阅读说明：本技术 异常检测装置以及异常检测方法 (Abnormality detection device and abnormality detection method ) 是由 西纳修一 矢崎彻 于 2019-11-29 设计创作，主要内容包括：异常检测装置具备：模型存储部,存储将成为基准的时间序列数据状态模型化的标准时间序列模型数值参数和将各状态的持续时间模型化的标准过渡时间模型数值参数；时间序列数据输入部,接受成为异常检测的对象的时间序列数据的输入；状态估计部,对成为异常检测的对象的时间序列数据进行基于标准时间序列模型数值参数的状态估计,来估计各数据点的状态变量；异常度计算部,使用标准过渡时间模型数值参数和状态估计部估计出的各数据点的状态变量来算出各数据点当中从给定的状态向其他状态的过渡所需的所需时间的异常度；和异常度输出部,将异常度作为每个经过时间的延迟度,进行图表化并输出。能在时间序列数据上进行考虑了所需时间异常的异常检测。(The abnormality detection device includes: a model storage unit that stores a standard time-series model numerical parameter for modeling a reference time-series data state and a standard transition time model numerical parameter for modeling a duration of each state; a time-series data input unit that receives input of time-series data to be detected for an abnormality; a state estimation unit that performs state estimation based on a standard time series model numerical parameter on time series data to be detected for an abnormality, and estimates a state variable of each data point; an abnormality degree calculation unit that calculates an abnormality degree of a time required for transition from a given state to another state among the data points using the standard transition time model numerical parameter and the state variables of the data points estimated by the state estimation unit; and an abnormality degree output unit that graphs and outputs the abnormality degree as a delay degree for each elapsed time. Abnormality detection can be performed on time-series data in consideration of a time abnormality required.)

1. An abnormality detection device is characterized by comprising:

a model storage unit that stores a standard time-series model numerical parameter for modeling a reference time-series data state and a standard transition time model numerical parameter for modeling a duration of each state;

a time-series data input unit that receives input of time-series data to be detected for an abnormality;

a state estimation unit that performs state estimation based on the standard time series model numerical parameter on time series data to be subjected to the abnormality detection, and estimates a state variable of each data point;

an abnormality degree calculation section that calculates an abnormality degree of a time required for transition from a given state to another state among the data points using the standard transition time model numerical parameter and the state variables of the data points estimated by the state estimation section; and

and an abnormality degree output unit that graphs and outputs the abnormality degree as a delay degree for each elapsed time.

2. The abnormality detection device according to claim 1,

the abnormality detection device includes:

a time stamp acquisition unit that acquires a time stamp from time-series data to be subjected to the abnormality detection,

the abnormality degree calculation unit calculates the degree of abnormality of the time required for transition from each state to another state using the standard transition time model numerical parameter, the state variable of each data point estimated by the state estimation unit, and the time stamp.

3. The abnormality detection device according to claim 1,

a plurality of the standard time series model numerical parameters and the plurality of the standard transition time model numerical parameters are stored in the model storage unit,

the state estimation unit estimates the state variable by combining results of state estimation based on a plurality of the standard time-series model numerical parameters.

4. The abnormality detection device according to claim 1,

the abnormality detection device includes:

and a standard time-series model estimating unit that estimates numerical parameters of one or more standard time-series models using the forwarded one or more standard time-series data, and stores the numerical parameters as the standard time-series model numerical parameters in the model storage unit.

5. The abnormality detection device according to claim 1,

the abnormality detection device includes:

and a standard transition time model estimating unit that calculates a transition time from a specific state to a transition to a next state in the one or more pieces of standard time-series data that have been forwarded, and stores the calculated transition time in the model storage unit as the standard transition time model numerical parameter.

6. The abnormality detection device according to claim 1,

the time-series data input unit receives input of time-series data output from a sensor that obtains a change amount of a position.

7. The abnormality detection device according to claim 1,

the abnormality detection device includes:

a feature value extracting unit for analyzing the moving image data to perform bone recognition,

the time-series data input unit receives input of moving image data output from a camera,

the state estimating unit estimates a state using, as time-series data, changes in joint positions of the bone obtained as a result of the analysis of the moving image data by the feature extraction unit.

8. The abnormality detection device according to claim 1,

the time-series data input unit receives input of time-series data output from a sensor that obtains a current value.

9. The abnormality detection device according to claim 1,

the abnormality degree output unit stores standard identification information including an identifier of an event occurring in the reference time-series data and time information on an occurrence time point and a duration of the event in advance,

in the graphing, an identifier of the respective instance is mapped to the given state.

10. The abnormality detection device according to claim 1,

the abnormality degree output unit displays an elapsed time during which a peak of the delay degree becomes large in association with the graph.

11. The abnormality detection device according to claim 1,

the abnormality degree output unit indicates a comment of occurrence of a delay at a position of an elapsed time at which a peak of the delay degree of the graph becomes large.

