CN115707913A - Abnormality detection system, abnormality detection method for abnormality detection system, and recording medium for abnormality detection system - Google Patents

Abstract

The invention provides an abnormality detection system which can detect abnormality of an air conditioner composed of an outdoor unit and an indoor unit, identify the cause of the abnormality, and identify which device has the abnormality. The method includes the steps of obtaining a plurality of measured values related to a plurality of devices from a plurality of sensors, selecting measured values of the sensors related to the device types, calculating statistical values based on the selected values, calculating a feature quantity set indicating a position in a feature quantity space based on the calculated plurality of statistical values for each of the plurality of device types, registering the feature quantity set as normal data in the feature quantity space in the case of learning, determining whether or not there is an abnormality based on a deviation between the feature quantity set and the normal data in the case of a monitored state after learning, determining a predetermined device type as a device type in which an abnormality has occurred from the plurality of device types in the case of determining an abnormality, and determining a device in which an abnormality has occurred from among 1 or more devices as the predetermined device types.

Description

Abnormality detection system, abnormality detection method for abnormality detection system, and recording medium for abnormality detection system

Technical Field

The present invention relates to an abnormality detection system, an abnormality detection method for an abnormality detection system, and a recording medium for an abnormality detection system, and more particularly to an abnormality detection system for detecting an abnormality of an air conditioning system, an abnormality detection method for an abnormality detection system, and a recording medium for an abnormality detection system.

Background

The detection of an abnormality or a failure of an air conditioning system (hereinafter, referred to as an abnormality) and a sign of the occurrence of the abnormality (abnormality detection) is very important for cost reduction of inspection work and maintenance of the air conditioning system. In addition, recently, there is a trend to operate services that are not only providing air conditioners to customers but also uniformly performing operation and maintenance of the air conditioners.

In addition, in the air conditioning system for business and the freezing and refrigerating equipment, "freon discharge inhibition law" of environmental province was implemented from 4 months in 2015, and "simple inspection" and "regular inspection" were performed on equipment obligations using freon gas held by the enterprises. Therefore, it becomes important to grasp the abnormality of the air conditioning system in terms of the service business.

From the viewpoint of maintenance and repair, it is important to detect an abnormality in an air conditioning system including an outdoor unit (one or more) and an indoor unit(s) connected by a pipe through which a refrigerant circulates, to identify the cause of the abnormality, and to identify which device has an abnormality.

By using such abnormality detection, the following environmental value, economic value, and social value can be provided.

That is, in terms of environmental value, it is possible to contribute to prevention of the greenhouse effect by reducing the amount of leakage caused by early detection of leakage of refrigerant (freon). In addition, in terms of economic value, it is possible to contribute to reduction of life cycle cost by suppressing a loss of business (production stoppage, yield, loss of food disposal, and the like) of a customer due to an unexpected failure of an air conditioner, or by changing maintenance from a time reference to a state reference. In addition, in terms of social value, it is possible to contribute to stable operation of air conditioning equipment, which is important in medical practice, and elimination of a shortage of maintenance personnel (improvement in work efficiency of maintenance workers).

However, in the abnormality detection of the air conditioning system, the following methods are known: the normal state is learned using as input a capability value (feature value) of each device calculated from sensor signals provided in the outdoor unit and the indoor unit, and abnormality detection is performed based on the degree of deviation from the learned normal state. As the learning method, there are individual learning by a combination of 1 indoor unit and 1 outdoor unit and collective learning by totaling all the indoor units and all the outdoor units.

The problem of the individual learning is that the capacity value (characteristic amount) of each device in the outdoor unit is affected by the operation states (operation/non-operation) of the plurality of indoor units. That is, in the case of the individual learning, the feature distribution of the normal state of the outdoor unit changes depending on the operation mode (combination of operation/non-operation of each indoor unit), and thus the abnormality detection cannot be performed accurately.

In order to avoid this problem, learning is required in accordance with the operation mode. However, if a Variable Refrigerant Flow (VRF) air conditioning system including a plurality of outdoor units and indoor units is used, the number of operation modes becomes enormous, which is not practical.

In addition, if the uniform learning is performed, the uniform learning has a problem in that different distributions are learned according to the operation mode. However, since the dimensionality is increased in the unified learning and the feature distribution in the normal state is made coarser, there are problems such as a large number of false positives or a high threshold value, which lowers the abnormality detection sensitivity.

Regarding abnormality detection of devices constituting a system, for example, the following system is disclosed in japanese patent laid-open No. 2001-289492 (patent document 1): in an engine-driven air conditioner, when an engine is in a non-operable state, whether an engine stall has occurred is monitored in advance, and when an engine stall has occurred, a throttle valve diagnosis mode is started and a fuel adjustment valve diagnosis mode is executed, thereby detecting a malfunction of an internal combustion engine that drives a compressor.

Documents of the prior art

Patent document

Patent document 1: japanese patent laid-open No. 2001-289492

Disclosure of Invention

Problems to be solved by the invention

The abnormality detection system described in patent document 1 is a system in which: in an engine-driven air conditioner, when an engine is in a non-operable state, whether or not an engine stall has occurred is monitored in advance, and when the engine stall has occurred, a throttle valve diagnosis mode is started and a fuel adjustment valve diagnosis mode is executed, thereby detecting a malfunction of an internal combustion engine that drives a compressor.

However, in patent document 1, no consideration is given to abnormality detection when the method is applied to an air conditioning system including a plurality of indoor units and outdoor units. As described above, patent document 1 does not disclose or suggest the following: the abnormality of an air conditioner comprising outdoor units (1 or more) and indoor units (more) connected by pipes through which a refrigerant circulates is detected, the cause of the abnormality is identified, and which of the units is abnormal is identified.

An object of the present invention is to provide an abnormality detection system, an abnormality detection method for an abnormality detection system, and a recording medium for an abnormality detection system, which are capable of detecting an abnormality of an air conditioning apparatus including an outdoor unit (1 or a plurality of) and an indoor unit(s) connected by a pipe through which a refrigerant circulates, identifying a cause of the abnormality, and identifying which apparatus has the abnormality.

Means for solving the problems

The invention provides an abnormality detection system, comprising:

an air conditioning system having a plurality of devices and sensors; and a monitoring computer that monitors the air conditioning system,

it is characterized in that the preparation method is characterized in that,

the plurality of devices is made up of a plurality of device classes,

the monitoring computer performs the following processing:

(1) Obtaining a plurality of measurements associated with a plurality of devices from a plurality of sensors,

(2) For each of the kinds of the devices,

(2A) The measured values of the sensors associated with the device class are selected,

(2B) A statistical value is calculated based on the selected value,

(3) Calculating a feature quantity set representing a position within the feature quantity space based on the calculated plurality of statistical values for each of the plurality of device types,

(4) In the case where it is in the learning state,

(4A) The feature quantity set is registered as normal data to the feature quantity space,

(5) In the case of the monitoring state after learning,

(5A) Judging whether the abnormal condition exists or not based on the deviation of the characteristic quantity set and the normal data,

(5B) In the case where it is determined that there is an abnormality,

(5 Ba) identifying a predetermined device type as the device type in which the abnormality has occurred from among the plurality of device types,

(5 Bb) identifying the device in which the abnormality has occurred from among 1 or more devices which are predetermined device types.

Further, the present invention is an abnormality detection method of an abnormality detection system including: an air conditioning system having a plurality of devices for air conditioning and a plurality of sensors for measuring operation state quantities of the plurality of devices; and a monitoring computer that monitors an operation state of the air conditioning system, characterized in that the monitoring computer executes the steps of:

a step of acquiring a plurality of measurement values related to a plurality of types of devices from a plurality of sensors, selecting measurement values related to the same type of device (hereinafter, referred to as the same type of device), and calculating a statistical value based on the selected measurement values;

calculating a feature quantity set indicating a position in a feature quantity space based on statistics of at least 2 types of devices of the same kind;

registering the feature quantity set as normal data in a feature quantity space when the learning state is present;

judging the cause of the abnormality based on the degree of deviation between the feature quantity set and the normal data when the monitoring state is in the learned state; and

based on the cause of the abnormality, a predetermined device of the same kind is identified as a device of the same kind in which the abnormality has occurred, and a device in which the abnormality has occurred is identified from the predetermined device of the same kind.

