WO2023127329A1

WO2023127329A1 - Abnormality detection device, method, and program

Info

Publication number: WO2023127329A1
Application number: PCT/JP2022/042183
Authority: WO
Inventors: 祥平山田; 学吉見; 伸一笠原
Original assignee: ダイキン工業株式会社
Priority date: 2021-12-28
Filing date: 2022-11-14
Publication date: 2023-07-06
Also published as: JP2023097730A; JP7231866B1

Abstract

The present invention improves the accuracy of abnormality detection for an apparatus. An abnormality detection device according to one embodiment of the present disclosure comprises a control unit, and is characterized in that the control unit: creates a N+1-dimensional topological space using a value at an arbitrary time and N values prior to the arbitrary time from time series data that is a target of abnormality detection in an air conditioner; creates a normal region and an abnormal region in the topological space according to a distribution when normal time series data and abnormal time series data have been mapped in the topological space; acquires at least N+1 values from the time series data that is the target of abnormality detection in the air conditioner; and specifies whether the at least N+1 values are mapped in the normal region or are mapped in the abnormal region, thereby determining and outputting whether the time series data is normal or abnormal.

Description

Anomaly detection device, method and program

The present disclosure relates to an anomaly detection device, method, and program.

Conventionally, in the anomaly detection of equipment such as air conditioners, there is a known method of judging an anomaly when a failure index exceeds a predetermined threshold. This threshold value is determined based on two distributions (probability distribution of normal data and probability distribution of abnormal data) by calculating the probability density of normal data and abnormal data of the indicator. there is

JP 2009-24923 A

However, the two distributions have a clearly normal range (a range that does not overlap with abnormality), a clearly abnormal range (range that does not overlap with normal), and a range that is difficult to distinguish between normal and abnormal (normal and abnormal). There are three types: the range covered) and . Therefore, if the time-series data of the failure index fluctuates due to noise and reaches a range where it is difficult to determine whether the failure is normal or abnormal, the threshold may be exceeded and an erroneous determination may occur. Description will be made below with reference to FIG. The left side of FIG. 1 explains the determination of the threshold, and the right side of FIG. 1 explains the detection of abnormality using the threshold.

The left side of FIG. 1 shows the probability density (or frequency) of the failure index (for example, the refrigerant amount index). “Normal distribution” is the probability density calculated by collecting the values of the refrigerant quantity index during normal times, and “Distribution at the time of leakage” collects the values of the refrigerant quantity index at abnormal times (at the time of refrigerant leakage). is the probability density calculated by A threshold value (broken line in FIG. 1) is determined based on the normal distribution and the leakage distribution. “Range where it can be correctly judged as normal” is a clearly normal range (range where it is not abnormal (leakage)), and “range where leakage can be correctly judged” is a clearly abnormal (leakage) range (normal) range), and there are few misjudgments. On the other hand, the “range of many misjudgments” is a range in which it is difficult to distinguish between normality and abnormality (leakage) (range where normality and abnormality (leakage) overlap), and there are many misjudgments.

The right side of FIG. 1 shows the value of the failure index (for example, refrigerant amount index) at each time (t). If the noise removal of the refrigerant amount index is insufficient, for example, even though the equipment is normal, the threshold value is exceeded due to noise, and an abnormality is erroneously detected.

The purpose of this disclosure is to improve the accuracy of device anomaly detection.

An abnormality detection device according to a first aspect of the present disclosure includes:
An anomaly detection device comprising a control unit,
The control unit
An N+1 dimensional phase space is created from the value of the time series data that is the abnormality detection target of the air conditioner and N values past the arbitrary time, and the normal time series data and the abnormality are created on the phase space. Create a normal region and an abnormal region on the phase space according to the distribution when the time series data of is mapped,
Acquiring at least N+1 values from the time-series data that is the abnormality detection target of the air conditioner, and specifying whether the at least N+1 values are mapped to the normal region or the abnormal region. , it is determined whether the time-series data is normal or abnormal, and output.

According to the first aspect of the present disclosure, erroneous detection due to noise can be reduced.

An abnormality detection device according to a second aspect of the present disclosure is the abnormality detection device according to the first aspect,
The normal region and the abnormal region on the phase space are learned models created by machine learning.

According to the second aspect of the present disclosure, normal regions and abnormal regions can be easily created.

An abnormality detection device according to a third aspect of the present disclosure is the abnormality detection device according to the first aspect or the second aspect,
The phase space is two-dimensional.

According to the third aspect of the present disclosure, it is possible to improve the accuracy of detecting an abnormality.