12. The abnormality detection device according to claim 1,

the abnormality degree output unit creates the graph for the abnormality degree amount, and the amount of delay degree is indicated by geometric information on the graph.

13. An abnormality detection method using an information processing apparatus, the abnormality detection method characterized in that,

the information processing device is provided with:

a model storage unit that stores a standard time-series model numerical parameter for modeling a reference time-series data state and a standard transition time model numerical parameter for modeling a duration of each state; and

a control part for controlling the operation of the display device,

the control unit performs:

a time-series data input step of receiving input of time-series data to be detected for an abnormality;

a state estimation step of performing state estimation based on the standard time series model numerical parameter on time series data to be detected for the abnormality, and estimating a state variable of each data point;

an abnormality degree calculation step of calculating an abnormality degree of a time required for transition from a given state to another state among the data points using the standard transition time model numerical parameter and the state variables of the data points estimated in the state estimation step;

and an abnormality degree output step of graphing and outputting the abnormality degree as a delay degree for each elapsed time.

Technical Field

The present invention relates to an abnormality detection device and an abnormality detection method.

Background

Patent document 1 discloses "an abnormality detection method including: a distance calculating step of calculating a distance between the standard time series data and the target time series data by arranging the standard time series data calculated in the standard time series data calculating step on one axis and arranging the target time series data acquired in the abnormality detection target data acquiring step on a DTW label on the other axis; and an abnormal data point detection step of detecting, as abnormal, a point in the time-series data at which the distance calculated in the distance calculation step is greater than a predetermined reference.

Disclosure of Invention

An object of the present invention is to provide a technique capable of detecting an abnormality in time-series data in consideration of a time abnormality required for delay, advance, or the like.

The present application includes various means for solving at least part of the above problems, and examples thereof are as follows.

One aspect of the present invention is an abnormality detection device including: a model storage unit that stores a standard time-series model numerical parameter for modeling a reference time-series data state and a standard transition time model numerical parameter for modeling a duration of each state; a time-series data input unit that receives input of time-series data to be detected for an abnormality; a state estimation unit that performs state estimation based on the standard time series model numerical parameter on time series data to be subjected to the abnormality detection, and estimates a state variable of each data point; an abnormality degree calculation unit that calculates an abnormality degree of a time required for transition from a predetermined state to another state among the data points, using the standard transition time model numerical parameter and the state variables of the data points estimated by the state estimation unit; and an abnormality degree output unit that graphs and outputs the abnormality degree as a delay degree for each elapsed time.

ADVANTAGEOUS EFFECTS OF INVENTION

According to the present invention, it is possible to provide a technique capable of performing abnormality detection in consideration of an abnormality in a required time on time-series data.

Problems, structures, and effects other than those described above will become apparent from the following description of the embodiments.

Drawings

Fig. 1 is a diagram schematically showing an example of a required time abnormality detection unit in the first embodiment.

Fig. 2 is a diagram showing an example of the initial probability table of the standard time series model numerical value parameter storage unit.

Fig. 3 is a diagram showing an example of an output probability table of the standard time-series model numerical parameter storage unit.

Fig. 4 is a diagram showing an example of a transition probability table in the standard time-series model numerical parameter storage unit.

Fig. 5 is a diagram showing an example of a transition time table in the standard transition time model numerical parameter storage unit.

Fig. 6 is a diagram showing an example of the abnormality degree storage table of the abnormality degree calculation unit.

Fig. 7 is a diagram showing an example of the effect of the smoothing processing.

Fig. 8 is a diagram showing an example of an abnormality degree output screen (peak list).

Fig. 9 is a diagram showing an example of an abnormality degree output screen (delay time point estimation).

Fig. 10 is a diagram showing an example of an abnormality degree output screen (degree of delay).

Fig. 11 is a diagram schematically showing an example of the required time abnormality detection unit in the second embodiment.

Fig. 12 is a diagram schematically showing an example of the required time abnormality detection unit in the third embodiment.

Fig. 13 is a diagram showing an example of the result of the matching operation.

Fig. 14 is a diagram showing an example of a required time abnormality detection system in the fourth embodiment.

Fig. 15 is a diagram showing a configuration example of a work abnormality detection system according to the fourth embodiment.

Fig. 16 is a diagram showing an example of the required time abnormality detection system in the fifth embodiment.

Fig. 17 is a diagram showing an example of the bone identification process in the fifth embodiment.

Fig. 18 is a diagram showing an example of a required time abnormality detection system in the sixth embodiment.

Fig. 19 is a diagram showing a configuration example of an abnormality degree output unit in the seventh embodiment.

Fig. 20 is a diagram showing an example of an abnormality degree output screen (job collation).

Description of reference numerals

100 abnormality degree calculation unit

101 state estimating unit

102 model storage unit

103 time-series data input unit

104 abnormality degree output unit

105 time stamp obtaining part

108 required time abnormality detection unit.