Further, the present invention is a recording medium of an abnormality detection system storing a program for operating a monitoring computer used in the abnormality detection system, the abnormality detection system including: an air conditioning system including a plurality of devices for air conditioning and a plurality of sensors for measuring operation state quantities of the plurality of devices; a monitoring computer for monitoring the operation state of the air conditioning system,

the recording medium includes:

a measurement value acquisition program that acquires a plurality of measurement values relating to a plurality of devices from a plurality of sensors;

a statistical value calculation program that calculates a statistical value of a plurality of devices of the same type from measurement values of the plurality of devices of the same type (hereinafter, the plurality of devices are referred to as the same type);

a statistical value distribution density calculation program that calculates a statistical value set representing positions of statistical values in a two-dimensional plane of statistical values with 2 groups of statistical values related to the same kind of equipment as parameters;

a statistical value normal data storage program that registers the statistical value set as normal data to a statistical value two-dimensional plane in a learning mode;

an abnormality detection program that, in a monitoring mode executed after the learning mode, determines the presence or absence of an abnormality of the same kind of equipment based on a state of deviation of the statistical value set calculated by the statistical value distribution density calculation program from the statistical value set stored by the normal data storage program;

an abnormality cause specifying program that specifies an abnormality cause based on a correlation between the plurality of abnormalities detected by the abnormality detecting program;

a measurement value distribution density calculation program that calculates a measurement value set indicating positions of measurement values in a two-dimensional plane of the measurement values using 2 groups of the measurement values as parameters;

a measurement value normal data storage program that registers a measurement value set as normal data in a measurement value two-dimensional plane in a learning mode; and

and an abnormal equipment specifying program for selecting a measurement value related to the cause of the abnormality found by the abnormal cause specifying program in a monitoring mode executed after the learning mode, specifying the same kind of equipment related to the cause of the abnormality based on a deviation state between a measurement value set calculated by the measurement value distribution density calculation program based on the selected measurement value and a measurement value set stored by the measurement value normal data storage program, and specifying an abnormal equipment in which the abnormality has occurred from the same kind of equipment.

Effects of the invention

According to the present invention, it is possible to detect an abnormality in an air conditioning system including an outdoor unit (1 or more) and an indoor unit (1 or more) connected by a pipe through which a refrigerant circulates, identify the cause of the abnormality, and further identify which of the systems has an abnormality.

Drawings

Fig. 1 is a block diagram showing a basic configuration of an abnormality detection system to which an air conditioning system according to the present invention is applied.

Fig. 2 is a control block diagram showing processing blocks of the abnormality detection system of the air conditioning system according to the embodiment of the present invention.

Fig. 3 is an explanatory diagram illustrating a configuration of an air conditioning system having a plurality of devices.

Fig. 4 is an explanatory diagram illustrating a relationship between a measured value and a statistical value of the same type of equipment for each equipment.

Fig. 5A is an explanatory diagram illustrating a two-dimensional distribution density of normal data in the learning mode.

Fig. 5B is an explanatory diagram illustrating in detail the two-dimensional distribution density of fig. 5A.

Fig. 6A is an explanatory diagram for explaining an example of specifying the cause of an abnormality due to the two-dimensional distribution density in the entire analysis unit shown in fig. 2.

Fig. 6B is an explanatory diagram illustrating an example of determining the cause of the abnormality illustrated in fig. 6A in more detail.

Fig. 7 is an explanatory diagram illustrating an example of the abnormality cause determination map based on the combination of the statistical values.

Fig. 8 is an explanatory diagram for explaining an example of specifying an abnormal device based on a two-dimensional distribution density generated by a measurement value set of a device related to a cause of the abnormality.

Fig. 9A is a flowchart showing the first half of the flow of the learning process based on the measured values and the statistical values of the devices in the air conditioning system at the normal time.

Fig. 9B is a flowchart showing the second half of the flow of the learning process based on the measured values and statistical values of the devices in the normal air conditioning system.

Fig. 10A is a flowchart showing the first half of a monitoring process flow based on measured values and statistical values of devices in the air conditioning system during monitoring.

Fig. 10B is a flowchart showing an intermediate portion of a monitoring processing flow based on measured values and statistical values of the devices in the air conditioning system during monitoring.

Fig. 10C is a flowchart showing the second half of the monitoring process flow based on the measured values and the statistical values of the devices in the air conditioning system during monitoring.

Fig. 11 is an explanatory diagram for explaining an example of display of a display screen of a display device constituting the abnormality detection system.

Description of the reference numerals

100 … air conditioning equipment system, sensor signals of Ssig multiple equipments, 102 … sensor signal input section, 200 … monitoring computer, 210 … overall analysis section, 220 … detailed analysis section, 211 … sensor signal processing section, statistical value of Ssta … "same kind of equipment", 212 … feature vector extraction section, 213 … abnormality measure calculation section, 214 … threshold calculation section, 215 … abnormality detection section, 216 … abnormality cause determination section, ssop … measured value for each equipment, 221 … abnormality cause feature quantity extraction section, 222 … abnormality cause feature quantity analysis section, 223 zxft 6223 zxft 6262 abnormality cause learning data storage section, 5662 abnormal cause learning section.

Detailed Description

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings, but the present invention is not limited to the embodiments described below, and various modifications and application examples are included in the technical concept of the present invention.

Fig. 1 shows a basic configuration of an abnormality detection system in an air conditioning system. An abnormality detection system in an air conditioning system is composed of an air conditioning system 100 and a monitoring computer 200, the air conditioning system 100 includes a plurality of devices and sensors for air conditioning, and the monitoring computer 200 monitors the operation of the air conditioning system 100.

The air conditioning system in the air conditioning system 100 basically includes an outdoor unit and an indoor unit, which are connected by a refrigerant pipe. The outdoor unit includes devices such as a compressor, a heat exchanger, an expansion valve, and a blower fan, and the indoor unit includes devices such as a heat exchanger, an expansion valve, and a blower fan. Each device is provided with a sensor for detecting an operation state quantity (for example, temperature, pressure, current, and the like) of the device.

Since the air conditioning system is configured by 1 or more outdoor units and 1 or more indoor units, a plurality of devices constituting the outdoor units and the indoor units are known as device types. Here, the device type refers to a type of device that performs a certain function. For example, when a compressor is focused on as a device type, compressors of different outdoor units are "the same type of device" when viewed as a compressor, and similarly, when an expansion valve and a blower fan are focused on as a device type, the expansion valve and the blower fan of different outdoor units are "the same type of device", respectively. The same is true of the indoor unit.

The operation state quantity includes a sensor signal that can be directly measured by a sensor, and a measurement value that cannot be measured by a sensor and is obtained by calculation based on the sensor signal. Note that, when the sensor signal or the measurement value is specially processed, the description is given to that effect.

The monitoring computer 200 includes an interface having an input/output function and a processor having an arithmetic function, and the processor can execute the arithmetic operation of the present embodiment described below based on a control program. As an example of the processor, a CPU (Central Processing Unit) or a GPU (Graphics Processing Unit) may be considered, but other semiconductor devices may be used as long as they are a main body for executing predetermined arithmetic Processing.

The monitoring computer 200 may be integrated with the air conditioner system, or may be a cloud system connected by wire/wireless. The control program has a control function, and thus can be understood as a control function block.

Fig. 2 shows a configuration of an abnormality detection system in an air conditioning system. The air conditioning system 100 is installed at a plurality of customer sites, and transmits sensor information and the like to the monitoring computer 200 via a network. The customer site corresponds to a building (such as a building), for example, a VRF air conditioning system provided in the building.

The air conditioning system 100 includes 1 or more air conditioning systems 101, and a sensor signal input unit 102 that inputs sensor signals (Ssig) of a plurality of devices provided in the air conditioning systems 101 and transmits the sensor signals to the monitoring computer 200. In the air conditioning system 101, sensors are provided in 1 or more of various devices. Further, the sensor signals (Ssig) of the plurality of devices are, for example, signals related to temperature, current, pressure, and the like.

The monitoring computer 200 is composed of the following parts: an overall analysis unit (abnormality cause specification analysis unit) 210 that detects an abnormality of the air conditioning system 101 and specifies the "abnormality cause"; and a detailed analysis unit (abnormal equipment specifying analysis unit) 220 that specifies the abnormal position of the air conditioning system 101, for example, the "abnormal equipment" of the outdoor units or the indoor units, based on the "abnormal cause" obtained by the entire analysis unit 210.

The overall analysis unit 210 and the detailed analysis unit 220 are functions executed by a control program of a processor. The flow of processing executed by these control programs will be described with reference to fig. 9A to 9B and fig. 10A to 10C.

The entire analysis unit 210 inputs the sensor signals (Ssig) of the plurality of devices transmitted from the air-conditioning equipment system 100 to the sensor signal processing unit 211, detects an abnormality of the air-conditioning system 101 using the statistical value (Ssta) of the "same kind of devices" generated by the sensor signal processing unit 211, and specifies the "cause of the abnormality".

The overall analysis unit 210 includes a feature vector extraction unit 212, an abnormality measure calculation unit 213, a threshold calculation unit 214, an abnormality detection unit 215, and an abnormality cause identification unit 216.

On the other hand, the detailed analysis unit 220 inputs the sensor signals (Ssig) of the plurality of devices transmitted from the air conditioning system 100 to the sensor signal processing unit 211, selects the measurement value (Ssop) based on the "abnormality cause" obtained by the entire analysis unit 210 using the measurement value (Ssop) for each device generated by the sensor signal processing unit 211, and specifies the "abnormal device" as the abnormal position of the air conditioning system 101.