An abnormality detection device according to a fourth aspect of the present disclosure is the abnormality detection device according to any one of the first to third aspects,
The phase space is created from at least one of a value of time-series data as an abnormality detection target for the air conditioner and a difference between time-series data as an abnormality detection target for the air conditioner.

According to the fourth aspect of the present disclosure, the normal region and the abnormal region can be created in consideration of at least one of the time-series data that is the target of anomaly detection and the difference between the data that is the target of anomaly detection.

An abnormality detection device according to a fifth aspect of the present disclosure is the abnormality detection device according to any one of the first to fourth aspects,
The control unit
Abnormality of the time-series data is determined based on at least one of a degree of deviation from normal and a degree of change.

According to the fifth aspect of the present disclosure, it is possible to detect an anomaly in consideration of at least one of the degree of deviation of data, which is the target of anomaly detection, from normal and the degree of change of the data, which is the target of anomaly detection. can.

An abnormality detection device according to a sixth aspect of the present disclosure is the abnormality detection device according to any one of the first to fifth aspects,
The control unit
When the normal region and the abnormal region overlap, the region where the normal region and the abnormal region overlap is defined as the normal region.

According to the sixth aspect of the present disclosure, it is possible to reduce erroneous detection as abnormal even though it is normal.

An abnormality detection device according to a seventh aspect of the present disclosure is the abnormality detection device according to any one of the first to sixth aspects,
The control unit
A relationship between the number of data mapped to the normal region and the number of data mapped to the abnormal region among the at least N + 1 values, and the normal center of the distribution of the at least N + 1 values after mapping Whether the time-series data is normal or abnormal is determined and output by calculating at least one of specifying whether it is in the region or in the abnormal region.

According to the seventh aspect of the present disclosure, an abnormality can be detected according to the state of distribution on the phase space after mapping.

An abnormality detection device according to an eighth aspect of the present disclosure is the abnormality detection device according to any one of the first to seventh aspects,
The control unit
Create multiple different topological spaces.

According to the eighth aspect of the present disclosure, anomalies can be detected based on a plurality of different phase spaces.

An abnormality detection device according to a ninth aspect of the present disclosure is the abnormality detection device according to the eighth aspect,
The control unit
By specifying whether the time-series data is mapped to a normal region or an abnormal region on each phase space of the plurality of different phase spaces, the time-series data is normal or abnormal. Determine whether or not there is, and output.

According to the ninth aspect of the present disclosure, anomalies can be detected based on a plurality of different phase spaces.

An abnormality detection device according to a tenth aspect of the present disclosure is the abnormality detection device according to any one of the first to ninth aspects,
The control unit
When creating the normal region and the abnormal region on the phase space, and when determining whether the time series data is normal or abnormal,
The value of the time-series data at the arbitrary time and past N consecutive values from the arbitrary time, or the value of the time-series data at the arbitrary time and past N values from the arbitrary time every M times get the value.

According to the tenth aspect of the present disclosure, it is also possible to thin out and use the data that is the target of anomaly detection.

An abnormality detection device according to an eleventh aspect of the present disclosure is the abnormality detection device according to any one of the first to tenth aspects,
The time-series data, which is the object of abnormality detection of the air conditioner, is data related to the amount of refrigerant contained in the air conditioner.

According to the eleventh aspect of the present disclosure, refrigerant leakage can be detected.

A method according to a twelfth aspect of the present disclosure comprises:
A method executed by a control unit of an anomaly detection device,
An N+1 dimensional phase space is created from the value of the time series data that is the abnormality detection target of the air conditioner and N values past the arbitrary time, and the normal time series data and the abnormality are created on the phase space. A step of creating a normal region and an abnormal region on the phase space according to the distribution when the time series data of is mapped;
Acquiring at least N+1 values from the time-series data that is the abnormality detection target of the air conditioner, and specifying whether the at least N+1 values are mapped to the normal region or the abnormal region. and determining whether the time-series data is normal or abnormal and outputting the data.

A program according to the thirteenth aspect of the present disclosure comprises:
In the control unit of the anomaly detection device,
An N+1 dimensional phase space is created from the value of the time series data that is the abnormality detection target of the air conditioner and N values past the arbitrary time, and the normal time series data and the abnormality are created on the phase space. A procedure for creating a normal region and an abnormal region on the phase space according to the distribution to which the time series data of is mapped,
Acquiring at least N+1 values from the time-series data that is the abnormality detection target of the air conditioner, and specifying whether the at least N+1 values are mapped to the normal region or the abnormal region. Then, a procedure for judging whether the time-series data is normal or abnormal and outputting the data is executed.

