time in rows and classification of process data (i.e., class) in columns (Φ ∈ R (T,N) ). The coefficient matrix Φ is also referred to as a feature amount matrix. The operation state matrix X is an N-row M-column matrix representing the classification of process data in rows and the data items of process data in columns (Φ ∈ R (N,M))。The operation state matrix X is also referred to as a feature matrix. The initial values of the coefficient matrix Φ and the operation state matrix X may be set randomly, for example.

[0053] Also, as an initial setting, the feature information generation unit 110 sets an upper limit value for the number of trials L. For example, the upper limit value of the number of trials L max may be set to 100. Note that the number of trials L starts from 1.

[0054] (S103: Generation of Feature Information) Next, when the feature information generation unit 110 acquires the past process data recorded in the process data storage unit 30, it uses NMF to approximate the past data matrix Y representing the past process data by the matrix product of the coefficient matrix Φ and the operation state matrix X (Y≈Φ·X) (S103). Here, the past process data is normal process data acquired when the process is in a normal state.

[0055] The coefficient matrix Φ is generally composed of sparse matrix elements as shown in FIG. 6. When each row of the coefficient matrix Φ is the matrix element Φ j (j = 1,…,T), the feature information generation unit 110 classifies (clusters) by associating the values of the data items with the matrix element Φ j by a scalar product. At this time, the matrix element Φ j represents the degree of contribution of each process characteristic of the process at each time. That is, each column of the matrix element Φ j represents a class corresponding to the process characteristic. Note that the operation state matrix X represents the characteristics of time-invariant process data independent of time. The operation state matrix X can be said to be feature information representing the characteristics of the normal process data obtained from the process 10. Each row of the operation state matrix X indicates a direction vector that is feature information of the normal process data having the same process characteristics. The feature information representing each process characteristic of the process 10 is, for example, the direction vectors a 1 , a 2 , a 3 shown in FIG. 1, and represents the proportional relationship of the normal process data in each process characteristic.

[0056] Note that, in the matrix element Φ of the coefficient matrix Φ obtained using NMF, since each column represents a class corresponding to a process characteristic, it can be said that the process data belongs to the class of the column having the maximum value. Thus, it can be seen that the class to which the process data at each time belongs is represented in the matrix element Φ of the coefficient matrix Φ, and each process characteristic is automatically specified. j In, since each column represents a class corresponding to a process characteristic, it can be said that the process data belongs to the class of the column having the maximum value. Thus, it can be seen that the class to which the process data at each time belongs is represented in the matrix element Φ of the coefficient matrix Φ, and each process characteristic is automatically specified. j Note that, in the matrix element Φ of the coefficient matrix Φ obtained using NMF, since each column represents a class corresponding to a process characteristic, it can be said that the process data belongs to the class of the column having the maximum value. Thus, it can be seen that the class to which the process data at each time belongs is represented in the matrix element Φ of the coefficient matrix Φ, and each process characteristic is automatically specified.

[0057] When the feature information generation unit 110 represents past normal process data as the matrix product of the coefficient matrix Φ and the operation state matrix X, the coefficient matrix Φ and the operation state matrix X are output to the search unit 120. At this time, the operation state matrix X output from the feature information generation unit 110 to the search unit 120 is a candidate for feature information.

[0058] (S105 - S121: Search for Optimal Feature Information) Next, the search unit 120 evaluates the candidate for feature information input from the feature information generation unit 110 and searches for the optimal feature information. The search unit 120 evaluates the candidate for feature information based on the error between the estimated process data calculated using the candidate for feature information and the normal process data.

[0059] Specifically, first, the search unit 120 calculates the error between the estimated process data calculated using the candidate for feature information and the normal process data, and calculates the Euclidean norm value for each row of the matrix (S105). The error between the estimated process data and the normal process data is represented by the difference (Y - Φ·X) between the past data matrix Y and the matrix product of the coefficient matrix Φ and the operation state matrix X. The search unit 120 obtains the Euclidean norm value for each row of the matrix representing the error. The Euclidean norm value represents the magnitude of the vector data represented by each row, and the larger the value, the larger the value.

[0060] Based on the calculated Euclidean norm value, the exploration unit 120 calculates an evaluation threshold for evaluating whether the candidate feature information appropriately represents the mainstream features of the normal process data (S107). In this example, the exploration unit 120 calculates the evaluation threshold from the quartiles of the error between the estimated process data and the normal process data. As described above, since the quartiles are values set based on the median, they are useful for setting the threshold based on the mainstream features of the process data. For example, the exploration unit 120 calculates the sum of the difference between the third quartile and the median and the third quartile as the evaluation threshold. Such an evaluation threshold means a value that is assumed to distinguish most normal process data as normal process data.

[0061] Then, the exploration unit 120 evaluates the candidate feature information using the evaluation threshold calculated in step S107. Specifically, the exploration unit 120 calculates the number of times the Euclidean norm value calculated in step S105 exceeds the evaluation threshold (hereinafter, also referred to as the "number of records exceeding the threshold") (S109), and evaluates the candidate feature information based on the number of records exceeding the threshold (S111).