The detailed analysis unit 220 includes an abnormality cause feature amount extraction unit 221, an abnormality cause feature amount analysis unit 222, an abnormality cause feature amount learning data storage unit 223, and an abnormality position specification unit 224.

Next, the control functions of the overall analysis unit 210 and the detailed analysis unit 220 will be described, and first, the "learning mode" in the normal state of the overall analysis unit 210 will be described.

In the entire analysis unit 210, the sensor signals (Ssig) of the plurality of devices transmitted from the air-conditioning equipment system 100 are converted into the statistical value (Ssta) of the "same type of device" by the sensor signal processing unit 211. A plurality of the statistical values (Ssta) are obtained. The statistical value (Ssta) is defined by "a total value, an average value, a maximum value, a minimum value, a standard deviation, a variance, a median, a histogram" or the like. The statistical value (Ssta) is input to the feature vector extraction unit 212 at the subsequent stage.

The feature vector extraction unit 212 extracts feature vectors using the statistical value (Ssta) of the "homogeneous plant", and inputs the extracted feature vectors to the abnormality measure calculation unit 213. The abnormal measure calculation unit 213 calculates an abnormal measure for each feature vector at predetermined time intervals (which may be expressed as time points) using the feature vector of the learning period specified in advance.

The calculated abnormality measure is input to the threshold value calculation unit 214, and the threshold value calculation unit 214 calculates a threshold value (SL) corresponding to the abnormality measure in the "learning mode". The threshold (SL) calculated by the threshold calculation unit 214 and the abnormality metric calculated by the abnormality metric calculation unit 213 are compared by the abnormality detection unit 215. Then, the comparison result is sent to the abnormality cause determination section 216, where "abnormality cause" is determined.

Next, the "monitoring mode" of the entire analysis unit 210 will be described. The "monitor mode" is executed after the "learning mode" ends. In the "monitoring mode" of the entire analysis unit 210, the sensor signals (Ssig) of the plurality of devices transmitted from the air-conditioning system 100 are converted into the statistical value (Ssta) of the "same type of device" by the sensor signal processing unit 211, and the feature vector extraction unit 212 extracts the feature vector using the statistical value (Ssta) of the "same type of device".

The abnormality metric calculation unit 213 calculates an abnormality metric for each feature vector at predetermined time intervals (which may be expressed as respective times) using the feature vector of the monitoring period specified in advance. The abnormality detector 215 determines an abnormality by comparing the abnormality measure calculated in the "monitoring mode" with a threshold (SL) obtained by the threshold calculator 214 from the abnormality measure in the "learning mode". The abnormality cause identification unit 216 identifies the "cause of abnormality" of the air-conditioning system 100 by analyzing the statistical value (Ssta) of the "same-type facility" related to the abnormality, that is, the feature vector corresponding to the statistical value (Ssta).

Next, the detailed functions of the sensor signal processing unit 211, the feature vector extraction unit 212, the abnormality metric calculation unit 213, the threshold calculation unit 214, the abnormality detection unit 215, and the abnormality cause determination unit 216 will be described.

The sensor signal processing unit 211 extracts steady operation data (hereinafter, referred to as normal data) from sensor signals (Ssig) of a plurality of devices provided in the air conditioning system 100, and calculates a measurement value (Ssop) for each device as a measurement value of a sensor related to the device type. The measurement value (Ssop) is also used in the detailed analysis unit 220.

The sensor signal processing unit 211 selects the measurement value (Ssop) for each facility type, and calculates a statistical value (Ssta) of "the same-type facility", which is a statistical value aggregated for each facility type, based on the selected measurement value (Ssop). As described above, the statistical value (Ssta) is "a total value, a mean value, a maximum value, a minimum value, a standard deviation, a variance, a median value, a histogram" or the like. Of course, other statistical values may be used.

The feature vector extraction unit 212 performs a process of normalizing the statistical value (Ssta) with the average value being "0" and the variance being "1". Then, at a certain time t (t =0, 1,2, …), 1 vector x (t) is extracted every 1 time as shown in the following [ expression 1 ].

x(t)＝(x1(t)，x2(t)，…，xM-1(t)，xM(t)) ^T … … [ formula 1]]。

Here, xm (t) (M =1,2, …, M) is the M-th statistical value (Ssta) at time t after normalization.

The abnormal measure calculation unit 213 calculates a measure by using a Local Subspace method (LSC). The LSC method is widely used as an anomaly detection technique, and generates a "k-1 dimensional" partial space using k pieces of adjacent data xi (i =1, …, k) for unknown data. The method is a method for determining whether or not an anomaly is present based on unknown data and the projection distance of the "k-1 dimensional" partial space. Since the anomaly measure is represented by the projection distance between the unknown data and the "k-1 dimensional" partial space, it is sufficient to find a point on the partial space closest to the unknown data.

The threshold value calculation unit 214 ranks the abnormality metrics calculated by the abnormality metric calculation unit 213 in ascending/descending order, and sets the value of the largest abnormality metric as the threshold value (SL). According to this method, the threshold (SL) can be easily obtained. The threshold (SL) may be obtained by other methods.

The abnormality detection unit 215 determines whether or not the abnormality measure calculated by the abnormality measure calculation unit 213 exceeds a threshold (SL) set by the threshold calculation unit 214, determines that the abnormality is present if the abnormality measure exceeds the threshold (SL), and determines that the abnormality is present if the abnormality measure does not exceed the threshold (SL).

The abnormality cause identification unit 216 obtains an abnormality section in which an abnormality is continuously detected, and extracts a statistical value (Ssta) of "same type of equipment" related to the abnormality for each section based on the two-dimensional distribution density. Therefore, in the "learning mode", the two-dimensional distribution density of the normal data based on the statistical value (Ssta) of the "same kind of equipment" is calculated in a cyclic manner based on the statistical values (Ssta) of the 2 kinds of "same kind of equipment", and the two-dimensional distribution density is stored in an image format. Therefore, a plurality of two-dimensional distribution densities calculated in a cyclic manner can be obtained. The details are shown in fig. 5A and 5B.

The two-dimensional distribution density is defined as "feature space" in the claim technical solution, and is a space for representing feature vectors. The statistical value (Ssta) is represented by a feature vector, and therefore can be handled as a gray-scale image that is a normal image obtained in the "learning mode".

On the other hand, in the "monitoring mode", the degree of deviation of the statistical values (Ssta) of the 2 types of "same type devices" is calculated from where the monitoring data of the statistical values (Ssta) of the "same type devices" obtained from the plurality of measurement values is plotted on the two-dimensional distribution density image. Then, the statistical values (Ssta) of the "same type devices" related to the abnormality are extracted in order from the value having the large degree of deviation.

This means that the statistical value (Ssta) of "homogeneous equipment" related to the abnormality far from the distribution of the normal data will be found in the distribution of the abnormal data. After the statistical value (Ssta) of the "same-type facility" related to the abnormality is specified, the abnormality cause is specified by referring to the abnormality cause specification map based on the combination of the statistical values (Ssta) of the "same-type facility" (details will be described with reference to fig. 7). In general, the cause of the abnormality may be considered as, for example, refrigerant leakage, expansion valve failure, compressor failure, or the like.

Next, the function of the detailed analysis unit 220 will be described, and first, the "learning mode" in the normal state of the detailed analysis unit 220 will be described.

In the "learning mode" of the detailed analysis unit 220, the sensor signals (Ssig) of the plurality of devices transmitted from the air-conditioning equipment system 100 are input to the sensor signal processing unit 211, and the sensor signal processing unit 211 converts the sensor signals into a measurement value (Ssop) for each device. The measurement values (Ssop) are input to the abnormality cause feature amount extraction unit 221, and the abnormality cause feature amount extraction unit 221 extracts a set of measurement values (Ssop) for each device relating to "the abnormality cause". The set of measurement values (Ssop) represents a set of multiple measurement values associated with a particular device.

The abnormality cause feature amount analysis unit 222 generates a two-dimensional distribution density of normal data using a set of measurement values (Ssop) for each device relating to each abnormality cause. The two-dimensional distribution density of the normal data is generated by the same method as the two-dimensional distribution density of the statistical value (Ssta) generated by the entire analysis unit 210. The two-dimensional distribution density of the normal data is sent to the abnormality cause feature amount learning data storage section 223 and stored in the memory of the abnormality cause feature amount learning data storage section 223.

The two-dimensional distribution density of the normal data is generated in accordance with the number of devices, for example, the number of compressors, expansion valves, blower fans, and heat exchangers corresponding to the number of outdoor units, and the number of expansion valves, blower fans, and heat exchangers corresponding to the number of indoor units. Of course, if there are other devices, the device is also generated for the device.