Detailed Description

In the following embodiments, the description will be made by dividing the embodiments into a plurality of parts or embodiments as necessary for convenience, and unless otherwise stated, these are not irrelevant to each other, and one part is in a relationship such as a modification, a detailed description, a supplementary description, or the like of a part or all of the other part.

In the following embodiments, when the number of elements and the like (including the number, numerical value, amount, range and the like) are mentioned, the number is not limited to a specific number unless otherwise specified or clearly limited to a specific number in principle, and may be equal to or greater than the specific number or may be equal to or less than the specific number.

Further, in the following embodiments, it is needless to say that the constituent elements (including the element steps) are not necessarily essential except for cases where they are explicitly shown in particular and cases where they are clearly considered to be essential in principle.

Similarly, in the following embodiments, when referring to the shape, positional relationship, and the like of the constituent elements and the like, the shape and the like substantially include the similar or analogous shape and the like, except for the case where the shape is specifically indicated and the case where it is considered that the shape is not clearly indicated in principle. The same applies to the above-mentioned values and ranges.

In the drawings for explaining the embodiments, the same members are denoted by the same reference numerals in principle, and redundant explanations thereof are omitted. Embodiments of the present invention will be described below with reference to the drawings.

DTW (Dynamic Time Warping) can be characterized by a generic concept of matching operation. In the present embodiment, a method of performing time scaling of time-series data and calculating a matching point is used as a matching calculation. The matching calculation can also obtain the matching point when the rates of change in the standard time-series data and the target time-series data differ with time scaling.

In general, standard time series data and matching operations can be characterized by a generic concept of standard time series models and state estimates, respectively. Here, the state estimation is an arithmetic method of estimating a state variable corresponding to each data point using numerical parameters with time series data as input, and the numerical parameters and the arithmetic method used are used as a standard time series model. The matching operation treats each data point of the standard time-series data as a different state, and can be considered as one of state estimations in the sense that each data point of given target time-series data is matched with any one of the data points. However, if an abnormality such as a delay is observed during the transition time, it cannot be detected as an abnormality.

Therefore, in the embodiment according to the present invention, the standard transient time model that represents the standard transient time of each state is newly introduced. For an arbitrary section from the data point a to the data point B for calculating the degree of abnormality of the target time-series data, a transition time distribution or a representative value thereof for the section is calculated from a state included in the target section obtained from the result of state estimation for the target time-series data and a transition time distribution of each state obtained with reference to a standard transition time model. The degree of abnormality is calculated as the degree of abnormality, such as the degree to which the section transition time from the data point a to the data point B in the target time-series data does not fit the section transition time distribution, or the difference between the transition time and the representative value of the section transition time distribution.

First embodiment fig. 1 is a diagram schematically showing an example of a required time abnormality detection unit according to a first embodiment. Here, the time-series data received by the time-series data input unit 103 is data in which a plurality of data points are connected in chronological order from the time point at which the data point is obtained. A variable representing the time point at which the data point was obtained is set as a time stamp, and a variable representing the value measured at the data point is set as a data value. The sequence in which the data values and the time stamps are arranged in time series is defined as a data value sequence and a time stamp sequence, and time series data can be held and expressed by both of them. However, in the expression of time-series data, it is not essential to use a time-stamp sequence, and when data points are sampled in accordance with a certain rule, the data points can be calculated later by holding the sampling rule instead of the time-stamp sequence. The data value includes not only a scalar but also a vector, an image, and the like.

In addition, information for modeling the state of a given single or multiple pieces of time-series data serving as a reference in advance is preset in the model storage unit 102. The predetermined method can be implemented by the method described in the second embodiment.

Fig. 2 is a diagram showing an example of the initial probability table of the standard time series model numerical value parameter storage unit. The initial probability table 1021 is a table indicating the probability of the initial state of the standard time-series model in a single time-series data. In the initial probability table 1021, a state 1021A and an initial probability 1021B are correspondingly included. That is, it can be said that the probability that the state is the initial state is stored for each state.

Fig. 3 is a diagram showing an example of an output probability table of the standard time-series model numerical parameter storage unit. The output probability table 1022 is a table indicating the output probability of the standard time-series model in a single time-series data. The output probability table 1022 is created to include a state 1022A, a mean 1022B, and a variance 1022C correspondingly. That is, it can be said that the distribution of values that can be taken in each state is stored for each state.

Fig. 4 is a diagram showing an example of a transition probability table in the standard time-series model numerical parameter storage unit. The transition probability table 1023 is a table indicating the probability of a transition between states of the standard time series model in a single time series data. The transition probability table 1023 is created to include probabilities of transition with respect to a combination of the original state 1023A and the transition target state 1023B. That is, the probability of transition from one state to another state is stored for each state.

Fig. 5 is a diagram showing an example of a transition time table in the standard transition time model numerical parameter storage unit. The transition time table 1024 is a table representing a distribution modeling of a duration in which a state lasts for each state in a single time series data having a standard time series model. A transition schedule 1024 is established that correspondingly contains states 1024A, mean 1024B, and variance 1024C. That is, it can be said that the distribution of values that can be taken for the duration of the state is stored for each state.