It is a figure for demonstrating the conventional abnormality detection. It is a configuration example of the entire present disclosure. FIG. 4 is a diagram for explaining mapping of time-series data, which is an anomaly detection target of the present disclosure, to a phase space; FIG. 4 is a diagram for explaining creation of a normal region and an abnormal region according to an embodiment of the present disclosure; FIG. FIG. 4 is a diagram for explaining overlap of a normal region and an abnormal region according to the present disclosure; FIG. FIG. 4 is a diagram for explaining abnormality detection according to an embodiment of the present disclosure; FIG. FIG. 4 is a diagram for explaining the number of dimensions of a phase space according to an embodiment of the present disclosure; FIG. 4 is an example of a phase space according to an embodiment of the present disclosure; 4 is an example of a phase space according to an embodiment of the present disclosure; FIG. 4 is a diagram for explaining creation of multiple phase spaces and anomaly detection using multiple phase spaces according to an embodiment of the present disclosure; 4 is a flowchart of normal region and abnormal region creation processing according to an embodiment of the present disclosure; 6 is a flowchart of normality/abnormality determination processing according to an embodiment of the present disclosure. 1 is a hardware configuration diagram of an abnormality detection device according to an embodiment of the present disclosure; FIG. 1 is a hardware configuration diagram of an air conditioner (for cooling operation) according to an embodiment of the present disclosure; FIG. 1 is a hardware configuration diagram of an air conditioner (for heating operation) according to an embodiment of the present disclosure; FIG. 1 is a hardware configuration diagram of an air conditioner (for simultaneous cooling and heating operation) according to an embodiment of the present disclosure; FIG.

Embodiments of the present disclosure will be described below based on the drawings.

<Overall configuration example>
FIG. 2 is an example of the overall configuration of the present disclosure. An abnormality detection device 10 is a device that detects an abnormality in a device 100 such as an air conditioner. The abnormality detection device 10 includes a control section (control section 1001 in FIG. 13).

<Example 1>
For example, the abnormality detection device 10 is a device different from the air conditioner 100 . The abnormality detection device 10 and the air conditioner 100 are communicably connected via an arbitrary network. For example, the abnormality detection device 10 may be a cloud server remote from the air conditioner 100, or may be a computer installed in the same building as the air conditioner. One abnormality detection device 10 may detect an abnormality in one air conditioner 100 or may detect an abnormality in a plurality of air conditioners 100 .

<Example 2>
For example, the abnormality detection device 10 is part of the air conditioner 100 . One abnormality detection device 10 may detect an abnormality in one air conditioner 100 or may detect an abnormality in a plurality of air conditioners 100 .

<Example 3>
The processing of the abnormality detection device described in this specification is distributed and executed by the abnormality detection device 10 that is a device separate from the air conditioner 100 and the abnormality detection device 10 that is a part of the air conditioner 100. may be

<Overview>
Processing for detecting an abnormality in the air conditioner 100 executed by the abnormality detection device 10 will be described. In the present specification, the case of using a trained model created by machine learning (normal regions and abnormal regions created by machine learning, which will be described later) will be mainly described, but the present disclosure is not based on machine learning. Generated normal and abnormal regions can also be used.

<Learning>
First, the control unit 1001 of the abnormality detection device 10 creates a normal region and an abnormality region on the phase space used for detecting abnormality of the air conditioner 100 .

Specifically, the control unit 1001 of the anomaly detection device 10 controls the time-series data of an index (hereinafter referred to as an anomaly detection target) for detecting an anomaly of the air conditioner 100, the value of an arbitrary time and the arbitrary time Create an N+1 dimensional topological space from the N older values. Next, the control unit 1001 of the abnormality detection device 10 puts normal time-series data (time-series data of an index when the air conditioner 100 is normal) and abnormality time-series data (the air conditioner 100 A normal region and an abnormal region are created on the phase space according to the distribution when the time-series data of the index when is abnormal) is mapped. For example, the control unit 1001 of the anomaly detection device 10 inputs a plurality of (for example, of a plurality of properties) normal time-series data (time-series data of indicators when the air conditioner 100 is normal) and a plurality of ( For example, using teacher data that is abnormal time-series data (time-series data of indicators when the air conditioner 100 is abnormal) of a plurality of properties and whose output is a normal region or an abnormal region on the phase space It is possible to create a normal region and an abnormal region on the phase space by machine learning with

<Inference>
Next, the control unit 1001 of the anomaly detection device 10 detects an anomaly of the air conditioner 100 using the created normal region and anomalous region in the phase space.

Specifically, the control unit 1001 of the abnormality detection device 10 acquires at least N+1 values from the time-series data that is the abnormality detection target of the air conditioner 100, and at least N+1 values are mapped to the normal region. It is determined whether the time-series data is normal or abnormal by specifying whether it is mapped to an abnormal area or not, and output.