On the other hand, in the "monitoring mode" of the detailed analysis unit 220, the sensor signals (Ssig) of the plurality of devices transmitted from the air-conditioning equipment system 100 are converted into measurement values (Ssop) for each device by the sensor signal processing unit 211. The abnormality feature amount extraction unit 221 extracts a set of measurement values (Ssop) for each device regarding the "abnormality cause", and the set of measurement values (Ssop) for each device is sent to the abnormality cause feature amount analysis unit 222.

Next, the abnormality cause feature quantity analysis unit 222 generates a two-dimensional distribution density of the monitoring data as an image using the input set of measurement values (Ssop) for each device relating to the "abnormality cause". Then, the abnormality cause feature amount learning data storage 223 selects the two-dimensional distribution density of the normal data based on the measurement value (Ssop) set for each device related to the "abnormality cause" based on the "abnormality cause" specified by the abnormality cause specifying unit 216 of the entire analysis unit 210.

Then, the two-dimensional distribution density of the monitoring data based on the measurement value (Ssop) set for each device is compared with the two-dimensional distribution density of the normal data, and "abnormal device" is specified. The two-dimensional distribution density of the monitoring data is generated in accordance with the number of devices, for example, the number of compressors, expansion valves, blower fans, and heat exchangers corresponding to the number of outdoor units, and the number of expansion valves, blower fans, and heat exchangers corresponding to the number of indoor units.

Next, the functions of the abnormality cause feature amount extraction unit 221, the abnormality cause feature amount analysis unit 222, the abnormality cause feature amount learning data storage unit 223, and the abnormality position specification unit 224 will be described.

The abnormality cause feature amount extraction unit 221 extracts a set of measurement values (Ssop) for each device associated with "the cause of abnormality" from the measurement values (Ssop) for each device obtained by the sensor signal processing unit 211.

The abnormality cause feature amount analysis unit 222 receives 2 types of measurement values (Ssop) related to the abnormality cause, which are collected from the extracted measurement values (Ssop) for each device, and generates a two-dimensional distribution density of normal data and monitor data based on a cycle combination for each device (for example, each of the compressors, expansion valves, and air blowing fans corresponding to the number of outdoor units and indoor units).

The abnormality cause feature amount learning data storage unit 223 stores the two-dimensional distribution density of the normal data obtained by the abnormality cause feature amount analysis unit 222 in a memory for each device (for example, each of the compressors, expansion valves, and blower fans corresponding to the number of outdoor units and indoor units).

The abnormal position specifying unit 224 selects the two-dimensional distribution density of the normal data based on the measurement value (Ssop) set for each facility related to the "abnormal cause" from the abnormal cause feature amount learning data storage unit 223 based on the "abnormal cause" specified by the abnormal cause specifying unit 216 of the entire analysis unit 210, and compares the two-dimensional distribution density of the normal data with the two-dimensional distribution density of the monitoring data based on the measurement value (Ssop) set for each facility related to the "abnormal cause". From the comparison result, "abnormal device" can be specified. The two-dimensional distribution density of the monitoring data is generated for each device (for example, a compressor, an expansion valve, and a blower fan corresponding to the number of outdoor units and indoor units).

Next, a configuration of an air conditioning system having a plurality of devices will be described. The air conditioning system 101 shown in fig. 3 is configured by, for example, a plurality of outdoor units 110, 120, and 130 and a plurality of indoor units 140, 150, 160, and 170. The number of the outdoor units and the indoor units is not limited, and the number thereof can be increased or decreased.

The outdoor units 110, 120, and 130 have the same types of devices 111, 121, and 131 and the same types of devices 112, 122, and 132. For example, the same-type facilities 111 to 131 are heat exchangers, and the same-type facilities 112 to 132 are compressors, and there may be a plurality of the same-type facilities.

Similarly, the indoor units 140, 150, 160, and 170 include the same types of devices 141, 151, 161, and 171, and the same types of devices 142, 152, 162, and 172. For example, the same type devices 141 to 171 are expansion valves, and the same type devices 142 to 172 are heat exchangers, and there may be a plurality of the same type devices.

The air conditioning system 101 including the outdoor units 110 to 130 and the indoor units 140 to 170 is connected by a pipe 201 through which a refrigerant circulates. Therefore, the sensors provided in the outdoor units 110 to 130 are affected by the operating states (operation/daily operation) of the indoor units 140 to 170.

On the other hand, a sensor signal (Ssig) of a sensor for detecting operation state amounts of a plurality of devices, a measurement value (Ssop) of each device obtained from the sensor signal (Ssig), and a statistical value (Ssta) of "devices of the same type" are regarded as values reflecting the operation states of the devices themselves. Fig. 4 shows an example thereof.

Fig. 4 shows the item number, the name of the feature amount, the measurement value (Ssop) of each device, and the statistical value (Ssta) of "the same kind of device". The capacity value (characteristic amount) of the equipment (compressor, heat exchanger, etc.) which cannot be directly measured by the sensor can be obtained as a simulated measurement value by a calculation formula from the operation state amount of the existing sensor which is actually installed. The abnormal state can be easily grasped using these values.

The first column in fig. 4 shows the item number, the second column shows the name of the feature value, the third column shows the measurement value (Ssop) of each device obtained by a calculation formula (for example, a fourth arithmetic operation) based on the operation state quantity of the conventional sensor, and the fourth column shows the statistical value (Ssta) of "the same type of device" obtained by summarizing the measurement values (Ssop) of each device for each of the same model.

Of the second series of characteristic values, for example, F1 is a characteristic value indicating the control of the expansion valve, F2 is a characteristic value indicating the state of the refrigerant, and F3 is a characteristic value indicating the efficiency of the compressor. The same applies to F4 to F7. In addition, more feature values can be set for these feature values.

The measurement value (Ssop) of each device in the third column is measured for each device type. For example, in the case of the characteristic amount associated with the outdoor unit, the measurement values (Ssop) of the devices corresponding to the number of the outdoor units, for example, the compressor, the heat exchanger, the expansion valve, and the blower fan are obtained. Similarly, in the case of the feature values associated with the indoor units, measurement values (Ssop) of the devices corresponding to the number of indoor units, for example, the heat exchanger, the expansion valve, and the blower fan are obtained. Of course, if the number of devices increases, the number of measurement values (Ssop) per device also increases.

The statistical value (Ssta) of the "same-type facilities" in the fourth column is obtained as a total of the "same-type facilities" for each facility type. For example, in the case where the feature value is related to the indoor unit and there are measurement values (Ssop) for each of a plurality of identical devices, the measurement values (Ssop) are collected as a statistical value (Ssta), and therefore the statistical value (Ssta) of the "identical device" is 1 type. The same feature amounts are also applied to the outdoor unit and the compressor, and the statistical value (Ssta) of "same type of equipment" is 1 type.

Note that, since the statistical value (Ssta) corresponds to the feature vector, the statistical value (Ssta) may be described below as referring to the feature vector.

The statistical value (Ssta) is defined as, for example, "total value, average value, maximum value, minimum value, standard deviation, variance, median value, histogram" or the like. For example, the feature amount F1 defines a statistical value (Ssta) as an "average value", the feature amount F2 defines a statistical value (Ssta) as an "total value", and the feature amount F3 also defines a statistical value (Ssta) as an "average value".

For example, the feature quantity relating to the heat quantity or the controlled quantity indicating the capacity of the plant is obtained by an operation of summing up the measured values (Ssop) for each of the same kind of plants, and the feature quantity indicating the state, efficiency, or index of the plant is obtained by an operation of averaging the measured values (Ssop) for each of the same kind of plants.

In this way, the statistical value (Ssta) of the measurement values (Ssop) for each of the same type of equipment may be obtained in accordance with the type and state of the feature amount. In the calculation of the statistical value (Ssta) of the "same type of equipment", it is preferable that the measurement value (Ssop) of each equipment which is a thermal shutdown condition is set to "0" when the equipment is thermally shut down (the outdoor unit is in a stopped state). This is because the signal output from the thermally-turned-off device becomes noise.

In this way, the plurality of devices have at least an operating state and a non-operating state, and the statistical value can be calculated by considering the measurement value related to the device being stopped as a value other than the measurement value, or by considering the measurement value as a value determined in advance.

Fig. 5A and 5B show examples of images of two-dimensional distribution densities of the normal data and the monitor data generated by the abnormality cause specification unit 216 of the entire analysis unit 210. Here, the same applies to the abnormality cause feature amount analysis unit 222 of the detailed analysis unit 220. In the following, statistical values are marked by "Fall" based on the description of fig. 4.

In fig. 5A, the abscissa axis represents the statistical value (Fall 1) of the feature quantity (F1) of a certain "same type of equipment" shown in fig. 4, and the ordinate axis represents the statistical value (Fall 2) of the feature quantity (F2) of a certain "same type of equipment" shown in fig. 4, which is associated with the "same type of equipment" on the abscissa axis.

Fig. 5A is a two-dimensional distribution density graph (G2 dg) in an image format of "feature space" representing a feature vector, and is a graph representing "0" of pixel values of two-dimensional distribution density as white, representing "maximum value" as black, and representing intermediate values as gray. As can be seen from this figure, the monitor data (Derr) has a position deviated from the normal data (Dnor), and an abnormality occurs.