The required time abnormality detection unit 108 receives the time-series data from the time-series data input unit 103, and forwards information on the degree of abnormality to the abnormality degree output unit 104. The time-series data input unit 103 inputs time-series data to be subjected to abnormality detection. The required time abnormality detection unit 108 performs matching calculation of the time series data with the model information preset in the model storage unit 102, specifies the degree of abnormality, and forwards the degree of abnormality to the abnormality degree output unit 104. The abnormality degree output unit 104 performs output processing such as displaying the information of the abnormality degree or recording the information of the abnormality degree in a storage area. In this case, the abnormality degree output unit 104 may output a warning when the abnormality degree is larger than a specific reference.

The required time abnormality detection unit 108 includes an abnormality degree calculation unit 100, a state estimation unit 101, a model storage unit 102, and a time stamp acquisition unit 105.

The time stamp acquisition unit 105 acquires the time stamp sequence of the target time series data received from the time series data input unit 103, and sends the time stamp sequence to the abnormality degree calculation unit 100.

The state estimating unit 101 performs state estimation based on the standard time-series model on the time-series data received by the time-series data input unit 103 using the standard time-series model numerical value parameters (including the initial probability table 1021 in fig. 2, the output probability table 1022 in fig. 3, and the transition probability table 1023 in fig. 4) read out by the model storage unit 102, estimates a state variable of each data point, and sends the estimated state variable to the abnormality degree calculating unit 100. In the estimation of the state variable, the state estimation unit 101 uses a state estimation method corresponding to the standard time series model. For example, when a hidden markov model is used as the standard time series model, the state estimation unit 101 performs state estimation using the viterbi algorithm.

The state estimation unit 101 integrates the results of state estimation based on the plurality of standard time series model numerical parameters to estimate a state variable. For example, the state estimating unit 101 may estimate the mean value and the median value of a plurality of estimated state variables as the state variables, or may determine the mean value and the variance to estimate the probability.

The abnormality degree calculation unit 100 reads the standard transition time model numerical value parameter from the model storage unit 102, calculates the abnormality degree using the state variable of each data point sent from the state estimation unit 101 and the time stamp sequence received from the time stamp acquisition unit 105, and stores the abnormality degree in the abnormality degree storage table 1001.

The abnormality degree calculation unit 100 obtains a transition time T1 required to transition from the data point a to the data point B from the time stamp sequence for a target section from the data point a to the data point B for calculating the abnormality degree of the target time-series data, and obtains the state included in the target section from the result of state estimation for the target time-series data. Further, the abnormality degree calculation unit 100 derives or approximates a section transition time distribution, which is a distribution of transition times over the entire section, or a representative value T2 thereof, from the transition time distribution of each state or a representative value thereof (an average value, a mode value, or the like) obtained by referring to the standard transition time model. The abnormality degree calculation unit 100 calculates a difference between the transition time T1 and the representative value T2 or between the transition time T1 and the section transition time distribution as the abnormality degree.

As an approximation method of the interval transition time distribution, the abnormality degree calculation unit 100 adopts a sampling-based method. As a method of approximating the representative value, the abnormality degree calculation unit 100 adopts a method of accumulating sums when the representative value is an average value.

As a method of calculating the difference between the section transition time distribution and the transition time T1, the following method is adopted: the degree to which the transition time T1 is suitable for the section transition time distribution is evaluated by the likelihood or the like, and the result of the inversion, the inversion of the sign, or the like is used as the degree of abnormality.

As a method of calculating the difference between the transition time T1 and the representative value T2, for example, a method of dividing the transition time T1 by the representative value T2 is employed. In this method, when the degree of abnormality, which is a quotient resulting from the division, is greater than "1", it is highly likely that a delay occurs in the target time-series data during the transition from the data point a to the data point B or a situation different from the standard time-series data is inserted between the data point a and the data point B. That is, it is highly likely that the operation of the job is delayed or another operation of the job is performed during the operation of the job. On the other hand, when the degree of abnormality is less than "1", it indicates that the change from the data point a to the data point B in the target time-series data is early or that there is a high possibility that the event existing in the standard time-series data is missing between the data point a and the data point B. That is, it indicates that the task operation is more likely to be performed earlier than usual or the task operation is omitted.

In addition, although the above description shows an example of calculating the degree of abnormality (degree of delay or degree of advance) between arbitrary 2 points (between the data point a and the data point B), the degree of abnormality calculation unit 100 is not limited to this, and may calculate the degree of abnormality as a sequence by repeating the degree of abnormality calculation for 2 points at a sufficiently short time interval on the target time-series data.

Fig. 6 is a diagram showing an example of the abnormality degree storage table of the abnormality degree calculation unit. The abnormality degree storage table 1001 includes a time point 1001A as information for specifying a time point and an abnormality degree 1001B associated therewith.