In this specification, the case where the abnormality detection target is the refrigerant amount index (data related to the amount of refrigerant contained in the air conditioner 100) will be mainly described, but the present disclosure can be applied to any failure index. can do.

Below, the processing for creating the normal region and the abnormal region will be described in detail separately for <mapping to the topological space>, <creation of the normal region and the abnormal region>, and <overlapping of the normal region and the abnormal region>.

<Mapping to topological space>
FIG. 3 is a diagram for explaining mapping of time-series data, which is an anomaly detection target of the present disclosure, onto a phase space. The control unit 1001 of the anomaly detection device 10 creates an N+1-dimensional phase space from an arbitrary time value of the time-series data, which is an anomaly detection target of the air conditioner 100, and N values past the arbitrary time. We map normal time-series data and abnormal time-series data onto it.

3, the controller 1001 of the anomaly detection device 10 detects an anomaly detection target of the air conditioner 100 from the arbitrary time value (refrigerant amount index value at time t) of the time-series data and the arbitrary time An N+1-dimensional phase space (two-dimensional plane) is created from past N values (refrigerant amount index values at time t−1). Then, the control unit 1001 of the abnormality detection device 10 puts normal time-series data (time-series data of the refrigerant amount index when the air conditioner 100 is normal) and abnormal time-series data (air conditioner Time-series data of the refrigerant amount index when 100 is abnormal (refrigerant leakage) is mapped.

<Creation of normal area and abnormal area>
FIG. 4 is a diagram for explaining creation of normal regions and abnormal regions according to an embodiment of the present disclosure. The control unit 1001 of the anomaly detection device 10 creates a normal region and an anomaly region on the phase space according to the distribution when the normal time-series data and the abnormal time-series data are mapped on the phase space. If the index (t-1) ~ reference value (approximation) and index (t) ~ reference value (approximation), then it is considered normal, and either index (t-1) or index (t) is greater than the reference value. If it is below (or above), it is determined to be normal (deemed to be noise), and if the index (t-1) < (or >) reference value and index (t) < (or >) reference value, it is determined to be abnormal. can be done.

In the example of FIG. 4, the control unit 1001 of the abnormality detection device 10 maps the normal region on the phase space according to the distribution when the normal time-series data and the abnormal time-series data are mapped on the phase space. and creating (eg, demarcating) anomalous regions. For example, if the index (t-1) to 0 (approximate) and the index (t) to 0 (approximate), it is determined to be normal, and if either the index (t-1) or the index (t) is negative, it is normal (assumed to be noise), and if index (t-1)<0 and index (t)<0, it can be determined as abnormal.

In this way, by creating a normal region and an abnormal region (for example, a boundary line) so as not to include data that would cause an erroneous judgment, even data judged to be abnormal in one dimension are mapped onto a plane. Plotted in the normal region.

<Overlap of normal and abnormal regions>
FIG. 5 is a diagram for explaining the overlap of the normal region and the abnormal region of the present disclosure. When the time-series data, which is the object of anomaly detection, is mapped, the normal region and the abnormal region may overlap on the topological space. When the normal region and the abnormal region overlap, the control unit 1001 of the abnormality detection device 10 detects the region where the normal region and the abnormal region overlap (for example, the “normal region” and the “leakage region” in FIG. The area indicated by the “overlap”) can be taken as the normal area.

The process of detecting an abnormality in the air conditioner 100 will be described in detail below.

<Anomaly detection>
FIG. 6 is a diagram for explaining abnormality detection according to an embodiment of the present disclosure. The control unit 1001 of the abnormality detection device 10 acquires at least N+1 values from the time-series data that is the abnormality detection target of the air conditioner 100, and at least N+1 values are mapped to the normal region or mapped to the abnormal region. It is determined whether the time-series data is normal or abnormal by specifying whether the time-series data is normal or abnormal (at least N+1 values within the dashed frame on the left side of FIG. 6 are mapped (moving the frame )).

For example, the control unit 1001 of the anomaly detection device 10 calculates the relationship between the number of data mapped to the normal region and the number of data mapped to the abnormal region among at least N+1 values, so that time-series data is normal or abnormal (for example, if the number of data mapped to the normal region is greater can be determined).

For example, the control unit 1001 of the abnormality detection device 10 determines whether the time-series data is normal by calculating whether the center of gravity of the distribution after mapping of at least N+1 values is in the normal region or in the abnormal region. or abnormal (for example, if the center of gravity of the distribution after mapping is in the normal region, it is determined to be normal, and if the center of gravity of the distribution after mapping is in the abnormal region, it is determined to be abnormal) can be done.

