JP7455667B2 - Sensor drift diagnosis device, sensor drift diagnosis method, and sensor drift diagnosis program - Google Patents

Description

Embodiments of the present invention relate to a sensor drift diagnosis device, a sensor drift diagnosis method, and a sensor drift diagnosis program.

Accurate temperature and humidity control is important in spaces such as clean rooms in semiconductor manufacturing plants. Therefore, it is necessary to measure the temperature by disposing a large number of temperature sensors and humidity sensors in the space. These sensors may exhibit values that deviate from accurate values over time, and this is generally referred to as sensor "drift." Administrators need to check for deviations in measured values from the reference sensor (hereinafter referred to as "reference sensor") for each sensor to be managed, and periodically perform maintenance work to calibrate such deviations. Therefore, a great deal of effort was required for maintenance work. To address this problem, conventionally, there have been cases in which an estimation model for estimating the measured value of a sensor is constructed and the estimated value is compared with the measured value of a reference sensor to reduce the maintenance load.

However, spaces that require temperature and humidity control, such as clean rooms, are often large spaces with many air conditioners and heat generating equipment, so temperature and humidity settings must be reviewed regularly, and equipment may need to be replaced. This may occur in some cases. Therefore, even if an estimation model is constructed for a space where such environmental changes occur, there is no guarantee that the model will be applicable at the time when drift diagnosis is desired. For example, even if an estimation model is created in which the difference between two temperature sensors becomes 2°C at a certain point in time, if the amount of operation of the heat generating equipment in the space increases, the temperature difference may rise to 3°C. Further, if the flow rate of air blown out from the air conditioner changes due to a change in the temperature setting of the air conditioner, the temperature difference may be reduced to 1°C.

That is, in the method of calculating the drift amount using an estimation model, it is necessary that the environmental factors that affect the sensor do not change from the premise of the estimation model at the time the estimation model is used. This is difficult in spaces where there are many air conditioners and heat generating devices, and where temperature settings are regularly reviewed, and is one of the reasons why it is difficult to apply a method that uses an estimation model to calculate the amount of drift. ing.

Japanese Patent Application Publication No. 2014-153957 Japanese Patent Application Publication No. 2018-124069 Special table 2014-528065 publication Patent No. 4436227

The problem to be solved by the present invention is to provide a sensor drift diagnosis device, a sensor drift diagnosis method, and a sensor drift diagnosis program that can more accurately diagnose the drift of a sensor that measures the temperature or humidity of a space. .

The sensor drift diagnosis device of the embodiment includes an absolute humidity calculation section, a reference data generation section, a drift score calculation section, and a drift diagnosis section. The absolute humidity calculation unit calculates an actual measured value of absolute humidity of the target space based on the measured values of the temperature and relative humidity of the target space. The reference data generation unit generates reference data, which is data used to diagnose drift of the sensor that measures the temperature and relative humidity of the space, and provides a comparison target for the measured values of the sensor. . The drift score calculation unit calculates an index value regarding the drift based on the reference data. The drift diagnosis section diagnoses drift of the sensor based on the index value.