The abnormality degree calculation unit 100 may perform post-processing such as smoothing on the calculated abnormality degree sequence. A known method such as a moving average method is used as an algorithm for smoothing.

Fig. 7 is a diagram showing an example of the effect of the smoothing processing. In fig. 7, before smoothing is represented by dotted lines, and after smoothing is represented by solid lines. For example, when the time-series data has an influence of noise or the like, the variation of the degree of abnormality becomes large. In such a case, information in which the degree of abnormality of the variation is suppressed as shown in fig. 7 can be obtained by performing smoothing.

The abnormality degree calculation unit 100 stores the calculated sequence of the abnormality degree or the information obtained by smoothing the calculated sequence of the abnormality degree in the abnormality degree storage table 1001, and passes the abnormality degree to the abnormality degree output unit 104.

Fig. 8 is a diagram showing an example of an abnormality degree output screen (peak list). The abnormality degree output screen (peak list) 1041 is screen information created by the abnormality degree output unit 104 by graphing a sequence of the abnormality degrees or information obtained by smoothing the sequence. The abnormality degree output screen (peak list) 1041 includes: a delay degree display area 1041A that displays a delay degree chart indicating the magnitude of the delay degree (abnormality degree) generated in each product; a peak list 1041B that displays, for each product, time points at which the peak of the delay degree is obtained in descending order of the delay degree; and a confirm button 1041C that accepts an instruction to end the screen display. From this picture, the user can compare the peak value of the delay degree for each product. The abnormality degree output unit 104 may generate and output the following output screen.

Fig. 9 is a diagram showing an example of an abnormality degree output screen (delay time point estimation). The abnormality degree output screen (delay time point estimation) 1042 is screen information created by the abnormality degree output unit 104 by graphing a sequence of the abnormality degrees or information obtained by smoothing the sequence. The abnormality degree output screen (delay time point estimation) 1042 includes: a delay degree display region 1042A that displays a delay degree chart indicating the magnitude of a delay degree (abnormality degree) generated in a product; a comment 1042B showing that a delay occurs at a point (peak) at which a degree of delay is detected to be particularly large; and a confirmation button 1042C for accepting an instruction to end the screen display. From this screen, the user can know the time point at which a large delay occurs with respect to the target product, and obtain clues to improve the contents of the job.

Fig. 10 is a diagram showing an example of an abnormality degree output screen (degree of delay). The abnormality degree output screen (delay degree) 1043 is screen information created by the abnormality degree output unit 104 by graphing a sequence of the abnormality degrees or information obtained by smoothing the sequence. The abnormality degree output screen (delay degree) 1043 includes: a delay degree display area 1043A that displays a delay degree chart indicating the magnitude of the delay degree (abnormality degree) occurring in the product; an annotation 1043B that displays the time zone in which the degree of delay is detected and the amount of delay in that time; and a confirm button 1043C that accepts an instruction to end the screen display. From this screen, the user can know the time point at which the delay continuously occurs with respect to the target product, and obtain a clue to improve the work content.

In order to generate the abnormality degree output screen (delay degree) 1043, it is necessary to specify the delay amount in the time zone in which the delay degree is detected. In this determination process, the abnormality degree calculation unit 100 determines the delay amount by calculating the difference (T1 to T2) between the transition time T1 and the representative value T2. Thus, the delay amount can be determined, and an abnormality degree output screen (delay degree) 1043 can be created. The user can also create a graph of the amount of abnormality and grasp the amount of delay from the geometric information on the graph (specifically, the area indicating the amount of delay).

Second embodiment fig. 11 is a diagram schematically showing an example of a required time abnormality detection unit in the second embodiment. The required time abnormality detection unit 108' according to the second embodiment is basically the same as the required time abnormality detection unit 108 according to the first embodiment, but is partially different. The following description will be made centering on differences from the required time abnormality detection unit according to the first embodiment.

In the second embodiment, the standard model stored in the model storage unit 102 can be learned from single or multiple pieces of standard time-series data obtained from the time-series data input unit 103.

The required time abnormality detection unit 108' according to the second embodiment further includes a standard time-series model estimation unit 106 and a standard transition time model estimation unit 107.

The standard time-series data input from the time-series data input unit 103 is forwarded to the standard time-series model estimation unit 106, the state estimation unit 101, and the time stamp acquisition unit 105. The time stamp obtaining unit 105 calculates or obtains a time stamp sequence, and passes the time stamp sequence to the standard transition time model estimating unit 107.

The standard time series model estimation unit 106 estimates numerical parameters of one or more standard time series models using the transferred one or more standard time series data, and stores the numerical parameters in the model storage unit 102. The learning of the standard time series model estimates numerical parameters using AI (artificial intelligence) and the like such as a framework of machine learning such as a hidden markov model and a probabilistic automaton. For example, when the hidden markov model is used, the initial probability, the transition probability, and the output probability are estimated as numerical parameters, and the model storage unit 102 stores the estimated numerical parameters (the initial probability table 1021, the output probability table 1022, and the transition probability table 1023) for each standard time series data.