<<Number of dimensions of topological space>>
FIG. 7 is a diagram for explaining the number of dimensions of the phase space according to one embodiment of the present disclosure. The topological space may be a topological space of any number of dimensions. The topological space may be two-dimensional, three-dimensional or more. For example, as shown in FIG. 7, the phase space can be represented by the value of the failure indicator (eg, refrigerant quantity indicator) at time t and the value of the failure indicator (eg, refrigerant quantity indicator) at time t−1 and the failure It is a phase space whose axis is the value at time t−2 of the index (eg, refrigerant amount index). As the number of dimensions increases, the boundary surface becomes more complicated, and false detection due to noise disappears.

<<Degree of change>>
FIG. 8 is an example of a phase space according to one embodiment of the present disclosure. The axes of the phase space may be computed fault indices. For example, as shown in FIG. 8, the phase space may include the difference between the value of the failure index (eg, refrigerant quantity index) at time t and the value at time t−1, and the failure index (eg, refrigerant quantity index) is a phase space whose axis is the difference between the value at time t−1 and the value at time t−2. As shown in Figure 8, the time-series data of a rapid leak (a large amount of refrigerant leaks in a short time and causes a lack of oxygen or ignition of a flammable refrigerant) Leakage leading to deterioration of function due to leakage of refrigerant) is mapped to the lower left of the time-series data. In this way, if the change is significant, it will be mapped to the lower left (in the case of the refrigerant amount index), and it is possible to determine the degree of progress, such as whether the abnormality is a fast-changing abnormality or a slow-changing abnormality. In addition, in the case of an index whose value increases when there is an abnormality, if the change is significant, it will be mapped to the upper right.

FIG. 9 is an example of a phase space according to an embodiment of the present disclosure. The axes of the phase space may be computed fault indices. For example, as shown in FIG. 9, the phase space may include the value of the failure indicator (eg, refrigerant quantity indicator) at time t, the value of the failure indicator (eg, refrigerant quantity indicator) at time t, and the value at time t− 1 is the topological space whose axes are the difference of the values in .

<<Multiple Topological Spaces>>
FIG. 10 is a diagram for explaining creation of multiple phase spaces and anomaly detection using multiple phase spaces according to an embodiment of the present disclosure.

In one embodiment of the present disclosure, the control unit 1001 of the anomaly detection device 10 can create multiple different phase spaces. For example, as shown in FIG. 10, the control unit 1001 of the abnormality detection device 10 controls the value of the failure indicator (eg, refrigerant amount indicator) at time t and the value of the failure indicator (eg, refrigerant amount indicator) at time t− 1 and the difference between the value at time t and the value at time t-1 of the index of failure (e.g. refrigerant amount index) and the time t of the index of failure (e.g. refrigerant amount index) A phase space can be created whose axis is the difference between the value at −1 and the value at time t−2.

In one embodiment of the present disclosure, the control unit 1001 of the anomaly detection device 10 determines whether the time-series data is mapped to a normal region or an abnormal region on each phase space of a plurality of different phase spaces. By specifying it, it is possible to determine whether the time-series data is normal or abnormal and output it. For example, as shown in FIG. 10, the control unit 1001 of the abnormality detection device 10 causes the time-series data of the failure index (for example, the refrigerant amount index) to fall into the normal region on each phase space of the two phase spaces. The degree of deviation from normal and the degree of change can be determined by specifying whether it is mapped or mapped to an abnormal region. Therefore, in addition to judging whether it is normal or abnormal (judging by the degree of deviation from normal), it is possible to judge the degree of progress, such as whether the abnormality is a fast-changing or slow-changing abnormality (judgment by the degree of change). can do.

<<Interval of time-series data>>
In an embodiment of the present disclosure, the control unit 1001 of the anomaly detection device 10 creates a normal region and an anomaly region in the phase space, and determines whether the time-series data is normal or abnormal. 2, N values in the past that are continuous with an arbitrary time value of the time series data, or an arbitrary time value of the time series data and the past N values that are obtained every M times from an arbitrary time. be able to.

<Method>
FIG. 11 is a flowchart of normal region and abnormal region creation processing according to an embodiment of the present disclosure.

At step 11 (S11), the control unit 1001 of the anomaly detection device 10 creates a phase space. Specifically, the control unit 1001 of the anomaly detection device 10 extracts an N+1-dimensional phase space from an arbitrary time value of the time-series data that is an anomaly detection target of the air conditioner 100 and N values past the arbitrary time. create.

At step 12 (S12), the control unit 1001 of the anomaly detection device 10 maps the time-series data, which is an anomaly detection target, onto the phase space. Specifically, the control unit 1001 of the abnormality detection device 10 maps the normal time-series data and the abnormal time-series data on the phase space created in S11.

