Method and device for detecting abnormality signs

Abstract

To easily detect a sign of an abnormality occurring in a target system based on time-series data obtained from the target system regarding an operation and a state of the target system.SOLUTION: A method for detecting abnormality signs includes: acquiring time series data for each of N types (N≥2) of feature quantities related to an operation and a state of a target system from the target system (step 101); calculating an average value of each of the time series data for each of a series of averaging periods defined in the target system (step 103); generating a data vector which is an N-dimensional vector in which the N average values calculated for each averaging period are arranged as components, and storing it (step 103); and calculating a Mahalanobis distance of the data vector corresponding to each averaging period for each averaging period based on the stored data vector (step 105).SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to a method and apparatus for detecting a sign of an abnormality occurring in a target system. [Background technology]

[0002] Water treatment systems, such as pure water production systems that produce pure water or ultrapure water from raw water or wastewater treatment systems that treat wastewater, are composed of various devices, such as filtration systems, activated carbon devices, ion exchange devices, reverse osmosis membrane devices, ultraviolet oxidation devices, coagulation sedimentation devices, and electrodeionization (EDI) devices, as well as piping, heat exchangers, and pumps. Since abnormalities can occur in the components of such water treatment systems, it is important to detect symptoms that predict the occurrence of abnormalities and identify the causes of the abnormalities during operation. Abnormalities here include not only equipment failures but also the need to replace consumables used in the system. For example, in an ion exchange device filled with ion exchange resins, the ion exchange resins must be replaced or regenerated before the adsorption capacity of impurity ions reaches saturation and breaks through the ion exchange resins. Performing ion exchange resin replacement or regeneration based on operating time or treated water volume can result in replacing or regenerating ion exchange resins that are still usable at that time, which can be cost-intensive. It is desirable to be able to detect the signs of breakthrough in ion exchange resins in order to optimize the timing of replacement or regeneration of the ion exchange resins.

[0003] A well-known method for detecting signs of abnormalities in various systems, including water treatment systems, involves placing sensors within the system and performing multivariate analysis on measurements acquired by the sensors. For example, Patent Document 1 discloses a method for detecting signs of abnormalities in a refrigeration system using helium gas, in which measurements of multiple different parameters representing the status of at least one of the refrigerator and compressor are acquired and multivariate analysis is performed on those measurements. Patent Document 1 also discloses the use of a Mahalanobis-Taguchi (MT) system for multivariate analysis. Patent Document 2 discloses a method for acquiring time-series data on multiple process variables in a water treatment system, generating anomaly detection data based on anomaly detection criteria and the time-series data, diagnosing the presence or absence of an abnormality, and, if an abnormality is determined to exist, making a rule-based judgment to identify the cause of the abnormality. Patent Document 2 also discloses that the Mahalanobis distance can be used to generate anomaly detection data.

[0004] Patent Document 3 discloses collecting measurement data for multiple items from a monitored system, such as a sewage treatment system, and calculating an evaluation index through statistical processing of the measurement data and normal data collected at a predetermined time in the past. The statistical calculation of the evaluation index uses the Mahalanobis distance between the principal component scores of the normal data obtained by principal component analysis and the principal component scores of the measurement data. Patent Document 4 discloses a method for calculating the cleaning cycle of a water-cooled water system of a refrigeration / air-conditioning equipment by defining a reference space based on water quality data of the water-cooled water system of the refrigeration / air-conditioning equipment that has already been determined to be fouling-resistant, calculating the Mahalanobis distance from the reference space for the water quality data of the water-cooled water system of the refrigeration / air-conditioning equipment to be evaluated, and determining whether the refrigeration / air-conditioning equipment is fouling-resistant or fouling-prone. Patent Document 5 discloses a method for diagnosing the corrosivity of water to metals in a system having metal-water contact points, in which the Mahalanobis distance of the water to be evaluated is calculated using data on non-corrosive water as a reference data set, with variables such as pH, conductivity, and various ion concentrations, and the corrosivity of the water to be evaluated based on the Mahalanobis distance value. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] Japanese Patent Application Laid-Open No. 2015-61993 [Patent Document 2] Japanese Patent Application Laid-Open No. 2013-61853 [Patent Document 3] Japanese Patent Application Laid-Open No. 2011-8403 [Patent Document 4] Japanese Patent Application Laid-Open No. 2006-349230 [Patent Document 5] Japanese Patent Application Laid-Open No. 2003-75325 Summary of the Invention [Problem to be solved by the invention]