The state estimation unit 101 performs state estimation using a standard time series model on a set of standard time series data input from the time series data input unit 103 using numerical parameters of the standard time series data read by the model storage unit 102, and estimates a state variable of each data point of the time series data. The state estimating unit 101 transfers the estimated state variable to the standard transition time model estimating unit 107. A state estimation method corresponding to a standard time series model is used for estimating the state variable. For example, in the case of hidden markov models, the state estimation can be performed using the viterbi algorithm. In addition, when a plurality of numerical parameters of the standard time-series data are used, since a plurality of states are estimated, the state estimation unit 101 may obtain the distribution thereof.

The standard transition time model estimation unit 107 receives the state variables of the data points from the state estimation unit 101, and receives the time stamps of the data points from the time stamp acquisition unit 105. By associating the state with the time stamp, the standard transition time model estimation unit 107 calculates a transition time from a specific state to a next state in the target time-series data.

Then, the standard transition time model estimation unit 107 calculates the transition time for all pairs of the plurality of standard transition time models of the state variables and the time stamps of the respective data points, thereby obtaining the distribution of the transition time for each state, and stores the distribution as the standard transition time model (transition time table 1024) of the model storage unit 102. The standard transient time model estimation unit 107 approximates the distribution of the transient time with a gaussian distribution or the like, and stores parameters (average, variance, and the like) thereof as a representative value. The present invention is not limited to this, and all the calculated transition times may be stored as they are.

In this way, a single or a plurality of standard models stored in the model storage unit 102 can be learned using the standard time-series data obtained from the time-series data input unit 103.

Third embodiment fig. 12 is a diagram schematically showing an example of a required time abnormality detection unit in the third embodiment. The required time abnormality detection unit 108 ″ according to the third embodiment is basically the same as the required time abnormality detection unit 108 according to the second embodiment, but is partially different. The following description will focus on differences from the required time abnormality detection unit according to the second embodiment.

In the third embodiment, the following is exemplified: the required time abnormality detection unit 108 ″ receives a single (a series of) standard time-series data and sets the standard time-series data as a standard model, thereby eliminating the need for the standard time-series model estimation unit 106 and the standard transient time model estimation unit 107.

The required time abnormality detection unit 108 ″ according to the third embodiment includes an abnormality degree calculation unit 100, a state estimation unit 101, a model storage unit 102, and a time stamp acquisition unit 105.

The time-series data input unit 103 receives input of single standard time-series data. The time-series data input unit 103 sends a data value sequence in which data values are arranged in time series to the model storage unit 102, and sends a time stamp sequence to the time stamp acquisition unit 105.

The time stamp obtaining unit 105 obtains a time stamp sequence of the input standard time series data, and transfers the time stamp sequence to the model storage unit 102.

Then, the model storage unit 102 stores the data value sequence of the standard time-series data transferred from the time-series data input unit 103 and the time stamp sequence transferred from the time stamp acquisition unit 105. This becomes the flow of learning of a single standard model. Next, a flow of abnormality degree calculation will be described.

When the time-series data input unit 103 receives the target time-series data to be subjected to the abnormality detection, the time-series data input unit 103 sends the data value sequence to the state estimating unit 101, and sends the time stamp sequence or information necessary for calculating the time stamp sequence to the time stamp obtaining unit 105. The time stamp acquisition unit 105 acquires or calculates a time stamp sequence of the target time series data, and passes the time stamp sequence to the abnormality degree calculation unit 100.

The state estimating unit 101 performs a matching operation on the data value sequence of the target time-series data and the data value sequence of the standard time-series data read from the model storage unit 102, and delivers the result of the matching operation to the abnormality degree calculating unit 100.

Fig. 13 is a diagram showing an example of the result of the matching operation. This result is not limited to the result of the third embodiment, and is the same in all embodiments from the first embodiment to the seventh embodiment described later. The data value sequence of the target time-series data indicated by the solid line and the data value sequence of the standard time-series data are connected by a diagonal line. The point at which the connection is made is identified by a matching operation, and the state estimation unit 101 identifies the point by using a known algorithm such as DTW.

The calculation of the degree of abnormality by the abnormality degree calculation unit 100 is basically the same as in the first and second embodiments. In the configuration of the third embodiment, since a single standard time-series data is used as the numerical parameter of the standard time-series model, the standard time-series model estimation unit 106 that performs modeling using a plurality of standard time-series data as described in the second embodiment is not necessary. Further, since the state estimation is realized by treating each data point of the standard time-series data as a different state and associating each data point of the target time-series by the matching operation, the standard transition time of each state is a single value, and the difference between adjacent time stamps in the time stamp series becomes the same. Therefore, the standard transition time model estimation unit 107 is not required, and the time stamp sequence may be stored as it is as a standard transition time model. Therefore, the required time abnormality detection unit 108 ″ according to the third embodiment can be configured simply.