[0062] Fig. 7 shows an image of the relationship between the calculated Euclidean norm value of the estimated error and the Euclidean norm value of the estimated error for a plurality of normal process data used for generating the candidate feature information. In the example of Fig. 7, among a part of the plurality of normal process data, there are records exceeding the threshold where the Euclidean norm value of the estimated error exceeds the evaluation threshold. Since the evaluation threshold means a value that can distinguish most normal process data as normal process data, when the candidate feature information correctly represents the features of the normal process data, the Euclidean norm values of all the estimated errors are below the evaluation threshold.

[0063] In other words, an over-threshold record in which the Euclidean norm value of the estimation error exceeds the evaluation threshold means that the candidate characteristic information deviates from the characteristics of the normal process data represented by the candidate characteristic information. In other words, the over-threshold record is normal process data that is determined to deviate from normal process data by the evaluation threshold despite being normal process data. The over-threshold record can also be said to be process data that is overdetected by the evaluation threshold as being process data acquired during a process abnormality. Therefore, the fewer the over-threshold records, the more accurately the candidate characteristic information used to calculate the evaluation threshold can be evaluated to represent the characteristics of normal process data.

[0064] Therefore, in step S111, the search unit 120 judges whether the number of records exceeding the threshold has updated the minimum value, thereby evaluating whether the candidate for characteristic information to be evaluated this time represents a more mainstream characteristic of normal process data than the characteristic information previously evaluated. If the number of records exceeding the threshold has updated the minimum value (S111: YES), the search unit 120 holds the number of classifications N of the current process data, the operation state matrix X, and the evaluation threshold (S113). On the other hand, if the number of records exceeding the threshold has not updated the minimum value (S111: NO), the search unit 120 proceeds to the process of step S115.

[0065] When the process up to step S113 is completed, the search unit 120 checks whether the number of trials L is equal to the upper limit L max It is determined whether the number of trials L reaches the upper limit L max If it has not been reached (S115: NO), the search unit 120 changes the coefficient matrix Φ and the operation state matrix X (S117), and adds 1 to the number of trials L (S119). After that, the processes of steps S103 to S115 are repeatedly performed.

[0066] On the other hand, the number of trials L is the upper limit of the number of trials L max If so (S115: YES), the search unit 120 checks whether the number of classifications N of the process data is equal to or smaller than the upper limit value N max It is determined whether the number of classifications N of the process data is equal to or exceeds the upper limit value Nmax If it has not reached (S121: NO), the search unit 120 changes the coefficient matrix Φ and the operation state matrix X (S123), resets the number of trials L to 1, and adds 1 to the number of classifications N (S125). Then, the processes of steps S103 to S121 are repeatedly executed.

[0067] And when the number of classifications N of the process data reaches the classification number upper limit value N max (S121: YES), the search unit 120 stores the currently held operation state matrix X, the number of classifications N, and the evaluation threshold value in the feature information storage unit 300 as the optimal operation state matrix X, the number of classifications N, and the evaluation threshold value (S127). The optimal operation state matrix X becomes the optimal feature information searched from the candidates of the feature information.

[0068] As described above, an example of the feature information generation method using NMF according to the present disclosure has been described. According to the present disclosure, by decomposing the normal process data used for feature information generation into a non - negative matrix using NMF, it is possible to automatically obtain the feature information of a plurality of process characteristics possessed by the process. Also, in a large - scale system, there are many types of process data obtained in the operation of the process, such as measurement values measured using the operation conditions of equipment, product types, sensors, etc. The operation state matrix X represented by these process data is not always uniquely determined. In such a large - scale system, it is significant to obtain the optimal operation state matrix X by search as in the feature information generation method according to the present disclosure.

[0069] [2 - 2. Anomaly detection method] (2 - 2 - 1. Overview) FIG. 8 is a flowchart showing an overview of the abnormality detection method according to the present disclosure. In the abnormality detection method according to the present disclosure, as shown in FIG. 8, first, the process data acquisition unit 210 acquires process data to be determined (S21: process data acquisition step). Next, the abnormality detection unit 220 detects the abnormal state of the process 10 based on the degree of abnormality indicating the difference between the acquired process data to be determined and the normal state of the process 10 represented by the feature information representing the characteristics of the normal process data for each process characteristic (S23: abnormality detection step).

[0070] Here, the feature information representing the normal state of the process can be obtained by the feature information generation method shown in FIG. 3 above. That is, the feature information representing the normal state of the process (the feature information representing the characteristics of the normal process data for each process characteristic described above) is the optimal feature information for each process characteristic, which is searched by evaluating candidates for feature information generated for each process characteristic from a plurality of normal process data acquired when the process is in a normal state based on the error between the estimated process data calculated using the candidates for feature information and the normal process data.

[0071] More specifically, in the abnormality detection method according to the present disclosure, as an example, the distance between the hyperplane representing the normal state of the process 10 by the feature information obtained using non-negative matrix factorization (NMF) and the value of the acquired process data is calculated as the degree of abnormality. By using NMF, the degree of abnormality can be determined by the feature information corresponding to the process characteristics of the process data to be determined among the plurality of process characteristics of the process 10 without specifying the process characteristics of the process data to be determined. Hereinafter, an example of the abnormality detection method using NMF will be described in detail.