[0006] As shown in Patent Documents 1-5, Mahalanobis distance is calculated based on various measurement data obtained from a water treatment system to detect signs of abnormalities in the water treatment system. However, in water treatment systems, the water supplied to the system and subjected to treatment, i.e., raw water, can fluctuate significantly in quality. When such fluctuations occur in the raw water, the measurement data obtained from the water treatment system can fluctuate significantly even if each device in the water treatment system is operating normally. When measurement data fluctuates significantly under normal conditions, even if measurement data indicating signs of abnormalities are generated, they are buried under the normal measurement data, making it difficult to detect signs of abnormalities even using Mahalanobis distance. Furthermore, water treatment systems often operate continuously for long periods of time, and detecting signs of abnormalities based on the massive amount of measurement data generated by the water treatment system during such periods presents a problem: the amount of calculation required is enormous.

[0007] An object of the present invention is to provide an abnormality sign detection method and an abnormality sign detection device that can easily detect signs of an abnormality occurring in a target system based on time series data obtained from the target system regarding the operation and status of the target system. [Means for solving the problem]

[0008] One embodiment of the abnormality sign detection method of the present invention is a method for detecting abnormality signs that detect signs of an abnormality occurring in a target system, which method acquires time series data for each of N types of quantities related to the operation and state of the target system, where N is an integer greater than or equal to 2, from the target system, calculates the average value of each of the time series data for each of a series of averaging periods defined in the target system, generates and accumulates a data vector that is an N-dimensional vector in which the N average values ​​calculated for each averaging period are arranged as components, and calculates the Mahalanobis distance of the data vector corresponding to each averaging period based on the accumulated data vector.

[0009] An abnormality sign detection device according to one embodiment of the present invention is an abnormality sign detection device that detects signs of an abnormality occurring in a target system, and includes an average value calculation unit that acquires time series data for each of N types of quantities related to the operation and state of the target system from the target system, where N is an integer greater than or equal to 2, and calculates an average value for each of the time series data for each of a series of averaging periods defined in the target system; a data storage unit that stores a data vector, which is an N-dimensional vector generated by arranging the N average values ​​calculated for each averaging period as components; and a distance calculation unit that calculates the Mahalanobis distance of the data vector corresponding to each averaging period for each averaging period based on the data vector stored in the data storage unit. [Effects of the Invention]

[0010] According to the present invention, it becomes possible to easily detect signs of abnormality in a target system based on time-series data obtained from the target system regarding the operation and state of the target system. [Brief explanation of the drawings]

[0011] [Figure 1] 1 is a flowchart illustrating an abnormality sign detection method according to an embodiment. [Figure 2] 1 is a block diagram showing an example of the configuration of an abnormality sign detection device; [Figure 3] FIG. 1 is a flow diagram illustrating an example of the configuration of a water treatment system. [Figure 4] 10(a) and 10(b) are graphs showing examples of Mahalanobis distances calculated based on time-series data from a water treatment system. DETAILED DESCRIPTION OF THE INVENTION

[0012] Next, an embodiment of the present invention will be described with reference to the drawings. First, the principle of an abnormality sign detection method according to an embodiment of the present invention will be described.