FIG. 2 is a block diagram showing a specific example of the functional configuration of the sensor drift diagnosis device in the first embodiment. FIG. 3 is a diagram showing a specific example of a space in which sensors to be diagnosed by the sensor drift diagnosis device are installed in the first embodiment. 2 is a flowchart illustrating a specific example of a process in which the sensor drift diagnosis device 1 of the first embodiment diagnoses the drift of each temperature and humidity sensor. FIG. 4 is a diagram showing a specific example of a temperature change in a space during a learning period in the first embodiment. FIG. 3 is a diagram showing a specific example of a change in humidity in a space during a learning period in the first embodiment. FIG. 7 is a diagram showing a specific example of a change in absolute humidity of a space during a learning period in the first embodiment. FIG. 4 is a diagram showing a specific example of a temperature change in a space during a diagnosis period in the first embodiment. FIG. 3 is a diagram showing a specific example of a change in humidity in a space during a diagnosis period in the first embodiment. FIG. 3 is a diagram showing a specific example of a change in absolute humidity of a space during a diagnosis period in the first embodiment. FIG. 3 is a diagram showing a specific example of a change in the absolute humidity (estimated value) of a space during a diagnosis period in the first embodiment. FIG. 2 is a block diagram showing a specific example of the configuration of a sensor drift diagnosis device according to a second embodiment. A histogram showing a specific example of a probability distribution generated for a learning period in the second embodiment. 7 is a histogram showing a specific example of a probability distribution generated for the diagnosis period in the second embodiment. FIG. 7 is a block diagram showing a specific example of the functional configuration of a sensor drift diagnosis device in a third embodiment. FIG. 7 is a diagram showing a specific example of the relationship between the Jensen-Shannon distance and the amount of drift in the third embodiment. FIG. 7 is a block diagram showing a specific example of the functional configuration of a sensor drift diagnosis device according to a fourth embodiment. FIG. 7 is a diagram showing a specific example of temperature change during a learning period in a situation where multiple drift occurs in the fourth embodiment. FIG. 7 is a diagram showing a specific example of humidity change during the learning period in a situation where multiple drift occurs in the fourth embodiment. FIG. 7 is a diagram showing a specific example of a change in absolute humidity during a learning period in a situation where multiple drift occurs in the fourth embodiment. FIG. 7 is a diagram showing a specific example of a change in absolute humidity during a diagnosis period in a situation where multiple drifts occur in the fourth embodiment. FIG. 7 is a diagram showing a specific example of a change in absolute humidity (estimated value) when multiple drift cannot be detected in the fourth embodiment. The figure which shows the specific example of the scatter diagram created about the time difference data of each temperature-humidity sensor in 4th Embodiment. FIG. 7 is an psychrometric diagram illustrating the correlation between temperature and absolute humidity when drift occurs in the fourth embodiment. FIG. 7 is a block diagram showing a specific example of the functional configuration of a sensor drift diagnosis device according to a fifth embodiment. 12 is a flowchart illustrating a specific example of a process in which the drift score calculation unit calculates the amount of drift of a temperature sensor and a humidity sensor when multiple drifts occur in the sensor drift diagnosis device of the fifth embodiment.

Hereinafter, a sensor drift diagnosis device, a sensor drift diagnosis method, and a sensor drift diagnosis program according to embodiments will be described with reference to the drawings.

(First embodiment)
FIG. 1 is a block diagram showing a specific example of the functional configuration of a sensor drift diagnosis device 1 according to the first embodiment. The sensor drift diagnosis device 1 includes a CPU (Central Processing Unit), a memory, an auxiliary storage device, etc. connected via a bus, and executes a program. The sensor drift diagnosis device 1 functions as a device including an absolute humidity calculation section 11, an estimation model construction section 12, a drift score calculation section 13, and a drift diagnosis section 14 by executing a program. Note that all or part of each function of the sensor drift diagnosis device 1 may be realized using hardware such as an ASIC (Application Specific Integrated Circuit), a PLD (Programmable Logic Device), or an FPGA (Field Programmable Gate Array). good. The program may be recorded on a computer-readable recording medium. The computer-readable recording medium is, for example, a portable medium such as a flexible disk, magneto-optical disk, ROM, or CD-ROM, or a storage device such as a hard disk built into a computer system. The program may be transmitted via a telecommunications line.

FIG. 2 is a diagram showing a specific example of a space in which sensors to be diagnosed by the sensor drift diagnosis device 1 are installed. For example, the sensor drift diagnosis device 1 diagnoses the drift of four temperature and humidity sensors 2a, 2b, 2c, and 2d installed in a space A as shown in FIG. The number of sensors to be diagnosed may be three or less, or five or more. It is assumed that the temperature and humidity of space A are controlled by the outdoor conditioner 3 and air conditioners 4a, 4b, 4c, and 4d. For example, it is assumed that the humidity in space A is controlled by the dew point temperature control of the outdoor air conditioner 3. Further, it is assumed that the plurality of air conditioners 4 installed on the ceiling of the space A can be controlled independently. In addition, it is assumed that equipment (not shown) that generates heat is operating in the space A, and the temperature distribution in the room can change from moment to moment.

Each temperature/humidity sensor 2 is equipped with a thermometer and a hygrometer on its upper part, and is capable of measuring the temperature/humidity at the same point. Further, each temperature and humidity sensor 2 is connected to a data collection device (not shown), and continuously outputs temperature and humidity measurement data to the data collection device. For example, it is assumed that each temperature/humidity sensor 2 outputs measurement data every minute, and the data collection device sequentially records the measurement data output from each temperature/humidity sensor 2. Hereinafter, humidity when simply written as "humidity" or "temperature and humidity" means "relative humidity", and absolute humidity will be explicitly written as "absolute humidity".