[0072] (2-2-2. Abnormality Detection Method Using NMF) FIG. 9 is an example of a flowchart showing an abnormality detection method using NMF.

[0073] (S200: Process Data Acquisition) First, the process data acquisition unit 210 acquires process data to be determined for calculating the degree of abnormality in order to detect an abnormality in process 10 (S200). The process data acquisition unit 210 may acquire, for example, process data obtained during the operation of process 10 as the process data to be determined, or may acquire process data input from the terminal 500 as the process data to be determined. The process data acquisition unit 210 outputs the acquired process data to be determined to the abnormality detection unit 220.

[0074] (S210 - S250: Abnormality detection process) The abnormality detection unit 220 calculates the degree of abnormality based on the acquired process data to be determined and detects the abnormal state of process 10.

[0075] First, the abnormality detection unit 220 acquires the characteristic information of the process 10 from the characteristic information storage unit 300 (S210). Specifically, the abnormality detection unit 220 acquires, for the process 10, the optimal operation state matrix X, the number of classifications N, and the evaluation threshold value generated by the characteristic information generation device 100 from the characteristic information storage unit 300.

[0076] Next, the abnormality detection unit 220 calculates the coefficient matrix φ using NMF from the number of classifications N and the operation state matrix X acquired in step S210 and the process data matrix y representing the process data to be determined acquired in step S200 (S220). Here, the process data matrix y is a 1-row M-column matrix in which the values of the data items are arranged (y ∈ R (1,M) ). The coefficient matrix φ is a 1-row N-column matrix corresponding to any of the matrix elements Φ j (j = 1,..., T) of the coefficient matrix Φ (φ ∈ R (1,N) ).

[0077] The coefficient matrix φ is determined from the input process data matrix y and the operating state matrix X. Once the coefficient matrix φ is determined, the anomaly detection unit 220 can identify the class to which the process data to be estimated belongs from the column with the maximum value. By identifying the class to which the process data to be estimated belongs, the process characteristics can be identified. The anomaly detection unit 220 calculates the coefficient matrix φ representing the characteristic information of the process data by approximating the process data matrix y with the matrix product (y ≈ φ·X) of the coefficient matrix φ and the operating state matrix X which is the feature matrix.

[0078] Note that in this example, since NMF is used, the coefficient matrix φ needs to be a non - negative matrix. To ensure that the coefficient matrix φ is a non - negative matrix, instead of multiplying both sides of the matrix product (y ≈ φ·X) of the coefficient matrix φ and the operating state matrix X by the pseudo - inverse matrix of the operating state matrix X from the right side, a method of approximating the process data matrix y with the matrix product of the coefficient matrix φ and the operating state matrix X is used.

[0079] Furthermore, the anomaly detection unit 220 calculates the absolute value of the reconstruction error (|y - φ·X|) as the degree of anomaly (S230). The degree of anomaly represented by the reconstruction error represents the distance between the hyperplane (φ·X) representing the normal state of the process 10 and the value of the process data (y) to be determined. Then, the anomaly detection unit 220 determines whether the calculated degree of anomaly exceeds the evaluation threshold value obtained in step S210 (S240). If the degree of anomaly is less than or equal to the evaluation threshold value (S240: NO), the anomaly detection unit 220 determines that the process data to be determined is normal process data and that no anomaly has occurred in the process 10, and ends the process shown in FIG. 9.

[0080] On the other hand, when the degree of anomaly exceeds the evaluation threshold value (S240: YES), the anomaly detection unit 220 evaluates that the process data to be determined is not normal process data and determines that there may be an anomaly in the process 10. At this time, the anomaly detection unit 220 notifies the terminal 500 that there may be an anomaly in the process 10 (S250).

[0081] At this time, the abnormality detection unit 220 may notify the terminal 500 of the degree of deviation representing the degree of abnormality for each operating condition of the process data, together with the degree of abnormality. The degree of deviation may be represented, for example, by the restoration error (y - φ·X). Since the restoration error represents the degree of abnormality in terms of components for each data item, it is possible to identify factors that increase the degree of abnormality among the operating conditions of the process data.

[0082] Alternatively, the degree of deviation may be represented, for example, by a vector from the center of gravity of the normal process data to the process data to be determined. At this time, the center of gravity of the normal process data is the center of gravity of the normal process data having the same process characteristics as the process data to be determined. FIG. 10 shows a conceptual diagram when the degree of deviation is represented by a vector. As shown in FIG. 10, when the process data PD to be determined has the same process characteristics as the normal process data ND in cluster c3, the abnormality detection unit 220 uses the average of the normal process data ND in cluster c3 as the center of gravity G, and obtains the vector from the center of gravity G to the process data PD as the degree of deviation. Note that the magnitude of the vector may be used as the degree of abnormality d.

[0083] As described above, an example of the abnormality detection method using NMF according to the present disclosure has been explained.