[0013] The abnormality sign detection method of this embodiment detects signs of an abnormality in a target system. While the target system is not particularly limited, this abnormality sign detection method of this embodiment is preferably used when detecting signs of an abnormality in a water treatment system, such as a pure water production system or a wastewater treatment system. This abnormality sign detection method acquires time-series data for each of multiple quantities (these multiple quantities are called feature quantities) related to the operation and status of the target system from the target system, calculates a Mahalanobis distance based on the multiple types of time-series data, and analyzes the calculated Mahalanobis distance to determine whether or not there is an abnormality sign and the cause of the abnormality. Time-series data contains hunting and noise, and it is difficult to measure each of the multiple feature quantities at exactly the same time in the target system. Therefore, in this embodiment, when time-series data is continuously input from the target system, an averaging period (also called a cycle) is determined, and the average value of each time-series data is calculated for each averaging period and used to calculate the Mahalanobis distance. The averaging period can be determined appropriately depending on the operating mode of the target system. For example, if the target system includes an ion exchange device that alternates between a water flow process and a regeneration process, one water flow process of the ion exchange device can be used as one averaging period. Of course, one water flow process can also be divided into multiple averaging periods. The lengths of the averaging periods can be different from one another.

[0014] More specifically, in the method of this embodiment, time-series data for each of N different types (N≧2) of feature quantities related to the operation and status of the target system are acquired from the target system, and the average value of each time-series data is calculated for each averaging period. Any average value can be used as long as it can represent the data values ​​within the averaging period. For example, the average value may be the center of gravity (i.e., arithmetic mean value) of the data values ​​within the averaging period, the geometric mean value, or the harmonic mean value. Alternatively, the average value may be calculated by taking the logarithm of each data value and then calculating the logarithm. However, typically, the arithmetic mean value (i.e., center of gravity value) of the data values ​​within the averaging period is used as the average value. The N average values ​​thus obtained are each a value representative of the N feature quantities of the target system for that averaging period. Considering an N-dimensional space with each of the N feature quantities as its coordinate axis, a point represented by the N average values ​​in the N-dimensional space represents the operation and status of the target system during that averaging period. This point can be represented by a data vector, which is an N-dimensional vector with the N average values ​​arranged as components. A data vector is calculated for each of a plurality of averaging periods, and the Mahalanobis distance can be calculated from the data vectors calculated in this manner.

[0015] Here, the calculation of the Mahalanobis distance MD will be explained. If a series of averaging periods are assigned sequential numbers starting from 1, then the data vector x in averaging period i is i is given by equation (1), where x i1 is the average value of the first type of time series data in the averaging period i, x i2 is the average value of the second time series data in the averaging period i, and so on for x iN is the mean value of the Nth time series data in the i-th averaging period. T represents the transpose. x i =(x i1 ,x i2 ,…,x iN ) T (1)

[0016] If μ is an N-dimensional vector indicating the average value of the data vector x over k averaging periods, μ is given by equation (2). μ = (μ1, μ2, …, μ N ) T (2) however, μ j =(x 1j +x 2j +...+x kj ) / k

[0017] If the covariance matrix calculated from the k accumulated data vectors is Σ, then the average value of N types of time series data obtained in averaging period i, that is, the data vector x i The Mahalanobis distance MD i is given by equation (3).

[0018]

number

[0019] A large Mahalanobis distance for a certain averaging period indicates that the data for that averaging period tend to deviate from the data for many other averaging periods. Therefore, by comparing the Mahalanobis distance with a predetermined threshold, it can be determined that a sign of an abnormality exists during that averaging period if the Mahalanobis distance is greater than the threshold. Furthermore, since the contribution of each of N types of feature quantities can be calculated for the calculated Mahalanobis distance, the cause of the abnormality can be estimated by determining which feature quantities have the greatest contribution. Furthermore, in this embodiment, average values ​​for new averaging periods can be calculated from time-series data over time, allowing new data vectors to be generated. When the Mahalanobis distance is recalculated, including the new data vector, the Mahalanobis distance of the newly generated data vector may be significantly greater than the Mahalanobis distance of the previously generated data vector. This strongly suggests that a new abnormality is occurring. By calculating the contribution of each feature quantity in the Mahalanobis distance corresponding to the newly generated data vector, the cause of the newly generated abnormality can be estimated.