At step 13 (S13), the control unit 1001 of the abnormality detection device 10 creates a normal region and an abnormal region on the phase space. Specifically, the control unit 1001 of the anomaly detection device 10 maps the normal region and the Create anomalous regions.

FIG. 12 is a flowchart of normality/abnormality determination processing according to an embodiment of the present disclosure.

At step 21 (S21), the control unit 1001 of the anomaly detection device 10 acquires time-series data that is an anomaly detection target. Specifically, the control unit 1001 of the abnormality detection device 10 acquires at least N+1 values from the time-series data that is the abnormality detection target of the air conditioner 100 .

At step 22 (S22), the control unit 1001 of the anomaly detection device 10 maps the time-series data, which is an anomaly detection target, onto the phase space. Specifically, the control unit 1001 of the abnormality detection device 10 maps at least N+1 values acquired in S21 onto the phase space created in FIG.

At step 23 (S23), the control unit 1001 of the anomaly detection device 10 determines whether the time series data to be anomaly detected is normal or abnormal. Specifically, the control unit 1001 of the abnormality detection device 10 determines whether the time-series data is normal by specifying whether at least N+1 values are mapped to the normal region or the abnormal region in S22. or abnormal.

At step 24 (S24), the control unit 1001 of the abnormality detection device 10 outputs the determination result. Specifically, the control unit 1001 of the abnormality detection device 10 outputs the result determined in S23 (for example, an abnormality of the air conditioner 100 is reported).

In this way, conventionally, erroneous judgments caused by noise due to failure index values before and after were not taken into consideration when determining the threshold value (for example, the past refrigerant amount index in FIG. 1 information, it can be understood that the value of the refrigerant quantity index has just become low by chance). In one embodiment of the present disclosure, a normal region and an abnormal region are created by mapping the failure index data onto a topological space centered on past data and current data of the failure index (for example, a boundary line or draw a boundary surface), it is possible to reduce false detections due to noise. Normal regions and abnormal regions (eg, boundary lines or boundary surfaces) can be easily created by machine learning with normal data and abnormal data. At the time of abnormality detection, it is possible to map chronologically continuous failure index data onto the phase space, and determine normality/abnormality based on whether it is mapped to the normal region or the abnormal region.

<Hardware configuration of anomaly detection device>
FIG. 13 is a hardware configuration diagram of the abnormality detection device 10 according to an embodiment of the present disclosure.

The anomaly detection device 10 has a CPU (Central Processing Unit) 1001, a ROM (Read Only Memory) 1002, and a RAM (Random Access Memory) 1003. The CPU 1001, ROM 1002, and RAM 1003 form a so-called computer. The abnormality detection device 10 also has an auxiliary storage device 1004 , a display device 1005 , an operation device 1006 , an I/F (Interface) device 1007 and a drive device 1008 . Each piece of hardware of the abnormality detection device 10 is connected to each other via a bus B. As shown in FIG.

The CPU 1001 is a computing device that executes various programs installed in the auxiliary storage device 1004 .

The ROM 1002 is a non-volatile memory. The ROM 1002 functions as a main storage device that stores various programs, data, etc. necessary for the CPU 1001 to execute various programs installed in the auxiliary storage device 1004 . Specifically, the ROM 1002 functions as a main storage device that stores boot programs such as BIOS (Basic Input/Output System) and EFI (Extensible Firmware Interface).

The RAM 1003 is a volatile memory such as DRAM (Dynamic Random Access Memory) or SRAM (Static Random Access Memory). The RAM 1003 functions as a main storage device that provides a work area that is developed when various programs installed in the auxiliary storage device 1004 are executed by the CPU 1001 .

The auxiliary storage device 1004 is an auxiliary storage device that stores various programs and information used when various programs are executed.

The display device 1005 is a display device that displays the internal state of the abnormality detection device 10 and the like.

The operation device 1006 is an input device through which the administrator of the anomaly detection device 10 inputs various instructions to the anomaly detection device 10 .

The I/F device 1007 is a communication device for connecting to a network and communicating with other devices.

A drive device 1008 is a device for setting a storage medium 1009 . The storage medium 1009 here includes media for optically, electrically, or magnetically recording information such as CD-ROMs, flexible disks, and magneto-optical disks. Also, the storage medium 1009 may include a semiconductor memory or the like that electrically records information such as a ROM or a flash memory.

Various programs to be installed in the auxiliary storage device 1004 are installed by, for example, setting the distributed storage medium 1009 in the drive device 1008 and reading the various programs recorded in the storage medium 1009 by the drive device 1008. be done. Alternatively, various programs installed in the auxiliary storage device 1004 may be installed by being downloaded from the network via the I/F device 1007 .