[0084] [2-3. Numerical example] An example of the feature information generation method and the abnormality detection method according to the present disclosure will be described using a simple example. In this example, two operating conditions (y1, y2) are used as data items of the process data matrix. First, based on the feature information generation method shown in FIG. 4, optimal feature information representing the normal state of the process is obtained from past normal process data using NMF. Here, the optimal feature information is obtained using the normal process data shown in Table 1 below.

[0085]

Table 1

[0086] In obtaining the optimal feature information, first, the normal process data in Table 1 above is represented by the past data matrix Y. In this example, the past data matrix Y is a matrix of T rows and M columns (Y ∈ R (T,M) ) where the rows correspond to unique numbers (T = 7) for the time when the process data was acquired and the columns represent the operating conditions (y1, y2) (M = 2). Then, the past data matrix Y is approximated as the matrix product of the coefficient matrix Φ and the operating state matrix X (Y ≈ Φ·X). For example, assuming the number of classifications N = 2 and using NMF, the input data matrix Y can be approximated as the matrix product of the coefficient matrix Φ and the operating state matrix X (Y ≈ Φ·X), and can be represented as shown in FIG. 11. Here, it is assumed that the coefficient matrix Φ and the operating state matrix X shown in FIG. 11 are the optimal feature information searched using the feature information generation method of FIG. 4.

[0087] Each row of the operating state matrix X, which is the feature matrix, represents a direction vector that is the feature information of normal process data having the same process characteristics. That is, as shown in FIG. 12, the data of No. 1 and No. 2 take values on the straight line L1 with the direction vector (2,1), and the data of No. 3 to No. 7 take values on the straight line L2 with the direction vector (3,1). Note that in the matrix element Φ j of the coefficient matrix Φ obtained using NMF, since each column represents the class corresponding to the process characteristic, it can be said that the process data belongs to the class of the column having the maximum value. Thus, it can be seen that the class to which the process data at each time belongs is represented in the matrix element Φ j of the coefficient matrix Φ, and each process characteristic is automatically specified.

[0088] In the feature information generation method according to the present disclosure, the past data matrix Y is approximated by the matrix product (Y≈Φ·X) of various coefficient matrices Φ and the operation state matrix X to generate candidates for feature information and search for optimal feature information. For example, as shown in FIG. 12, among the candidates for feature information, there is feature information obtained by combining the direction vectors of line C1 and line C2, feature information obtained by combining the direction vectors of line D1 and line D2, and the like. From such various pieces of feature information representing normal process data, the one that appropriately represents the main features of the normal process data is searched for and used as the optimal feature information.

[0089] Further, in order to calculate the evaluation threshold value for the searched optimal feature information, for example, as shown in FIG. 13, the difference (Y - Φ·X) between the past data matrix Y and the matrix product of the coefficient matrix Φ and the operation state matrix X is taken, and the Euclidean norm value is obtained for each row. Then, for example, using the quartiles of the angular Euclidean norm values, the evaluation threshold value is calculated. As shown in FIG. 13, the Euclidean norm values of the data from No. 1 to No. 7 are 0.0, 1.0, 1.4, 2.0, 2.0, 2.3, and 3.0, and the median value is 2.0. For example, if the evaluation threshold value is set as the sum of the difference between the third quartile and the median value and the third quartile, the evaluation threshold value is 2.3 + 0.3 = 2.6. In this example, the number of records exceeding the threshold value where the Euclidean norm value exceeds the evaluation threshold value is 1.

[0090] Then, based on the obtained optimal feature information, based on the anomaly detection method shown in FIG. 9, the anomaly degree of the process at a certain point in time is calculated from the process data obtained at that point in time. For example, it is assumed that process data with operation conditions y1 = 180 and y2 = 95 is obtained. At this time, when the coefficient matrix φ is calculated using the 1-row M-column process data matrix y representing the obtained process data and the operation state matrix X obtained as the optimal feature information, as shown in FIG. 14, a 1-row 2-column coefficient matrix φ with φ1 = 90 and φ2 = 0 is obtained.

[0091] Thereafter, the reconstruction error (y - φ·X) is calculated, and its absolute value (|y - φ·X|) is calculated as the degree of abnormality. The reconstruction error is the component display of the degree of abnormality for each data item, and is also referred to as the degree of deviation. For example, in the example of FIG. 14, the reconstruction error (degree of deviation) is 0 for the operating condition y1 and does not deviate from the normal state, but is 5 for the operating condition y2, indicating that it deviates from the normal state. At this time, the absolute value of the reconstruction error is 5.

[0092] Note that the degree of deviation may also be represented by a vector from the centroid of the normal process data shown in FIG. 10 to the process data to be determined. In this case, since the process data matrix y has the same process characteristics as No. 1 and No. 2, the degree of deviation may be obtained using the average vector (180 90.25) of No. 1 and No. 2 as the centroid G. At this time, the degree of deviation is (0 4.75) (=(180 95)-(180 90.25)).

[0093] As described above, by using the feature information generation method and the abnormality detection method according to the present disclosure, the normal process data used for feature information generation is decomposed into a non-negative matrix using NMF, so that the feature information of a plurality of process characteristics possessed by the process can be automatically obtained. Further, based on the feature information corresponding to the process characteristics when the process data is acquired, it is determined how much the process deviates from the normal state when the process data is acquired, and the abnormality of the process is detected. Thereby, without clustering the process data according to the process characteristics, the abnormality of the process can be detected using the feature information corresponding to the process characteristics of the acquired process data, and the abnormality can be detected with high accuracy.