The method of creating an image is not limited to the above method, and for example, 1 piece of data may be superimposed by assigning a gaussian distribution or another weighting filter to the data, instead of a simple frequency distribution.

Further, a maximum filter of a predetermined size, an average filter, or another weighting filter may be applied to the image obtained by the above method. In addition, the two-dimensional array may not necessarily be stored in an image format, and may be stored in a text format. The pixel values may be stored in a text format of a binary two-dimensional array in which pixels having a distribution are set to 1 and pixels having no distribution are set to 0, instead of gray shades.

Fig. 5B shows, as an example, a two-dimensional distribution density chart (G2 dg) relating to a certain 2 kinds of "same kind of equipment" related to each other in the air conditioning system 101. The statistical value (Fall 1) of the first "same type of equipment" on the horizontal axis is, for example, a characteristic quantity indicating the control of the expansion valve, and the statistical value (Fall 2) of the second "same type of equipment" on the vertical axis is, for example, a characteristic quantity indicating the state of the refrigerant.

In fig. 5B, the horizontal axis represents the minimum value (MIN) to the maximum value (MAX) of the scaled statistical values (Fall 1) of the first "same kind of equipment", and the vertical axis represents the minimum value (MIN) to the maximum value (MAX) of the similarly scaled statistical values (Fall 2) of the second "same kind of equipment".

Further, a characteristic amount (p 1) indicating the refrigerant state among characteristic amounts (t 1) indicating the control of the expansion valve as the abscissa of the first parameter is plotted as (X1, Y1), and a characteristic amount (p 2) indicating the refrigerant state among characteristic amounts (t 2) indicating the control of the expansion valve as the second parameter is plotted as (X2, Y2).

Identification information "S0001 a0001" (here, S0001 is a customer site identification number, and a0001 is an air conditioner identification number) for identifying a customer site and an air conditioner of an air conditioning system is given to the generated two-dimensional distribution density graph (G2 dg) in the image format. In this way, based on the statistical values (Fall) of all the "same kind devices" associated with the air conditioning system 101, the two-dimensional distribution density chart (G2 dg) as shown in fig. 5A and 5B after being imaged is generated for the combination of the statistical values (Fall) of the 2 kinds of "same kind devices", and is associated with the identification number.

The two-dimensional distribution density graph (G2 dg) generated in the above processing can be said to be a graph showing a correlation between the statistical values (Fall) of 2 types of "same kind of equipment". The two-dimensional distribution density graph (G2 dg) can be applied not only to the statistical value (Fall) of the "same kind of equipment" but also to the measurement value (Ssop) of each equipment. This applies to the detailed analysis unit 220 described later.

Fig. 6A shows an example of specifying "the cause of an abnormality" by using a two-dimensional distribution density chart (G2 dg) generated by using the statistical value (Fall) collected for each equipment type in the monitoring mode of the entire analysis unit 210. In particular, the outputs of the abnormality metric calculation section 213, threshold calculation section 214, abnormality detection section 215, and abnormality cause determination section 216 are shown.

In the abnormality detection table (Grp) of fig. 6A, the horizontal axis represents the passage of time (date), "AM" represents the abnormality measure generated by the abnormality measure calculation unit 213, "SL" represents the threshold value (SL) generated by the threshold value calculation unit 214, and "AD" represents the presence or absence of abnormality detection by the abnormality detection unit 215. Then, it is shown that in case the Anomaly Measure (AM) exceeds the threshold (SL), the Anomaly Detection (AD) is outputted.

In fig. 6A, for example, abnormality detection is performed continuously in the abnormal section (Wd 1) and the abnormal section (Wd 2). Then, for each abnormal section, an abnormal-related feature quantity (statistical value) corresponding to the monitoring data (Derr) in fig. 5A is extracted based on the two-dimensional distribution density graph (G2 dg) for the abnormal section in which the abnormality is continuously detected.

Fig. 6B shows two-dimensional distribution density charts (G2 dg-1) and (G2 dg-2) having the same form as that of fig. 5A, and contribution degree charts (i.e., gcnt-1) and (Gcnt-2) of abnormality-related feature quantities (statistical values) based on the two-dimensional distribution density chart (G2 dg), in which a statistical value (Fall) summarized for each equipment type is flexibly used for the abnormality detection abnormality section (Wd 1) and the abnormality section (Wd 2) in the monitoring mode.

Here, the two-dimensional distribution density graph (G2 dg-1) and the contribution degree graph (Gcnt-1) show the case of the abnormal section (Wd 1), and the two-dimensional distribution density graph (G2 dg-2) and the contribution degree graph (Gcnt-2) show the case of the abnormal section (Wd 2).

In the "learning mode", the two-dimensional distribution density graph (G2 dg) of the normal data shown in fig. 5A calculates statistics (Fall) of 2 kinds of "same kind of equipment" in a loop manner, and stores them in an image format. Therefore, a plurality of two-dimensional distribution density charts (G2 dg) calculated in a cyclic manner can be obtained. Hereinafter, this image is expressed as a "distribution density image".

Then, at the time of abnormality detection, the degree of contribution of the feature amount is calculated from where the abnormality data (Derr) shown in fig. 5A is plotted on the distribution density image, and the feature amount having a high degree of contribution is sequentially extracted as the abnormality-related feature amount. This means that a feature quantity far from the distribution of normal data will be found in the distribution of abnormal data.

In fig. 6B, extraction is performed in the order of the statistical values (Fall 7), (Fall 2), and (Fall 1) … as in the contribution graph (Gcnt-1) in the abnormal section (Wd 1), and extraction is performed in the order of the statistical values (Fall 7), (Fall 4), and (Fall 1) … as in the contribution graph (Gcnt-2) in the abnormal section (Wd 2).

Fig. 7 shows an example of the abnormality cause determination map based on a combination of statistical values (Fall) collected for each device type. The horizontal axis represents a statistical value (Fall) that is aggregated for each equipment type, that is, a characteristic quantity (for example, cooling/heating capacity, refrigerant state, compressor efficiency, etc.), and the vertical axis represents a cause of an abnormality (for example, refrigerant leakage, expansion valve failure, compressor failure, etc.). The low/high in the map indicates the state (degree of coincidence) of the abnormality-related feature quantity (abnormality statistic) based on the state of the feature quantity (normal statistic) in the "learning mode". Here, "high" means close to normal statistical value.

Then, as shown in fig. 7, the correlation between the feature amount set (the set of a plurality of feature amounts Fall1 to Fall 7) detected as an abnormality and each abnormality cause is compared, and the "abnormality cause" having a high correlation is identified. For example, when the feature amount sets detected as abnormalities are combinations of (Fall 1), (Fall 2), and (Fall 7), "abnormality cause" is AC5. That is, the abnormal state and the abnormal causes of a plurality of statistical values (feature quantities) are registered in the form of a map, and the abnormal cause is extracted from the map based on the correlation between the combinations of the abnormal states of the respective statistical values (feature quantities).

In this way, it is possible to evaluate the presence of an abnormality of a predetermined equipment type from a plurality of equipment types based on a predetermined statistical value associated with the predetermined equipment type, and to specify "the cause of the abnormality".

Next, the detailed analysis unit 220 for identifying "abnormal equipment" from the "abnormal cause" will be described. Here, the number of indoor units is shown as 3. Fig. 8 shows an example of specifying "abnormal equipment" by a two-dimensional distribution density chart generated from a measurement value set of equipment related to "abnormal cause" inputted from the whole analysis unit 210 to the detailed analysis unit 220 shown in fig. 2. The idea of the two-dimensional distribution density graph is the same as in the case of the statistical values described earlier.

For example, the abnormality cause specification map of fig. 7 is flexibly used for the abnormality section (Wd 1) and the abnormality section (Wd 2) in which abnormality is detected in the "monitoring mode" of fig. 6A, and it is assumed that the "abnormality cause" is AC5 and the "abnormality cause" is an abnormality associated with the indoor unit (the same type of equipment).

Next, an example of the two-dimensional distribution density image of the normal data and the monitor data generated by the abnormality cause feature amount analysis unit 222 of the detail analysis unit 220 will be described.

FIG. 8 shows the two-dimensional distribution density charts (G2 sdg-1), (G2 sdg-2) and the abnormality occurrence frequency charts (Gscnt-1), (Gscnt-2) of the facilities based on the two-dimensional distribution density chart (G2 sdg) in the same form as FIG. 5A.

The two-dimensional distribution density graph (G2 sdg-1) and the equipment abnormality occurrence frequency graph (Gscnt-1) show the case of the abnormal section (Wd 1), and the two-dimensional distribution density graph (G2 sdg-2) and the equipment abnormality occurrence frequency graph (Gscnt-2) show the case of the abnormal section (Wd 2).