Fourth embodiment fig. 14 is a diagram showing an example of a required time abnormality detection system in the fourth embodiment. In the fourth embodiment, the work abnormality detection system 205 for detecting an abnormality in a manual work by a wearable sensor (a sensor capable of wearing and taking off a change in position) is used as an example of the required time abnormality detection unit.

The worker 201 attaches the wearable sensor 202 to the body and repeats the work. Sensor data of the activity of the worker 201 measured by the wearable sensor 202 is sent to the information processing terminal 203 via wireless communication. As the wearable sensor 202 for measuring the movement, a sensor capable of acquiring the amount of change in position, such as an acceleration sensor or a gyro sensor, can be used. Upon receiving the sensor data at the sensor receiving unit 204, the information processing terminal 203 inputs the sensor data to the work abnormality detection system 205.

Fig. 15 is a diagram showing a configuration example of a work abnormality detection system according to the fourth embodiment. The sensor receiving unit 204 sends the sensor data (time-series data) to the feature extraction unit 206, and the feature extraction unit 206 converts the sensor data into time-series data composed of a feature suitable for abnormality detection. For example, when there is a defect in the sensor data, the feature amount extraction unit 206 performs interpolation processing to remove noise included in the sensor data by a filter such as a low-pass filter, and thus can expect improvement in the accuracy of abnormality detection. The feature extraction unit 206 obtains a feature by performing fourier transform on the sensor data.

The time-series data from which the feature amount is extracted by the feature amount extraction unit 206 is forwarded to the sequence section extraction unit 207, and a section corresponding to one of the repetitive jobs is extracted based on a control signal from the extraction control unit 210, and forwarded to the required time abnormality detection unit 108.

As described in the first, second, and third embodiments, the required time abnormality detection unit 108 calculates the degree of abnormality for the received time-series data, and sends the calculation result to the degree of abnormality output unit 104. The abnormality degree output unit 104 records the abnormality degree and displays it on a display or the like. In this way, for example, by checking the degree of abnormality displayed on the screen, it is possible to determine whether or not the job is being performed normally. When the degree of abnormality is larger than a specific reference, a warning can be output to support the response of an operator, a supervisor, or the like.

Fifth embodiment fig. 16 is a diagram showing an example of a required time abnormality detection system in the fifth embodiment. In the fifth embodiment, an example is described in which a required time abnormality detection unit is operated for the job abnormality detection system 205 that performs abnormality detection of a manual job by a camera.

The operator 301 performs the repetitive operation. Moving image data of the operator 301 captured by the camera 302 is sent to the information processing terminal 303 via wired communication or wireless communication. The camera 302 is not limited to a camera that captures visible light, and may be a camera that captures invisible light such as infrared light, as long as it is a camera that arranges still images in time series to form a moving image.

Upon receiving the moving image data from the camera 302, the information processing terminal 303 inputs the sensor data to the job abnormality detection system 205.

The information processing terminal 303 can regard moving image data as time-series data in which frame data are arranged in time series. The work abnormality detection system 205 according to the present embodiment is basically the same as the work abnormality detection system 205 according to the fourth embodiment, but is partially different. The following description will focus on differences from the work abnormality detection system 205 according to the fourth embodiment.

The feature value extraction unit 206 extracts the joint position of the subject in the moving image data as a feature value. Specifically, the feature extraction unit 206 analyzes the moving image data to perform bone recognition. For example, the feature amount extraction unit 206 uses a bone recognition algorithm to recognize a bone.

Fig. 17 is a diagram showing an example of the bone identification process in the fifth embodiment. In the skeleton recognition processing, skeleton recognition is performed on each frame 310 of the moving image data of the worker 301, and the joint position of the worker is extracted (frame 311). By using the joint position as the feature value, the time-series data reflecting the movement of the operator 301 can be converted, and the abnormality detection becomes easy.

The process of the time-series data from which the feature amount is extracted being handed over to the sequence section clipping unit 207 is the same as the fourth embodiment.

According to the fifth embodiment, it is possible to determine whether or not a job is being performed at a normal speed by, for example, checking the degree of abnormality displayed on the screen. If the degree of abnormality is larger than a specific reference, a warning may be output, and an operator, a supervisor, or the like may support the operation while confirming the operation in the image captured by the camera 302, so as to facilitate the response.

Sixth embodiment fig. 18 is a diagram showing an example of a required time abnormality detection system in the sixth embodiment. The sixth embodiment is an example in which the required time abnormality detecting unit is operated for the work abnormality detecting system 205 that detects an abnormality in the work of the work machine by the current sensor.

The machine tool 501 is a machine that performs cutting, and the magnitude of the current value of the main shaft electric power line is a value corresponding to the operating state of the machine tool 501 such as cutting, idling, and stop. The current sensor 502 transmits the measured current value to the information processing terminal 503 via wired communication or wireless communication.

Upon receiving the time-series data of the current value from the current sensor 502, the information processing terminal 503 inputs sensor data to the work abnormality detection system 20.