A hardware configuration example of the air conditioner 100 will be described below with reference to FIGS. 14 to 16. FIG. The air conditioner 100 may be an arbitrary air conditioning system such as a multi-air conditioner such as a multi-air conditioner for buildings, a central air-conditioning system using a chiller as a heat source, an air conditioner for stores and offices, and a room air conditioner. Besides, it may be a refrigerating/freezing system. The air conditioner 100 can have multiple indoor units 300 . The plurality of indoor units 300 may include indoor units with different performance, may include indoor units with the same performance, or may include indoor units that are not in operation.

<Hardware configuration example of air conditioner (for cooling operation)>
FIG. 14 is a hardware configuration diagram of the air conditioner 100 according to an embodiment of the present disclosure. Air conditioner 100 has outdoor unit 200 and one or more indoor units 300 .

In the example of FIG. 14, the outdoor heat exchanger 201, the outdoor unit main expansion valve 205, the supercooling heat exchanger 203, the indoor heat exchanger expansion valve 302, the four-way switching valve 206, and the indoor heat exchanger 301 , and the compressor 202 are connected by a refrigerant pipe to form a main refrigerant circuit. The four-way switching valve 206 has a flow path set so as to supply the discharge gas of the compressor 202 to the outdoor heat exchanger 201 . In the example of FIG. 14, the supercooling heat exchanger expansion valve 204 is further connected to the bypass pipe connected from the pipe between the outdoor heat exchanger 201 and the supercooling heat exchanger 203 to the suction side pipe of the compressor 202. is provided. The supercooling heat exchanger 203 includes a supercooling heat exchanger expansion valve 204 provided in a bypass pipe connected between the outdoor heat exchanger 201 and the supercooling heat exchanger 203 to the suction side pipe of the compressor 202. is a heat exchanger that exchanges heat between the refrigerant that has passed through and the refrigerant in the main refrigerant circuit. Note that the bypass example in FIG. 14 is an example.

<<Outdoor unit>>
On the outdoor unit 200 side, an outdoor heat exchanger 201, a compressor 202, a supercooling heat exchanger 203, a supercooling heat exchanger expansion valve (bypass circuit) 204, and an outdoor unit main expansion valve (main refrigerant circuit) 205 are connected to the piping. The outdoor unit 200 has various sensors (temperature sensors (eg, thermistors) (1), (3), (4), (6), (7), pressure sensors (2), (5), etc.).

<<Indoor unit>>
On the indoor unit 300 side, an indoor heat exchanger 301 and an indoor heat exchanger expansion valve 302 are connected to pipes. The indoor unit 300 has various sensors (temperature sensors (eg, thermistors) (8), (9), etc.).

<Hardware configuration example of air conditioner (for heating operation)>
FIG. 15 is a hardware configuration diagram of an air conditioner (for heating operation) 100 according to an embodiment of the present disclosure. Air conditioner 100 has outdoor unit 200 and one or more indoor units 300 .

In the example of FIG. 15, an outdoor heat exchanger 201, a compressor 202, a four-way switching valve 206, an indoor heat exchanger 301, an indoor heat exchanger expansion valve 302, a supercooling heat exchanger 203, and an outdoor The machine main expansion valve 205 is connected by a refrigerant pipe to form a main refrigerant circuit. The four-way switching valve 206 has a flow path set so as to supply the discharge gas of the compressor 202 to the indoor heat exchanger 301 .

<Hardware configuration example of air conditioner (for simultaneous cooling and heating operation)>
The present disclosure can be applied not only to cooling operation and heating operation, but also to simultaneous cooling and heating operation. The simultaneous cooling and heating operation will be described below with reference to FIG. 16 .

FIG. 16 is a hardware configuration diagram of an air conditioner (in the case of simultaneous cooling and heating operation) 100 according to an embodiment of the present disclosure. The air conditioner 100 has an outdoor heat exchanger 201-1 and an outdoor heat exchanger 201-2 with a two-part structure, and a plurality of indoor units, which are connected by three connecting pipes, and simultaneous cooling and heating operation is possible. is. FIG. 16 shows an example of cooling-dominant operation, in which indoor unit 300-1 is operated in heating mode and indoor unit 300-2 is operated in cooling mode. At this time, the outdoor heat exchanger 201-1 functions as a condenser, and the outdoor heat exchanger 201-2 functions as an evaporator.

Although the embodiments have been described above, it will be understood that various changes in form and detail are possible without departing from the spirit and scope of the claims.