[0020] In this embodiment, any data continuously obtained from the target system can be used as the time-series data. When the target system is a water treatment system such as a pure water production system, data such as flow rate, temperature, pressure, pH, concentration of a specific component, conductivity (or resistivity), and turbidity can be used as the time-series data. The time-series data may be obtained directly from a sensor provided in the target system, or may be obtained by performing calculations on data obtained from the sensor. For example, when an absorbance sensor is provided, the concentration of a specific component calculated from absorbance may be used as the time-series data. When two pressure sensors are provided, the differential pressure, which is the difference between the measurements from the two sensors, may be used as the time-series data.

[0021] If the target system includes an ion exchange device that repeatedly performs a water flow process and a regeneration process, the entire water flow process between the regeneration processes can be used as the averaging period, as described above. However, the interval between regeneration processes in an ion exchange device can be as short as 24 hours or as long as 2 to 3 weeks or more. If a single water flow process is used as an averaging period, the averaging period would be approximately one day to several weeks. The abnormality sign detection method of this embodiment calculates a Mahalanobis distance based on the average value calculated for each averaging period to detect the occurrence of an abnormality in the target system. A certain number of data vectors are required to calculate the Mahalanobis distance. The longer the averaging period, the longer it takes to collect the required number of data vectors, and therefore the longer it takes to detect an abnormality sign. Therefore, by dividing a water flow process into multiple sections and using each section as an averaging period, the time required to accumulate a sufficient number of data vectors for determination is shortened, thereby shortening the time required to determine whether an abnormality has occurred. Furthermore, by dividing a single water flow process into multiple averaging periods, it is possible to perform evaluation and judgment by dividing the process into sections such as the early, middle, and late stages, thereby enabling more sensitive determination of the occurrence of an abnormality. In other words, by dividing a single water flow process into multiple averaging periods, it is possible to more sensitively determine the occurrence of an abnormality and shorten the time required to determine the occurrence of an abnormality. The criteria and methods for dividing the process into multiple sections can be based on time, water flow rate, or other related parameters (e.g., time-series data such as pressure and conductivity), either alone or in combination. For example, if the regeneration process is performed at an interval of 24 hours, this can be divided into five periods, so that one water flow process consists of five averaging periods.

[0022] 1 is a flowchart illustrating an abnormality sign detection method according to one embodiment. First, in step 101, time-series data is acquired from a target system for each of N different types (N≧2) of feature quantities. In step 102, an average value within an averaging period is calculated for each time-series data of each feature quantity. In step 103, a data vector having components each consisting of the N average values ​​calculated within the averaging period is generated and stored. In step 104, it is determined whether the number of data vectors already stored and accumulated is sufficient for calculating the Mahalanobis distance. If not, the process repeats from step 101 to continue accumulating data vectors. On the other hand, if a sufficient number of data vectors have been accumulated, in step 105, the Mahalanobis distance is calculated for each accumulated data vector. In step 106, a sign of an abnormality is detected based on the calculated Mahalanobis distance. Subsequently, in step 107, it is determined whether to continue detecting a sign of an abnormality. If the detection of a sign of an abnormality is to continue, in step 108, each time-series data for the next averaging period is acquired, and in step 109, a data vector corresponding to the next averaging period is generated and stored. Thereafter, the process from step 105 is repeated. By executing steps 108 and 109, the number of accumulated data vectors increases. However, when step 105 is performed after executing step 109, the Mahalanobis distance is recalculated for all accumulated data vectors, including the increased data vectors. Alternatively, the number M of data vectors used to calculate the Mahalanobis distance may be predetermined, and when step 105 is performed, the Mahalanobis distance may be calculated using the most recent M data vectors among the accumulated data vectors. M may be, for example, 5. On the other hand, if it is determined in step 107 that the detection of a sign of an abnormality is not to be continued, the process of detecting a sign of an abnormality ends immediately.