[0094] [3. Hardware Configuration] Based on FIG. 15, the hardware configurations of the feature information generation device 100 and the abnormality detection device 200 according to the present disclosure will be described. FIG. 15 is a block diagram showing an example of the hardware configuration of an information processing device 900 that functions as the feature information generation device 100 or the abnormality detection device 200 according to the present disclosure.

[0095] The information processing apparatus 900 includes one or more hardware processors such as the CPU 901, and one or more memories such as a RAM (Random Access Memory) 905 and a ROM (Read Only Memory) 903. Various operations are executed by one or more programs stored in the memory being executed by one or more hardware processors. The information processing apparatus 900 also includes a bus 907, an input I / F 909, an output I / F 911, a storage device 913, a drive 915, a connection port 917, and a communication device 919.

[0096] For example, the CPU 901 functions as an arithmetic processing unit and a control unit. The CPU 901 controls all or part of the operations within the information processing apparatus 900 according to various programs recorded in the ROM 903, the RAM 905, the storage device 913, or the removable recording medium 925. The ROM 903 stores programs, arithmetic parameters, etc. used by the CPU 901. The RAM 905 temporarily stores programs used by the CPU 901 or parameters that change as appropriate during the execution of the programs. These are interconnected by a bus 907 formed by an internal bus such as a CPU bus. The bus 907 is connected to an external bus such as a PCI (Peripheral Component Interconnect / Interface) bus or a PCI Express (registered trademark) via a bridge.

[0097] Note that the arithmetic processing unit and the control unit may be realized by a PLC (Programmable Logic Controller) in addition to the CPU 901, or may be realized by dedicated hardware such as an ASIC (Application Specific Integrated Circuit).

[0098] The input I / F 909 is an interface that receives input from an input device 921, which is an operation means operated by a user, such as a mouse, keyboard, touch panel, button, switch, and lever. The input I / F 909 is configured as, for example, an input control circuit that generates an input signal based on information input by the user using the input device 921 and outputs it to the CPU 901. The input device 921 may be, for example, a remote control device using infrared rays or other radio waves, or an external device 927 such as a PDA corresponding to the operation of the information processing device 900. The user of the information processing device 900 can operate the input device 921 to input various data to the information processing device 900 or instruct processing operations.

[0099] The output I / F 911 is an interface that outputs the input information to an output device 923 that can notify the user visually or audibly. The output device 923 may be, for example, a display device such as a CRT display device, a liquid crystal display device, a plasma display device, an EL display device, and a lamp. Alternatively, the output device 923 may be an audio output device such as a speaker and headphones, a printer, a mobile communication terminal, a facsimile, or the like. The output I / F 911 instructs the output device 923 to output, for example, the processing results obtained by various processes executed by the information processing device 900. Specifically, the output I / F 911 instructs the display device to display the processing results by the information processing device 900 in text or image. Further, the output I / F 911 instructs the audio output device to convert an audio signal such as voice data received a reproduction instruction into an analog signal and output it.

[0100] The storage device 913 is one of the storage units of the information processing device 900 and is a device for storing data. The storage device 913 is a non-temporary tangible computer-readable recording medium. The storage device 913 is composed of, for example, a magnetic storage device such as an HDD (Hard Disk Drive), a semiconductor storage device such as an SSD (Solid State Drive), an optical storage device, or a magneto-optical storage device. The storage device 913 stores programs executed by the CPU 901, various data generated by the execution of the programs, and various data acquired from the outside.

[0101] The drive 915 is a reader / writer for a recording medium and is built-in or externally attached to the information processing device 900. The drive 915 reads the information recorded on the removable recording medium 925 that is mounted and outputs it to the RAM 905. Also, the drive 915 can write information to the removable recording medium 925 that is mounted. The removable recording medium 925 is, for example, a magnetic disk, an optical disk, a magneto-optical disk, or a semiconductor memory. Specifically, the removable recording medium 925 may be a CD medium, a DVD medium, a Blu-ray (registered trademark) medium, a CompactFlash (registered trademark) (CF), a flash memory, an SD memory card (Secure Digital memory card), or the like. Also, the removable recording medium 925 may be, for example, an IC card (Integrated Circuit card) or an electronic device equipped with a non-contact type IC chip.

[0102] The connection port 917 is a port for directly connecting a device to the information processing apparatus 900. The connection port 917 is, for example, a USB (Universal Serial Bus) port, an eSATA (external Serial Advanced Technology Attachment), a SAS (Serial Attached SCSI (Small Computer System Interface)) port, or the like. The information processing apparatus 900 can directly acquire various data from the external device 927 connected to the connection port 917 or provide various data to the external device 927. For example, an alarm notification device such as a rotating light for notifying alarm information may be connected via the connection port 917. Further, as the external device 927, a NAS (Network Attached Storage) may be connected and used as a storage device.