In the "learning mode", the two-dimensional distribution density chart (G2 sdg) calculates the measurement values (Ssop) of 2 types of "same equipment" in a loop manner, and stores them in an image format. Therefore, a plurality of two-dimensional distribution densities (G2 sdg) calculated in a cycle can be obtained. This is also the case with statistical values.

Then, at the time of abnormality detection in the "monitoring mode", the occurrence of an abnormality is detected from where the abnormality data (Dserr) shown in fig. 8 is plotted in the plurality of two-dimensional distribution density charts (G2 sdg) calculated in a cyclic manner. In this case, it is understood that an abnormality has occurred in the indoor unit 3 in the abnormal section (Wd 1) and the abnormal section (Wd 2).

Further, the number of occurrences of an abnormality in the two-dimensional distribution density graph (G2 sdg) in which the measurement values (Ssop) of 2 types of "same-kind equipment" are calculated in a cyclic manner is counted, and it is considered that there is a high possibility that an abnormality occurs in the indoor unit 3 with respect to a specific "abnormal equipment" having a large number of occurrences of an abnormality, as in the abnormality occurrence number graphs (Gscnt-1) and (Gscnt-2).

That is, the two-dimensional distribution density charts (G2 sdg-1) and (G2 sdg-2) compare the normal data in the "learning mode" and the monitoring data in the "monitoring mode" when the measurement value sets of the individual devices relating to the "causes of abnormalities" are normal, count the number of occurrences of abnormalities in each indoor unit, and consider that the more the count value is, the higher the possibility of occurrence of abnormalities is. In this case, it is considered that an abnormality has occurred in the indoor unit 3. Note that, here, the "abnormal device" is captured as an indoor unit, and it is needless to say that a component device of a lower indoor unit can be targeted.

In this way, it is possible to evaluate whether or not there is an abnormality in a predetermined device from among predetermined types of devices in which an abnormality is detected, based on predetermined measurement values associated with the predetermined devices, and to specify an abnormal position (abnormal device).

According to the present embodiment described above, it is possible,

(1) By introducing the statistical value aggregated for each equipment type, even when a part of the equipment is stopped or the operation state of a part of the equipment is changed, the statistical value can be obtained, and thus the monitoring can be continued;

(2) By introducing the statistical value collected for each equipment type, it is possible to detect an abnormality by learning with a small amount of operation data, and therefore, it is possible to shorten the initial learning period;

(3) The feature space is generated with a measurement value set for each device, and the feature space is not sparse because of the reduced dimension. Therefore, it is possible to realize an operation and an effect that the normal data becomes sufficiently dense and the abnormality can be determined with high accuracy.

Next, a flow of processing of the "learning mode" executed by the processor of the monitoring computer 200 will be described. Fig. 9 shows a flow of learning processing based on measured values of individual devices and statistical values collected for each device type (the same type of device) in a normal air conditioning system.

Step S901

In step S901, the sensor signals (Ssig) of the plurality of devices transmitted from the air conditioning system 101 are first taken into the sensor signal input section 102. When the sensor signal (Ssig) is input, the process of step S902 is performed.

Step S902

In step S902, the loop process 1 is started during an arbitrary period of the "learning mode", and the following processes are repeatedly executed until the arbitrary period ends. When the loop processing 1 is started, the processing of step S903 is executed.

Step S903

In step S903, when the loop processing 1 is started in step S902, the loop processing 2 is started, data in a normal state is acquired for each sensor signal, and the following processing is repeatedly executed until the loop processing 2 is ended. When the loop processing 2 is started, the processing of step S904 is executed.

Step S904

In step S904, the measurement value for each device is obtained from the acquired data. For example, measurement values at 1 minute intervals, 1 hour intervals, and 1 day intervals are obtained. The measurement value is obtained by converting the sensor signal into a measurement value for each device by the sensor signal processing unit 211, and the measurement value set for each device relating to the "cause of abnormality" is input to the abnormality feature amount extraction unit 221. When this processing is executed, the processing of step S905 is executed.

Step S905

In step S905, the detailed analysis unit 220 starts the loop processing 3 for each abnormality factor, and repeats the following processing until the loop processing 3 is completed. When this processing is executed, the processing of step S906 is executed.

Step S906

In step S906, the detailed analysis unit 220 further starts the loop process 4 for each number of devices (for example, for each number of indoor units, for each number of outdoor units, and for each number of compressors), and repeats the following processes until the loop process 4 is completed. When this processing is executed, the processing of step S907 is executed.

Step S907

In step S907, the abnormality feature extraction unit 222 extracts the measurement value set for each device associated with the "abnormality cause", and the abnormality cause feature analysis unit 222 generates the two-dimensional distribution density of the normal data using the measurement value sets for each device associated with the abnormality causes. When this processing is executed, the processing of step S908 is executed.

Step S908

In step S908, the two-dimensional distribution density of the normal data is stored in the abnormality cause feature amount learning data storage unit 223. By this storing process, the loop process 4 ends. When this processing is executed, the processing of step S909 is executed.

Step S909

In step S909, the sensor signal processing unit 211 calculates a statistical value of the "same type of equipment" based on the measurement value of each equipment, and inputs the statistical value set of the "same type of equipment" to the abnormality cause determination unit 216. When this processing is performed, the processing of step S910 is performed.

Step S910

In step S910, the abnormality cause determination portion 216 extracts the statistical value set of "same-type devices" associated with "abnormality causes", and the abnormality cause determination portion 216 generates the two-dimensional distribution density of the normal data using the statistical value set of "same-type devices" associated with each abnormality cause. When this processing is executed, the processing of step S911 is executed.

Step S911

In step S911, the two-dimensional distribution density of the normal data is stored in the abnormality cause determination unit 217. By this storing process, the loop process 3 ends. When this processing is executed, the processing of step S912 is executed.

Step S912

In step S912, the entire analysis unit 210 inputs the statistical value set of "the same kind of equipment" to the feature vector extraction unit 212. When this processing is executed, the processing of step S913 is executed.

Step S913

In step S913, the feature vector extraction unit 212 extracts a feature vector using the statistical value of the "same kind of equipment". When this processing is executed, the processing of step S914 is executed.

Step S914

In step S914, the abnormality metric calculation unit 213 calculates an abnormality metric for each feature vector at predetermined time intervals (which may be expressed as respective times) using the feature vectors in the learning period specified in advance. When this processing is executed, the processing of step S915 is executed.

Step S915

In step S915, the threshold value (SL) corresponding to the abnormality degree is calculated by the threshold value calculation unit 214. When this processing is executed, the processing of step S916 is executed.

Step S916

In step S916, the loop process 2 ends. When this processing is performed, the processing of step S917 is performed.

Step S917

In step S917, the loop process 1 ends. When this process is executed, the "learning mode" ends.

Next, a process flow of the "monitoring mode" executed by the processor of the monitoring computer 200 will be described. Fig. 10A to 10C show a monitoring process flow based on the learned measured values of the individual devices of the air conditioning system in the monitoring state and the statistical values collected for each device type. First, description is made with reference to fig. 10A.

Step S1001

In step S1001, the sensor signals (Ssig) of the plurality of devices transmitted from the air conditioning system 101 are first taken into the sensor signal input unit 102. When the sensor signal (Ssig) is input, the process of step S1002 is executed.

Step S1002

In step S1002, the loop process 1 is started during an arbitrary period of the monitoring mode, and the following processes are repeatedly executed until the arbitrary period ends. When the loop processing 1 is started, the processing of step S1003 is executed.

Step S1003

In step S1003, when the loop processing 1 is started in step S1002, the loop processing 2 is started, data at the time of monitoring is acquired for each sensor signal, and the following processing is repeatedly executed. When the loop processing 2 is started, the processing of step S1004 is executed.

Step S1004

In step S1004, a measurement value for each device is obtained. For example, measurement values at 1 minute intervals, 1 hour intervals, and 1 day intervals are obtained. The measurement value is obtained by converting the sensor signal into a measurement value for each device by the sensor signal processing unit 211, and the measurement value set for each device associated with the "abnormality cause" is input to the abnormality feature amount extraction unit 221. When this processing is executed, the processing of step S1005 is executed.

Step S1005

In step S1005, the sensor signal processing unit 211 calculates a statistical value of the "same type of equipment" based on the measurement value for each equipment, and inputs the statistical value set of the "same type of equipment" to the feature vector extraction unit 212. When this processing is executed, the processing of step S1006 is executed.

Step S1006

In step S1006, the feature vector extraction unit 212 extracts a feature vector using the statistical value of "the same kind of equipment". When this process is executed, the process of step S1007 is executed.

Step S1007

In step S1007, the abnormality metric calculation unit 213 calculates a metric for each abnormality of the feature vector at predetermined time intervals (which may be expressed as respective times) using the feature vector in the monitoring period specified in advance. This corresponds to the Anomaly Measure (AM) shown in fig. 6A. When this processing is executed, the processing of step S1008 is executed.