[0023] While manual analysis of time-series data itself requires skilled techniques, the method of this embodiment automatically calculates the Mahalanobis distance and the contribution, making it possible to easily detect signs of anomalies in the target system and analyze the causes of the anomalies. The method of this embodiment also has the advantage of not requiring data (reference data) from when the target system is operating normally, making it easy to apply to predictive analysis using techniques such as machine learning.

[0024] 2 shows an example of the configuration of an abnormality sign detection device 10 that implements the abnormality sign detection method described above. This abnormality sign detection device 10 detects signs of abnormality in a target system 20 in order to manage the operation of the target system 20. The target system 20 is provided with N (N≧2) sensors 21-1 to 21-N to acquire time series data on different quantities related to the operation and state of the target system 20. The time series data on each of the N different feature quantities obtained by measurement by the sensors 21-1 to 21-N is input to the abnormality sign detection device 10.

[0025] The abnormality sign detection device 10 includes an average value calculation unit 11 that acquires time-series data for each of N feature quantities from the target system 20 and calculates an average value for each of the time-series data for each averaging period, a data storage unit 13 that stores the N average values ​​for each averaging period as a data vector, and a distance calculation unit 14 that calculates the Mahalanobis distance for each of the data vectors stored in the data storage unit 13. The average value calculation unit 11 includes a calculation unit 12 provided for each input time-series data, and the calculation unit 12 calculates an average value for the corresponding time-series data for each averaging period. When a data vector has a large Mahalanobis distance, an administrator operating the target system 20 can determine that there was a sign of an abnormality in the target system 20 during the averaging period corresponding to that data vector.

[0026] The abnormality sign detection device 10 may include a determination unit 15 that determines whether or not there is a sign of an abnormality in the target system 20 based on the calculated Mahalanobis distance, and, if there is a sign of an abnormality, estimates the cause of the abnormality. The determination unit 15 outputs a determination result indicating whether or not an abnormality has occurred and the estimated cause of the abnormality. Furthermore, by incorporating a machine learning engine that performs machine learning into the determination unit 15, it is possible to improve the accuracy of detecting signs of an abnormality and identifying the location of the abnormality.

[0027] Next, we will explain an example in which the above-mentioned abnormality sign detection method is applied to an actual system. Here, we will explain the case where the target system 20 is a water treatment system that produces pure water from raw water such as city water or well water. Figure 3 shows an example of the configuration of a water treatment system.

[0028] A storage tank 51 is provided for storing raw water, and the raw water in storage tank 51 is pumped by pump 52, and solids and other impurities are removed by filter 53. The outlet water of filter 53 is stored in storage tank 54. Storage tank 54 is also supplied with recovered water recovered from equipment that uses pure water. The water in storage tank 54 is pumped by pump 55 to activated carbon tower 56, where it undergoes activated carbon treatment. The outlet water of activated carbon tower 56 is supplied to ion exchange device 60. Ion exchange device 60 includes a first cation exchange resin tower 61 filled with cation exchange resin and supplied with the outlet water of activated carbon tower 56, a decarbonation tower 62 supplied with the outlet water of first cation exchange resin tower 61, an anion exchange resin tower 63 filled with anion exchange resin and supplied with the outlet water of decarbonation tower 62, and a second cation exchange resin tower 64 filled with cation exchange resin and supplied with the outlet water of anion exchange resin tower 63. The outlet water of the second cation exchange resin tower 64 is the outlet water of the ion exchange device 60, and this outlet water is fed by a pump 71 and supplied to a reverse osmosis membrane device 72. Pure water is discharged from the reverse osmosis membrane device 72.