[0103] The communication device 919 is, for example, a communication interface composed of a communication device or the like for connecting to the communication network 929. The communication device 919 is, for example, a communication card for wired or wireless LAN (Local Area Network), Bluetooth (registered trademark), or WUSB (Wireless USB). Further, the communication device 919 may be an optical communication router, a router for ADSL (Asymmetric Digital Subscriber Line), or a modem for various communications. The communication device 919 can transmit and receive signals or the like in accordance with a predetermined protocol such as TCP / IP, for example, between the Internet and other communication devices. For example, a computer for operating the information processing apparatus 900 can also be connected via the communication device 919. Further, the communication network 929 connected to the communication device 919 is composed of a network or the like connected by wire or wirelessly. For example, the communication network 929 is the Internet, a home LAN, infrared communication, radio wave communication, or satellite communication.

[0104] An example of the hardware configuration of the information processing apparatus 900 has been described above. Each of the above-described components may be configured using general-purpose members, or may be configured by hardware specialized for the functions of each component. The hardware configuration of the information processing apparatus 900 can be appropriately changed according to the technical level at the time of implementing the present disclosure.

[0105] As described above, the preferred embodiments of the present disclosure have been described in detail with reference to the accompanying drawings, but the present disclosure is not limited to such examples. It is obvious that those having ordinary knowledge in the technical field to which the present disclosure pertains can conceive of various modification examples or correction examples within the scope of the technical idea described in the claims, and it is naturally understood that these also belong to the technical scope of the present disclosure.

[0106] For example, the abnormality detection device may be installed in a plant system including a plant having a process with a plurality of process characteristics, rather than an abnormality detection system that detects abnormalities occurring in a process. The plant system includes plants with various operating conditions, such as a steel plant, a chemical plant, a power generation plant, a power plant, etc., and controls the process of the plant by a control device including a processor that controls the plant. The plant system can calculate the degree of abnormality or the degree of deviation by the abnormality detection device. In the plant system, the plant is controlled based on the calculated degree of abnormality or the degree of deviation.

[0107] For example, the process data changes by changing the operating conditions of the plant, such as the operation amount of the plant and the types of operating facilities. When the process data input to the abnormality detection device changes, the abnormality degree and the deviation degree output by the abnormality detection device based on this change. Therefore, according to the present disclosure, the abnormality degree can be reduced by appropriately changing the operating conditions of the plant under equipment constraints. Further, from the deviation degree output from the abnormality detection device, the operating conditions of the plant that are factors increasing the abnormality degree can be specified. By appropriately changing the operating conditions that are factors increasing the specified abnormality degree under equipment constraints, it becomes possible to reduce the deviation degree, and as a result, it becomes possible to reduce the abnormality degree.

[0108] In this way, the plant system controls the plant based on the abnormality degree or the deviation degree calculated by the abnormality detection device. For example, the plant system changes the operating conditions of the process, such as changing the operating equipment, to control the plant so that the process operates normally. Thereby, damage to the plant can be prevented.