Step S1008

In step S1008, the abnormality detection unit 215 compares the abnormality measure calculated in step S1007 with a threshold (SL) obtained from the abnormality measure in the "learning mode", and determines whether or not there is an abnormality based on the magnitude of the abnormality measure. This corresponds to a comparison of the Anomaly Measure (AM) with the threshold (SL) shown in fig. 6A, and if it is determined to be abnormal, an Anomaly Detection (AD) is generated. When this processing is executed, the processing of step S1009 is executed.

Step S1009

If the abnormality detection unit 216 determines in step S1009 that the abnormality remedy obtained in step S1007 is normal, the process returns to step S1003, and the processes of S1003 to S1008 are repeated for the next abnormality remedy. On the other hand, if the abnormality measure obtained in step S1007 is determined to be abnormal, the process of step S1010 is executed. The processing after step S1010 is as shown in fig. 10B.

Step S1010

In step S1010, the entire analysis unit 210 starts the loop process 3 for each abnormality factor, and repeats the following processes. When this processing is executed, the processing of step S1011 is executed.

Step S1011

In step S1011, the abnormality cause determination section 216 extracts a statistical value set of "same kind of equipment" associated with "abnormality cause". Then, a two-dimensional distribution density of the monitoring data is generated using a statistical value set of "homogeneous devices" associated with each of the heterogeneous causes. The "cause of abnormality" includes, for example, refrigerant leakage, expansion valve failure, and compressor failure. When this processing is executed, the processing of step S1012 is executed.

Step S1012

In step S1012, the loop process 3 ends. When this processing is executed, the processing of step S1013 is executed.

Step S1013

In step S1013, the abnormality cause determination section 216 compares the two-dimensional distribution density of the monitoring data with the normal data using the statistical value set of the "same kind of equipment" associated with each abnormality cause. When this processing is executed, the processing of step S1014 is executed.

Step S1014

In step S1014, the statistical value of "the same kind of equipment" as an abnormality (hereinafter referred to as an abnormality feature amount) is counted for each abnormality feature amount by the comparison in step S1013. This corresponds to the degree-of-contribution graph (Gcnt) of the abnormal feature quantity (statistical value) in fig. 6B. When this processing is executed, the processing of step S1015 is executed.

Step S1015

In step S1015, ranks are assigned in the order of the large number of abnormal feature amounts counted in step S1014. This also corresponds to the degree of contribution graph (Gcnt) of the abnormal feature quantity (statistical value) in fig. 6B, and is sorted, for example, in the manner of "Fall7", "Fall2", …. When this processing is executed, the processing of step S1016 is executed.

Step S1016

In step S1016, a predetermined upper N-bit abnormal feature amount is extracted from the sorted abnormal feature amounts. This also corresponds to the degree-of-contribution graph (Gcnt) of the abnormal feature quantity (statistical value) in fig. 6B, and for example, "Fall7", "Fall2", "Fall1", and "Fall4" are extracted. The upper N bits are determined according to the main characteristic quantity which can distinguish the abnormal reason. When this processing is executed, the processing of step S1017 is executed.

Step S1017

In step S1017, the degree of matching with the main feature amount of each anomaly is calculated. This corresponds to "low/high" assigned to each feature amount in fig. 7. When this processing is executed, the processing of step S1018 is executed.

Step S1018

In step S1018, the rate of occurrence of each anomaly (correlation) is calculated from the degree of matching of the feature values. This is calculated by evaluating the set of matching degrees shown in fig. 7. When this processing is executed, the processing of step S1018 is executed.

Step S1019

In step S1019, the sequence is given in order from the abnormality cause having a high occurrence ratio of each Abnormality Cause (AC). For example, the sorting is performed in the order of "AC5", "AC4" …. When this processing is executed, the processing of step S1020 is executed. The processing after step S1020 is as shown in fig. 10C.

Step S1020

In step S1020, the first digit of the occurrence ratio of the "abnormality cause" is determined as the "abnormality cause" from the ranking obtained in step S1019.

Step S1021

In step S1021, the abnormality cause feature amount extraction unit 222 of the detailed analysis unit 220 extracts a measurement value set for each device associated with the "abnormality cause" from the measurement values for each device stored in the sensor signal processing unit 211 based on the "abnormality cause" specified by the abnormality cause specification unit 216. When this processing is performed, the processing of step S1021 is performed.

Step S1022

In step S1021, the loop processing 4 is started for each number of devices associated with the "abnormality cause", and the following processing is repeatedly executed. When this processing is executed, the processing of step S1023 is executed.

Step S1023

In step S1023, the abnormality cause feature amount analysis unit 222 receives 2 types of the extracted measurement values for each device, which are associated with the respective abnormality causes, and generates a two-dimensional distribution density of the monitoring data based on the loop combination for each device. This corresponds to the two-dimensional distribution density graph of FIG. 8 (G2 sdg). When this processing is executed, the processing of step S1024 is executed.

Step S1024

In step S1024, the loop processing 4 ends. When this processing is performed, the processing of step S1025 is performed.

Step S1025

In step S1025, the abnormality position determination unit 224 selects, from the abnormality cause feature amount learning data storage unit 223, the two-dimensional distribution density of the normal data generated from the measurement value set for each device related to the "abnormality cause" based on the "abnormality cause" determined by the abnormality cause determination unit 216, and compares the two-dimensional distribution density of the monitoring data generated from the measurement value set for each device in the "monitoring mode". When this processing is executed, the processing of step S1026 is executed.

Step S1026

In step S1026, the number of abnormal devices is counted for each abnormal device by the comparison in step S1025. This corresponds to the abnormality occurrence number chart (Gscnt) of the device in fig. 8. Then, the device having the largest count value is determined as the "abnormal device". When this processing is executed, the processing of step S1027 is executed.

Step S1027

In step S1027, the loop process 2 ends. When this processing is executed, the processing of step S1028 is executed.

Step S1028

In step S1028, loop process 1 ends. When this processing is executed, the monitoring mode ends.

Next, a screen configuration of a display screen connected to the monitoring computer 200 will be described. Fig. 11 shows an example of a screen displaying the abnormality sign detection, the cause of the abnormality, and the abnormal device (abnormal position).

In fig. 11, a Display Screen (DSPY) is displayed on a screen of a display device connected to the monitoring computer 200. On the Display Screen (DSPY), an abnormality sign detection display frame (Dpd), an abnormality cause determination frame (Dic), and an abnormal device determination frame (Ddi) are displayed.

The abnormality detection graph (Grp) shown in fig. 6A along the passage of time is displayed in the abnormality sign detection display frame (Dpd). This makes it possible to read the time lapse of the abnormality detection, and thus to read the sign of the occurrence of the abnormality.

In the abnormality cause determination block (Dic), a two-dimensional distribution density graph (G2 dg) shown in fig. 6B and a contribution degree graph (Gcnt) of the abnormality-related feature amount are displayed. This visualizes the basis for identifying the "cause of abnormality", and thus the understanding of the user can be enhanced. Further, since a specific character display frame (SC) of the cause of the abnormality is also displayed, the cause of the failure can be clearly grasped.

The two-dimensional distribution density graph (G2 sdg) shown in fig. 8 and the abnormality occurrence number graph (Gscnt) of the device are displayed in the abnormal device determination box (Ddi). Thus, the basis for determining the "abnormal device" is visualized, and therefore the understanding of the user can be deepened. Further, since a specific character display box (SE) of the abnormal device is also displayed, it is possible to clearly grasp the "abnormal device".

The present invention is characterized in that a plurality of measured values related to a plurality of types of equipment are acquired from a plurality of sensors, measured values of the sensors related to the types of equipment are selected, statistical values are calculated based on the selected values, a feature quantity set indicating a position in a feature quantity space is calculated based on the calculated plurality of statistical values for each of the plurality of types of equipment, the feature quantity set is registered as normal data in the feature quantity space when learning is performed, whether or not an abnormality exists is determined based on a deviation of the feature quantity set from the normal data when a monitored state after learning is performed, a predetermined type of equipment is determined as a type of equipment in which an abnormality has occurred from the plurality of types of equipment when an abnormality is determined, and the equipment in which an abnormality has occurred is determined from 1 or more types of equipment which are the predetermined type of equipment.

This makes it possible to detect an abnormality in an air conditioning system including outdoor units (1 or more) and indoor units (1 or more) connected by pipes through which a refrigerant circulates, identify the cause of the abnormality, and identify which of the units has the abnormality.

The present invention is not limited to the above-described embodiments, and various modifications are possible. The above-described embodiments are described in detail to explain the present invention easily and understandably, and are not limited to having all the configurations described. In addition, a part of the structure of one embodiment may be replaced with the structure of another embodiment, and the structure of another embodiment may be added to the structure of one embodiment. The structure of each embodiment can be added, deleted, or replaced with another structure.