[0029] In this ion exchange device 60, as a result of the ion exchange treatment in the first cation exchange resin tower 61, the outlet water of the first cation exchange resin tower 61 contains hydrogen ions (H+ ), making the water acidic, and the carbonate ions and bicarbonate ions in the outlet water are converted to free carbon dioxide. The decarbonation tower 62 is provided to remove this free carbon dioxide by blowing in air or by membrane deaeration. As a result of providing the decarbonation tower 62, water that does not contain carbon dioxide components is supplied to the anion exchange resin tower 63.

[0030] The ion exchange device 60 further includes sensors, such as a flow meter (FI) 65, conductivity meters (CI) 66 and 69, pressure gauges (PI) 67 and 68, and a resistivity meter (RI) 70, for measuring different types of characteristic quantities related to the operation and state of the ion exchange device 60 and outputting the results as time-series data. The flow meter 65 is provided at the inlet of the ion exchange device 60 and measures the flow rate of water supplied to the ion exchange device 60, i.e., the flow rate of treated water in the ion exchange device 60. The conductivity meter 66 is provided at point A, which is between the flow meter 65 and the first cation exchange resin tower 61, and measures the conductivity of the inlet water of the first cation exchange resin tower 61. The pressure gauge 67 measures the pressure at the inlet of the anion exchange resin tower 63. The pressure gauge 68 measures the pressure at the outlet of the anion exchange resin tower 63. The conductivity meter 69 is provided at point B, which is a position between the outlet of the anion exchange resin tower 63 and the inlet of the second cation exchange resin tower 64, and measures the conductivity of the outlet water of the anion exchange resin tower 63. The resistivity meter 70 is provided at point C, which is a position on the piping connected to the outlet of the second cation exchange resin tower 64, and measures the resistivity of the outlet water of the second cation exchange resin tower 64.

[0031] If the flow meter 65, the conductivity meters 66, 69, and the pressure gauges 67, 68 are provided as sensors in the ion exchange device 60, the Mahalanobis distance can be calculated based on the time-series data from these sensors in accordance with the procedure described above, and signs of abnormality can be detected in the water treatment system including the ion exchange device 60. In this case, since the difference in the pressure measurements from the pressure gauges 67, 68 indicates the water flow differential pressure in the anion exchange resin tower 63, from the viewpoint of detecting signs of abnormality in the anion exchange resin tower 63, it is preferable to generate time-series data of the water flow differential pressure from the pressure measurements from the pressure gauges 67, 68 and use this time-series data of the water flow differential pressure to calculate the Mahalanobis distance, rather than handling the time-series data from the pressure gauges 67, 68 separately.

[0032] The abnormality sign detection method of this embodiment was applied to the water treatment system shown in FIG. 3 to calculate the Mahalanobis distance. The time-series data used to calculate the Mahalanobis distance were time-series data related to the water quality in the water treatment system, namely, time-series data on the conductivity of the inlet water of the first cation exchange resin tower 61 obtained by a conductivity meter 66 installed at point A, time-series data on the conductivity of the outlet water of the anion exchange resin tower 63 obtained by a conductivity meter 69 installed at point B, and time-series data on the resistivity of the outlet water of the second cation exchange resin tower 64 obtained by a resistivity meter 70 installed at point C. One water flow process in the ion exchange device 60 was defined as one averaging period. The results are shown in FIG. 4. In FIG. 4, (a) shows the Mahalanobis distance for each of a series of averaging periods (the first to fortieth averaging periods), and (b) shows the results of calculating the contribution to each Mahalanobis distance shown in (a).