[0109] Note that the following configurations are also included in the technical scope of the present disclosure. (1) A process data acquisition unit that acquires process data of a process having a plurality of process characteristics, Based on the acquired process data and the abnormality degree indicating the difference from the normal state of the process represented by the feature information representing the characteristics of the normal process data for each process characteristic, an abnormality detection unit that detects the abnormal state of the process, Comprising The feature information representing the normal state of the process is Candidate feature information generated for each process characteristic from a plurality of normal process data acquired when the process is in a normal state is searched by evaluating based on the error between the estimated process data calculated using the candidate feature information and the normal process data, and is the optimal feature information for each process characteristic, an abnormality detection device. (2) The abnormality detection unit Factorize the matrix representing the obtained process data with a feature matrix representing the feature information representing the characteristics of the normal process data of the process to obtain a coefficient matrix representing the feature information of the process data. The abnormality detection device according to (1) above, which calculates the degree of abnormality based on the matrix representing the process data, the feature matrix, and the coefficient matrix. (3) The degree of abnormality is the distance between the value of the process data and the hyperplane representing the normal state of the process by the feature information representing the characteristics of the normal process data for each process characteristic. The abnormality detection device according to (1) or (2) above. (4) The abnormality detection unit calculates the degree of deviation representing the degree of abnormality for each operating condition of the process data. The abnormality detection device according to (3) above. (5) The abnormality detection device includes a processing unit for generating feature information representing the normal state of the process. The processing unit A feature information generation unit that generates candidates for feature information representing the characteristics of the normal process data for each process characteristic from a plurality of normal process data obtained when a process having a plurality of process characteristics is in a normal state. A search unit that evaluates the candidates for the feature information based on the error between the estimated process data calculated using the candidates for the feature information and the normal process data, and searches for the optimal feature information for each of the process characteristics. The abnormality detection device according to any one of (1) to (4) above, which has the above. (6) The search unit Calculates an evaluation threshold from the quartiles of the error. The abnormality detection device according to (5) above, which evaluates the feature information based on the number of the normal process data exceeding the evaluation threshold. (7) The search unit searches for the feature information that minimizes the number of the normal process data exceeding the evaluation threshold as the optimal feature information. The abnormality detection device according to (6) above. (8) A plant system comprising a plant having a process with a plurality of process characteristics and the abnormality detection device according to any one of (1) to (7) above, wherein the abnormality detection device calculates the degree of abnormality or the degree of deviation representing the degree of abnormality for each operating condition of the process data, and controls the plant based on the degree of abnormality or the degree of deviation. (9) A process data acquisition step of acquiring process data of a process having a plurality of process characteristics, and an abnormality detection step of detecting an abnormal state of the process based on the acquired process data and the degree of abnormality indicating a difference from the normal state of the process represented by the feature information representing the characteristics of the normal process data for each process characteristic. including wherein the feature information representing the normal state of the process is the optimal feature information for each of the process characteristics, which is searched by evaluating candidates for the feature information generated for each process characteristic from a plurality of normal process data acquired when the process is in a normal state based on the error between the estimated process data calculated using the candidates for the feature information and the normal process data. (10) In the abnormality detection step, a matrix representing the acquired process data is factorized by a feature matrix representing the feature information of the normal process data of the process to obtain a coefficient matrix representing the feature information of the process data, and the degree of abnormality is calculated based on the matrix representing the process data, the feature matrix, and the coefficient matrix. The abnormality detection method according to (9) above. (11) The degree of abnormality is the distance between the value of the process data and a hyperplane representing the normal state of the process by the feature information representing the characteristics of the normal process data for each process characteristic. The abnormality detection method according to (9) or (10) above. (12) In the abnormality detection step, the abnormality degree is calculated as the degree of deviation representing the abnormality degree for each operating condition of the process data, and the abnormality detection method according to (11) above. (13) As a process for generating characteristic information representing the normal state of the process, From a plurality of normal process data obtained when a process having a plurality of process characteristics is in a normal state, for each process characteristic, a characteristic information candidate representing the characteristics of the normal process data is generated, a characteristic information generation step; An exploration step of evaluating the characteristic information candidates based on the error between the estimated process data calculated using the characteristic information candidates and the normal process data, and searching for the optimal characteristic information for each of the process characteristics; The abnormality detection method according to any one of (9) to (12) above, including (14) In the exploration step, An evaluation threshold is calculated from the quartiles of the error, Based on the number of normal process data exceeding the evaluation threshold, the characteristic information is evaluated, and the abnormality detection method according to (13) above. (15) In the exploration step, as the optimal characteristic information, the characteristic information for which the number of normal process data exceeding the evaluation threshold is minimized is searched, and the abnormality detection method according to (14) above. (16) A computer, A process data acquisition unit that acquires process data of a process having a plurality of process characteristics, Based on the acquired process data and the degree of abnormality indicating the difference from the normal state of the process represented by the characteristic information representing the characteristics of the normal process data for each process characteristic, an abnormality detection unit that detects the abnormal state of the process, Function as an abnormality detection device including The characteristic information representing the normal state of the process is A program that searches for optimal feature information for each of the process characteristics by evaluating candidates for the feature information generated for each of the process characteristics from a plurality of normal process data obtained when the process is in a normal state based on the error between the estimated process data calculated using the candidates for the feature information and the normal process data. (17) The abnormality detection unit Factorizes a matrix representing the acquired process data with a feature matrix representing the feature information representing the features of the normal process data of the process to obtain a coefficient matrix representing the feature information of the process data, The program according to (16) above, which calculates the degree of abnormality based on the matrix representing the process data, the feature matrix, and the coefficient matrix. (18) The degree of abnormality is the distance between the value of the process data and the hyperplane representing the normal state of the process by the feature information representing the features of the normal process data for each process characteristic. The program according to (16) or (17) above. (19) The abnormality detection unit calculates the degree of deviation representing the degree of abnormality for each operating condition of the process data. The program according to (18) above. (20) Function the computer as an abnormality detection device including a processing unit for generating feature information representing the normal state of the process, The processing unit A feature information generation unit that generates candidates for feature information representing the features of the normal process data for each process characteristic from a plurality of normal process data obtained when a process having a plurality of process characteristics is in a normal state, A search unit that evaluates the candidates for the feature information based on the error between the estimated process data calculated using the candidates for the feature information and the normal process data, and searches for the optimal feature information for each of the process characteristics, The program according to any one of (16) to (19) above, comprising (21) The search unit The program according to (20) above, which calculates an evaluation threshold from the quartiles of the error, and evaluates the feature information based on the number of the normal process data exceeding the evaluation threshold. (22) The program according to (21) above, wherein the search unit searches for the feature information as the optimal feature information, in which the number of the normal process data exceeding the evaluation threshold is minimized.