Claims (11)

1. An abnormality detection system having: an air conditioning system having a plurality of devices and sensors; and a monitoring computer that monitors the air conditioning system,

it is characterized in that the preparation method is characterized in that,

the plurality of devices are made up of a plurality of device classes,

the monitoring computer performs the following processing:

(1) Obtaining a plurality of measurements associated with the plurality of devices from the plurality of sensors,

(2) For the plurality of device classes:

(2A) Selecting a measurement value of the sensor associated with the equipment category,

(2B) Calculating a statistical value based on the selected measurement values,

(3) Calculating a feature quantity set representing a position within a feature quantity space based on the calculated plurality of statistical values for each of the plurality of device classes,

(4) In the case of being in learning:

(4A) Registering the feature quantity set as normal data to the feature quantity space,

(5) In the case of the learned monitoring state:

(5A) Judging whether there is abnormality based on the deviation of the feature quantity set from normal data,

(5B) If the judgment result is abnormal:

(5 Ba) identifying a predetermined device type as the device type in which the abnormality has occurred from among the plurality of device types,

(5 Bb) identifying the device having the abnormality from among 1 or more devices which are the predetermined device types.

2. The anomaly detection system of claim 1,

and (5 Ba) evaluating whether or not there is an abnormality in the predetermined device type based on the predetermined statistical value associated with the predetermined device type from among the plurality of device types, and specifying a cause of the abnormality.

3. The abnormality detection system according to claim 2,

the item (5 Bb) is a item that evaluates whether or not a predetermined device is abnormal based on a predetermined measurement value associated with the predetermined device, from among predetermined device types detected from the abnormality, and specifies the abnormal position.

4. The abnormality detection system according to claim 3,

the plurality of devices having at least an operational and a non-operational state,

the statistical value in (2B) is calculated by considering the measurement value related to the stopped device as a value other than the measurement target or a value determined in advance.

5. A monitoring computer, having a processor and an interface, for monitoring an air conditioning system,

the air conditioning system has a plurality of devices and sensors,

the plurality of devices are constituted by a plurality of kinds of devices,

the processor performs the following processing:

(1) Obtaining measurement values from the plurality of sensors via the interface,

(2) For each of the kinds of devices constituting the air conditioning system:

(2A) Selecting a measurement value of a sensor associated with the device class,

(2B) Calculating a statistical value based on the selected measurement values,

(3) Calculating a feature quantity set representing a position in the feature quantity space based on the calculated plurality of statistical values for each of the plurality of device types,

(4) In the case of learning:

(4A) Registering the feature quantity set as normal data to the feature quantity space,

(5) In the case of the post-learning monitoring state:

(5A) Judging whether there is abnormality based on the deviation of the feature quantity set from normal data,

(5B) If the judgment result is abnormal:

(5 Ba) identifying a predetermined device type as the device type in which the abnormality has occurred from among the plurality of device types,

(5 Bb) identifying the device in which the abnormality has occurred from among 1 or more devices of the predetermined device type.

6. An abnormality detection method of an abnormality detection system having: an air conditioning system having a plurality of devices for air conditioning and a plurality of sensors for measuring operation state quantities of the plurality of devices; and a monitoring computer for monitoring the operating state of the air conditioning system,

it is characterized in that the preparation method is characterized in that,

the monitoring computer performs the steps of:

acquiring a plurality of measurement values related to a plurality of types of devices from a plurality of sensors, selecting a measurement value related to the same type of device, and calculating a statistical value based on the selected measurement value;

calculating a feature quantity set indicating a position in a feature quantity space based on statistics of at least 2 types of the same kind of devices;

a step of registering the feature quantity set as normal data to a feature quantity space in a learning state;

determining a cause of abnormality based on a degree of deviation between the feature quantity set and the normal data in the case of the learned monitor state;

and identifying, from among the plurality of devices of the same type, the predetermined device of the same type as the device of the same type having the abnormality based on the cause of the abnormality, and further identifying the device having the abnormality from among the predetermined device of the same type.

7. An abnormality detection system having: an air conditioning system including a plurality of devices for air conditioning and a plurality of sensors for measuring operation state quantities of the plurality of devices; and a monitoring computer for monitoring the operating state of the air conditioning system,

it is characterized in that the preparation method is characterized in that,

the monitoring computer includes an overall analysis unit and a detailed analysis unit,

the whole analysis unit includes:

a measurement value acquisition unit that acquires a plurality of measurement values related to the plurality of devices from the plurality of sensors;

a statistical value calculation unit that calculates a statistical value of a same type of equipment among the plurality of equipments based on the measurement value related to the same type of equipment;

a statistical value distribution density calculation unit that calculates a statistical value set indicating a position of the statistical value in a statistical value two-dimensional plane in which the statistical values of 2 groups of the same kind of equipment are used as parameters;

a statistical value normal data storage unit configured to register the statistical value set as normal data in the statistical value two-dimensional plane in a learning mode;

an abnormality detection unit that determines the presence or absence of an abnormality in the same type of equipment in a monitoring mode executed after a learning mode, based on a state of deviation between the statistical value set calculated by the statistical value distribution density calculation unit and the statistical value set stored in the statistical value normal data storage unit; and

an abnormality cause specifying unit that specifies a cause of an abnormality based on a correlation between the plurality of abnormalities detected by the abnormality detecting unit,

the detailed analysis unit includes:

a measurement value distribution density calculation unit that calculates a measurement value set indicating positions of the measurement values in a two-dimensional plane of the measurement values using 2 groups of the measurement values as parameters;

a measured value normal data storage unit that registers the measured value set as normal data in the measured value two-dimensional plane in a learning mode; and

and an abnormal equipment specifying unit that selects the measurement value associated with the cause of the abnormality from the abnormality cause specifying unit in a monitoring mode executed after a learning mode, specifies the equipment of the same type associated with the cause of the abnormality based on a state of deviation between the measurement value set calculated by the measurement value distribution density calculating unit based on the selected measurement value and the measurement value set stored in the measurement value normal data storage unit, and specifies an abnormal equipment in which the abnormality has occurred from the equipment of the same type.

8. The abnormality detection system according to claim 7,

the statistical value two-dimensional plane obtained by the statistical value distribution density calculating unit obtains the number of only 2 groups of statistical values out of the plurality of statistical values calculated in a cyclic manner,

the two-dimensional plane of measurement values obtained by the measurement value distribution density calculation unit is obtained by calculating only the number of measurement values of 2 groups of the measurement values out of the plurality of measurement values in a cyclic manner.

9. The anomaly detection system of claim 7,

the statistical value distribution density calculating unit converts the statistical value into a feature vector and registers the feature vector in the statistical value two-dimensional plane in an image form,

the measured value distribution density calculation unit converts the measured value into a feature vector and registers the feature vector in an image format on the measured value two-dimensional plane.

10. The abnormality detection system according to claim 7,

the abnormality cause determination unit registers abnormality states of a plurality of statistical values and a plurality of the abnormality causes in a map format, and extracts the abnormality causes from the map according to a correlation between combinations of the abnormality states of the respective statistical values.

11. A recording medium of an abnormality detection system storing a program for operating a monitoring computer used in the abnormality detection system,

the abnormality detection system includes: an air conditioning system including a plurality of devices for air conditioning and a plurality of sensors for measuring operation state quantities of the plurality of devices; and the monitoring computer, it monitors the action state of the air conditioning system,

it is characterized in that the preparation method is characterized in that,

the recording medium includes:

a measurement value acquisition program that acquires a plurality of measurement values relating to the plurality of devices from the plurality of sensors;

a statistical value calculation program that calculates a statistical value of a same type of device among the plurality of devices, based on the measurement values related to the same type of device;

a statistical value distribution density calculation program that calculates a statistical value set representing a position of the statistical value in a statistical value two-dimensional plane having 2 sets of the statistical values related to the same kind of equipment as parameters;

a statistical value normal data storage program that registers the statistical value set as normal data to the statistical value two-dimensional plane in a learning mode;

an abnormality detection program that, in a monitor mode executed after a learning mode, determines the presence or absence of an abnormality of the same type of equipment based on a state of deviation of the statistical value set calculated by the statistical value distribution density calculation program from the statistical value set stored by the normal data storage program;

an abnormality cause specifying program that specifies a cause of an abnormality based on a correlation between the plurality of abnormalities detected by the abnormality detecting program;

a measurement value distribution density calculation program for calculating a measurement value set indicating positions of the measurement values in a two-dimensional plane of the measurement values using 2 groups of the measurement values as parameters;

a measurement value normal data storage program that registers the measurement value set as normal data in a two-dimensional plane of the measurement value in a learning mode; and

and an abnormal equipment specifying program that selects the measurement values associated with the cause of the abnormality found by the abnormal cause specifying program in a monitoring mode executed after a learning mode, specifies the same kind of equipment associated with the cause of the abnormality based on a state of deviation between the measurement value set calculated by the measurement value distribution density calculation program based on the selected measurement values and the measurement value set stored by the measurement value normal data storage program, and specifies an abnormal equipment in which an abnormality has occurred from the same kind of equipment.

Images