[0033] The results shown in Figure 4(a) show that the Mahalanobis distance is large during the first through fifteenth averaging periods, slightly large during the twenty-fourth through thirty-sixth averaging periods, and significantly larger during the thirty-seventh and subsequent averaging periods. These averaging periods are considered to be signs of an abnormality. Based on the contribution analysis results shown in Figure 4(b), the contribution of the conductivity of the inlet water of the first cation exchange resin tower 61 measured at point A during the first through fifteenth averaging periods is large, suggesting that a decline in the quality of the water supplied to the ion exchange device 60 is the cause of the abnormality. Similarly, the contribution of the conductivity of the outlet water of the anion exchange resin tower 63 measured at point B during the twenty-fourth through thirty-sixth averaging periods is large, suggesting that a leak in the anion exchange resin tower 63 is the cause of the abnormality. In the averaging period from the 37th onwards, the contribution of the resistivity of the outlet water of the second cation exchange resin tower 64 measured at point C is significantly large, so it can be assumed that the cause of the abnormality is breakthrough in the anion exchange resin tower 63 or the second cation exchange resin tower 64, etc. [Explanation of symbols]

[0034] 10. Anomaly detection device 11 Average value calculation section 12 Arithmetic section 13 Data storage section 14 Distance calculation unit 15 Judgment section 20 Target System 21-1~21-N Sensor 51,54 Storage tank 52, 55, 71 Pump 53 Filter 55 Activated carbon tower 60 Ion exchange device 61,64 Cation exchange resin tower 62 Decarboxylation tower 63 Anion exchange resin tower 65 Flow meter 66,69 Conductivity meter 67,68 Pressure gauge 70 Resistivity meter 72 Reverse Osmosis Membrane Device

Claims

1. An abnormality sign detection method for detecting a sign of an abnormality occurring in a target system, comprising: acquiring time-series data for each of N types of quantities related to the operation and state of the target system from the target system, where N is an integer of 2 or more; calculating an average value of each of the time series data for each of a series of averaging periods defined in the target system; generating and storing a data vector that is an N-dimensional vector in which the N average values ​​calculated for each averaging period are arranged as components; The abnormality sign detection method calculates, for each averaging period, a Mahalanobis distance of the data vector corresponding to the averaging period based on the accumulated data vector.

2. The abnormality sign detection method according to claim 1 , further comprising detecting a sign of an abnormality occurring in the target system based on the Mahalanobis distance for each of the averaging periods.

3. repeating the generation and accumulation of said data vectors for new averaging periods occurring over time; 3. The abnormality sign detection method according to claim 2, further comprising detecting a sign of an abnormality occurring in the target system based on a change in the Mahalanobis distance corresponding to the new averaging period added over time.

4. 4. The abnormality sign detection method according to claim 2, further comprising the step of analyzing the degree of contribution of the Mahalanobis distance for the averaging period in which a sign of an abnormality is detected to estimate the cause of the abnormality.

5. The abnormality sign detection method according to claim 1 , wherein the average value is expressed as a centroid value of the corresponding values ​​of the time series data within the averaging period.

6. 4. The abnormality sign detection method according to claim 1, wherein the target system is a water treatment system including at least an ion exchange device, and the time series data includes data regarding water quality within the water treatment system.

7. An abnormality sign detection device that detects a sign of an abnormality occurring in a target system, an average value calculation unit that acquires time series data for each of N types of quantities related to the operation and state of the target system from the target system, where N is an integer of 2 or more, and calculates an average value of each of the time series data for each of a series of averaging periods defined in the target system; a data storage unit for storing a data vector, which is an N-dimensional vector generated by arranging the N average values ​​calculated for each averaging period as components; a distance calculation unit that calculates, for each averaging period, a Mahalanobis distance of the data vector corresponding to the averaging period based on the data vector stored in the data storage unit; An abnormality sign detection device having the above.

8. The abnormality sign detection device according to claim 7 , further comprising a determination unit that determines whether or not a sign of an abnormality occurs in the target system based on the Mahalanobis distance for each averaging period.

9. 9. The abnormality sign detection device according to claim 7, wherein the average value calculation unit calculates, as the average value, a centroid value of the corresponding values ​​of the time series data within the averaging period.

Images