Explanation of Signs

[0110] 1 Abnormality detection system 10 Process 30 Process data storage unit 100 Feature information generation device 110 Feature information generation unit 120 Search unit 200 Abnormality detection device 210 Process data acquisition unit 220 Abnormality detection unit 300 Feature information storage unit 500 Terminal 900 Information processing device 901 CPU 903 ROM 905 RAM 907 Bus 909 Input I / F 911 Output I / F 913 Storage device 915 Drive 917 Connection port 919 Communication device 921 Input device 923 Output device 925 Removable recording medium 927 External device 929 Communication network

Claims

1. A process data acquisition unit that acquires process data of a process having process characteristics representing the trend of a plurality of process data, Based on the degree of abnormality indicating the distance from a hyperplane indicating the boundary of the space representing the normal state of the process, which is generated by characteristic information representing the characteristics of normal process data for each of the process characteristics in the acquired process data, an abnormality detection unit that detects an abnormal state of the process, Comprising, The characteristic information representing the normal state of the process is, Candidate characteristic information generated for each of the process characteristics from a plurality of normal process data acquired when the process is in a normal state is searched by evaluating based on the error between the estimated process data calculated using the candidate characteristic information and the normal process data, and is the optimal characteristic information for each of the process characteristics, an abnormality detection device.

2. The abnormality detection unit, Factorizes the matrix representing the acquired process data with a characteristic matrix representing the characteristics of the normal process data of the process to obtain a coefficient matrix representing the characteristic information of the process data, The abnormality detection device according to claim 1, wherein the degree of abnormality is calculated based on the matrix representing the process data, the characteristic matrix, and the coefficient matrix.

3. The degree of abnormality is the distance between the value of the process data and a hyperplane representing the normal state of the process by characteristic information representing the characteristics of normal process data for each of the process characteristics, the abnormality detection device according to claim 1 or 2.

4. The abnormality detection unit calculates a degree of deviation representing the degree of abnormality for each operating condition of the process data, the abnormality detection device according to claim 3.

5. A plant system comprising a plant having a process having process characteristics representing the trend of a plurality of process data and the abnormality detection device according to claim 1, The abnormality detection device calculates the degree of abnormality or the degree of deviation representing the degree of abnormality for each operating condition of the process data, A plant system that controls the plant based on the degree of abnormality or the degree of deviation.

6. A characteristic information generation unit that generates candidates for characteristic information representing the characteristics of the normal process data for each of the process characteristics from a plurality of normal process data acquired when a process having process characteristics representing the trend of a plurality of process data is in a normal state, A search unit that evaluates candidates for the characteristic information based on the error between the estimated process data calculated using the candidates for the characteristic information and the normal process data, and searches for the optimal characteristic information for each of the process characteristics; comprising; A characteristic information generation device in which a space representing the normal state of the process is generated from the characteristic information. **Claim 7** The search unit calculates an evaluation threshold from the quartiles of the error, The characteristic information generation device according to claim 6, wherein the characteristic information is evaluated based on the number of the normal process data exceeding the evaluation threshold. **Claim 8** The search unit searches for the characteristic information that minimizes the number of the normal process data exceeding the evaluation threshold as the optimal characteristic information. The characteristic information generation device according to claim 7. **Claim 9** A process data acquisition step of acquiring process data of a process having a process characteristic representing the tendency of a plurality of process data; An abnormality detection step of detecting an abnormal state of the process based on an abnormality degree indicating a distance from a hyperplane indicating a boundary of a space representing the normal state of the process, which is generated by characteristic information representing characteristics of the normal process data for each of the process characteristics in the acquired process data; including; The characteristic information representing the normal state of the process is the optimal characteristic information for each of the process characteristics, which is searched by evaluating candidates for the characteristic information generated for each of the process characteristics from a plurality of normal process data acquired when the process is in a normal state, based on the error between the estimated process data calculated using the candidates for the characteristic information and the normal process data. An abnormality detection method. **Claim 10** A characteristic information generation step of generating candidates for characteristic information representing the characteristics of the normal process data for each of the process characteristics from a plurality of normal process data acquired when a process having a process characteristic representing the tendency of a plurality of process data is in a normal state; A search step of evaluating the candidates for the characteristic information based on the error between the estimated process data calculated using the candidates for the characteristic information and the normal process data, and searching for the optimal characteristic information for each of the process characteristics; including; A space representing the normal state of the process is generated from the characteristic information. A characteristic information generation method. **Claim 11** A computer A process data acquisition unit that acquires process data of a process having process characteristics representing the trends of a plurality of process data; Based on the degree of abnormality indicating the distance from a hyperplane indicating the boundary of a space representing the normal state of the process, generated by feature information representing the characteristics of normal process data for each of the process characteristics in the acquired process data, an abnormality detection unit that detects an abnormal state of the process; Function as an abnormality detection device comprising; The feature information representing the normal state of the process is; A program that is the optimal feature information for each of the process characteristics, searched by evaluating candidates for the feature information generated for each of the process characteristics from a plurality of normal process data acquired when the process is in a normal state, based on the error between the estimated process data calculated using the candidates for the feature information and the normal process data.

12. A computer, A feature information generation unit that generates candidates for feature information representing the characteristics of the normal process data for each of the process characteristics from a plurality of normal process data acquired when a process having process characteristics representing the trends of a plurality of process data is in a normal state; A search unit that evaluates the candidates for the feature information based on the error between the estimated process data calculated using the candidates for the feature information and the normal process data, and searches for the optimal feature information for each of the process characteristics; Comprising; A space representing the normal state of the process is generated from the feature information, A program that functions as a feature information generation device.

Images