This international application claims priority based on Japanese Patent Application No. 2021-213979 filed on December 28, 2021, and the entire contents of No. 2021-213979 are hereby incorporated into this international application.

10 Abnormality detection device 100 Air conditioner 200 Outdoor unit 201 Outdoor heat exchanger 201-1 Outdoor heat exchanger (condenser)
201-2 Outdoor heat exchanger (evaporator)
202 Compressor 203 Supercooling heat exchanger 204 Supercooling heat exchanger expansion valve 205 Outdoor unit main expansion valve 206 Four-way switching valve 300 Indoor unit 300-1 Heating indoor unit 300-2 Cooling indoor unit 301 Indoor heat exchanger 302 Indoor Heat exchanger expansion valve 1001 Control unit 1002 ROM
1003 RAM
1004 auxiliary storage device 1005 display device 1006 operation device 1007 I/F device 1008 drive device 1009 storage medium

Claims

An anomaly detection device comprising a control unit,
The control unit
An N+1 dimensional phase space is created from the value of the time series data that is the abnormality detection target of the air conditioner and N values past the arbitrary time, and the normal time series data and the abnormality are created on the phase space. Create a normal region and an abnormal region on the phase space according to the distribution when the time series data of is mapped,
Acquiring at least N+1 values from the time-series data that is the abnormality detection target of the air conditioner, and specifying whether the at least N+1 values are mapped to the normal region or the abnormal region. and an anomaly detection device that determines whether the time-series data is normal or abnormal and outputs the result.
The abnormality detection device according to claim 1, wherein the normal region and the abnormal region on the phase space are learned models created by machine learning.
The anomaly detection device according to claim 1 or 2, wherein the phase space is two-dimensional.
2. The phase space is created from at least one of a value of time-series data as an abnormality detection target for the air conditioner and a difference between time-series data as an abnormality detection target for the air conditioner. 4. The abnormality detection device according to any one of 3.
The control unit
5. The anomaly detection device according to claim 1, wherein the anomaly of the time-series data is determined based on at least one of a degree of deviation from normal and a degree of change.
The control unit
The abnormality detection device according to any one of claims 1 to 5, wherein when the normal region and the abnormal region overlap, the region where the normal region and the abnormal region overlap is defined as the normal region.
The control unit
A relationship between the number of data mapped to the normal region and the number of data mapped to the abnormal region among the at least N + 1 values, and the normal center of the distribution of the at least N + 1 values after mapping Specifying whether it is in the region or in the abnormal region, and determining and outputting whether the time-series data is normal or abnormal by calculating at least one of The anomaly detection device according to any one of the items.
The control unit
8. The anomaly detection device according to any one of claims 1 to 7, which creates a plurality of different phase spaces.
The control unit
By specifying whether the time-series data is mapped to a normal region or an abnormal region on each phase space of the plurality of different phase spaces, the time-series data is normal or abnormal. 9. The anomaly detection device according to claim 8, which determines whether or not there is an abnormality and outputs the result.
The control unit
When creating the normal region and the abnormal region on the phase space, and when determining whether the time series data is normal or abnormal,
The value of the time-series data at the arbitrary time and past N consecutive values from the arbitrary time, or the value of the time-series data at the arbitrary time and past N values from the arbitrary time every M times 10. The anomaly detection device according to any one of claims 1 to 9, which acquires a value.
The anomaly detection device according to any one of claims 1 to 10, wherein the time-series data, which is an anomaly detection target for the air conditioner, is data related to the amount of refrigerant contained in the air conditioner.
A method executed by a control unit of an anomaly detection device,
An N+1 dimensional phase space is created from the value of the time series data that is the abnormality detection target of the air conditioner and N values past the arbitrary time, and the normal time series data and the abnormality are created on the phase space. A step of creating a normal region and an abnormal region on the phase space according to the distribution when the time series data of is mapped;
Acquiring at least N+1 values from the time-series data that is the abnormality detection target of the air conditioner, and specifying whether the at least N+1 values are mapped to the normal region or the abnormal region. and determining whether the time-series data is normal or abnormal and outputting the result.
In the control unit of the anomaly detection device,
An N+1 dimensional phase space is created from the value of the time series data that is the abnormality detection target of the air conditioner and N values past the arbitrary time, and the normal time series data and the abnormality are created on the phase space. A procedure for creating a normal region and an abnormal region on the phase space according to the distribution to which the time series data of is mapped,
Acquiring at least N+1 values from the time-series data that is the abnormality detection target of the air conditioner, and specifying whether the at least N+1 values are mapped to the normal region or the abnormal region. A program for executing a procedure for judging whether the time-series data is normal or abnormal and outputting it.