The work abnormality detection system 205 according to the present embodiment is basically the same as the work abnormality detection system 205 according to the fourth embodiment, but is partially different. The following description will focus on differences from the work abnormality detection system 205 according to the fourth embodiment.

The feature value extraction unit 206 converts the sensor data into time-series data composed of feature values suitable for abnormality detection, with respect to the time-series data of the current values. For example, by removing noise included in the sensor data with a filter such as a low-pass filter, it is expected that the accuracy of abnormality detection will be improved. For example, when there is a defect in the sensor data, the feature amount extraction unit 206 performs interpolation processing to remove noise included in the sensor data by a filter such as a low-pass filter, and thus can expect improvement in the accuracy of abnormality detection. The feature extraction unit 206 obtains a feature by performing fourier transform on the sensor data.

The processing after the time-series data of the extracted feature amount is transferred to the sequence section clipping unit 207 is the same as the fourth embodiment.

According to the sixth embodiment, it is possible to determine whether or not the work machine 501 is performing a work operation at a normal speed, for example, by checking the degree of abnormality displayed on the screen. When the abnormality degree is larger than a specific reference, a warning may be output to assist in response to repair of the machine tool or provision of a state maintaining operation.

Fig. 19 is a diagram showing a configuration example of an abnormality degree output unit in the seventh embodiment. In the seventh embodiment, the output of the degree of abnormality is expanded. The seventh embodiment has basically the same configuration as that of the required time abnormality detection system according to any one of the first to sixth embodiments, but is partly different. The difference will be described below.

As shown in fig. 19, in the seventh embodiment, the abnormality degree output unit 104 includes a standard identification information 402 that holds an identifier of an event occurring in the standard time-series data and time information on the occurrence time point and the duration of the event, and the standard identification information 402 is connected to the abnormality degree display unit 401. For example, in many cases, in the examples of the jobs in the plant as shown in the fourth to sixth embodiments, the order of the jobs in the series of jobs and the respective standard job times are determined in advance, and the standard identification information 402 is stored in advance based on the determined order and standard job times.

When the information on the degree of abnormality at a specific time is passed from the required time abnormality detection unit 108, the abnormality degree display unit 401 refers to the standard identification information 402, searches the standard time series data for an identifier of a situation occurring at the corresponding time, and creates a display screen by associating the degree of abnormality with the identifier.

Fig. 20 is a diagram showing an example of an abnormality degree output screen (job collation). The abnormality degree output screen (work comparison) 1044 is screen information created by the abnormality degree output unit 104 by graphing a sequence of the abnormality degrees or information obtained by smoothing the sequence. The abnormality degree output screen (job comparison) 1044 includes: a delay degree display area 1044A that displays a delay degree chart indicating the magnitude of the delay degree (abnormality degree) occurring in the product; an identifier display area 1044B that displays the job identifier for each time point; and a confirmation button 1044C for accepting an instruction to end the screen display. From this screen, the user can know the job at the time point when the large delay occurs with respect to the target product, and get a clue to improve the contents of the job.

Here, the time information of the identifier held in the standard identification information 402 is only time information of the standard time at all times, and when a time abnormality occurs in the target time-series data, there is a possibility that the error is too large to be ignored. Therefore, the abnormality degree display unit 401 can calculate the time offset from the standard time in the target time-series data based on the abnormality degree, correct the time information, and use the identifier. Specifically, when the section from the start time to the specific elapsed time is divided into local sections in the abnormality degree calculation unit 100, and the abnormality degree of each local section AB is obtained by the difference between T1 and T2 (the definition of T1 and T2 is the same as that in the first embodiment), the accumulated value is calculated, whereby the time shift from the standard time to the specific time can be calculated, and the time information can be used for display after being corrected using the calculated value. The seventh embodiment is described above.

The present invention is not limited to the above-described embodiments, and includes various modifications. For example, the above-described embodiments are described in detail to facilitate understanding of the present invention, but the present invention is not limited to having all of the structures described. In addition, a part of the structure of one embodiment may be replaced with the structure of another embodiment, and the structure of one embodiment may be added to the structure of another embodiment.

Further, addition, deletion, and replacement of another configuration may be performed on a part of the configurations of the embodiments. The above-described structures, functions, processing units, and the like may be partially or entirely implemented in hardware by designing them with, for example, an integrated circuit. The above-described structures, functions, and the like may be implemented in software by a processor interpreting and executing a program for implementing the functions. Information such as programs, tables, and files for realizing the respective functions can be stored in a memory, a recording device such as a hard disk or an SSD (Solid State Drive), or a recording medium such as an IC card, an SD card, or a DVD (digital versatile Disc).

The control lines and the information lines are lines which are considered necessary for the description, and not all the control lines and the information lines are necessarily shown in the product. In practice, almost all structures can be considered to be interconnected.

The present invention is not limited to the abnormality detection unit, the abnormality detection device, and the abnormality detection method, and may be provided in various forms such as an abnormality detection system, a computer-readable program, and an image processing circuit.