The absolute humidity calculation unit 11 has a function of calculating the absolute humidity of the space A based on the temperature and humidity of the space A measured by each temperature and humidity sensor 2. The absolute humidity calculation unit 11 outputs information indicating the calculated absolute humidity to the estimation model construction unit 12 in the subsequent stage.

The estimated model construction unit 12 has a function of constructing a drift score calculation model that calculates the drift score of each temperature/humidity sensor 2 based on the calculated absolute humidity of the space A. The drift score here is a value representing the degree of drift occurring in each temperature/humidity sensor 2, and is used as an index value for determining whether each temperature/humidity sensor 2 should be calibrated. Needless to say, the drift score may be the actual drift amount of each temperature/humidity sensor 2 (namely, the temperature drift amount or the humidity drift amount). The details and construction method of the drift score calculation model will be described later. The estimated model construction unit 12 outputs the constructed drift score calculation model to the drift score calculation unit 13.

The drift score calculation unit 13 calculates the drift score of each temperature/humidity sensor 2 by applying the measured value of each temperature/humidity sensor 2 at the diagnosis target time to the constructed drift score calculation model. The drift score calculation unit 13 outputs information indicating the calculated drift score to the drift diagnosis unit 14.

The drift diagnosis section 14 diagnoses whether or not each temperature/humidity sensor 2 needs to be calibrated based on the calculated drift score. The drift diagnosis section 14 outputs information regarding diagnosis results (hereinafter referred to as "diagnosis result information"). The diagnosis result information may include information indicating whether or not calibration is necessary, as well as supplementary information indicating a drift score, a drift amount, and the like.

FIG. 3 is a flowchart showing a specific example of a process in which the sensor drift diagnosis device 1 of the first embodiment diagnoses the drift of each temperature/humidity sensor 2. First, the absolute humidity calculation unit 11 inputs measurement data for constructing a drift score calculation model from each temperature and humidity sensor 2 (step S101), and calculates each temperature and humidity sensor 2 in space A based on the input measurement data. The absolute humidity at the installation position is calculated (step S102). The absolute humidity calculation unit 11 outputs information indicating the calculated absolute humidity to the estimation model construction unit 12.

Hereinafter, the temperature and humidity measurement data input in step S101 will be referred to as "learning measurement data", and the period for acquiring the learning measurement data will be referred to as "learning period". For example, the learning period is a period during which it is assumed that no sensor drift occurs, such as one month after calibration. FIG. 4 is a diagram showing a specific example of temperature change in space A during the learning period, and FIG. 5 is a diagram showing a specific example of humidity change in space A during the learning period. As can be seen from FIGS. 4 and 5, in the space A during the learning period, the temperature and humidity change significantly at the installation position of the temperature/humidity sensor 2c.

On the other hand, FIG. 6 is a diagram showing a specific example of changes in absolute humidity in space A during the learning period. As can be seen from FIG. 6, even at the installation position of the temperature/humidity sensor 2c, where the relative humidity showed a large change, the absolute humidity did not change significantly. In this way, even when the relative humidity and temperature change significantly, the amount of moisture in the air (ie, absolute humidity) does not change significantly.

For example, the absolute humidity calculation unit 11 can calculate the absolute humidity by applying the learning measurement data to the following equations (1) to (4).

Here, the absolute humidity at each installation point of the temperature/humidity sensors 2a to 2d to be calculated is represented by AHa[t], AHb[t], AHc[t], and AHd[t], respectively. In addition, the temperatures measured by the temperature and humidity sensors 2a to 2d at time t are expressed as Ta[t], Tb[t], Tc[t], and Td[t], respectively, and the temperatures measured by the temperature and humidity sensors 2a to 2d at time t are The relative humidity is expressed as RHa[t], RHb[t], RHc[t], and RHd[t], respectively. mixr is a known function that calculates absolute humidity based on temperature and relative humidity. If information indicating the atmospheric pressure of the space A can be acquired, the absolute humidity calculation unit 11 may calculate the absolute humidity using a function that uses the atmospheric pressure as a variable in addition to the temperature and relative humidity.

Note that in the following, when explaining the processing common to each temperature/humidity sensor 2, the subscripts a to d may be omitted. For example, when explaining similar processing using RHa[t], RHb[t], RHc[t], and RHd[t], they may be collectively written as RH[t] in the calculation formula. be.

Further, a value indicating a predetermined learning period may be set in advance in the sensor drift diagnosis device 1, and the sensor drift diagnosis device 1 inputs the learning period when executing the drift score calculation model construction process. It may be configured to accept. The absolute humidity calculation unit 11 outputs information indicating the calculated absolute humidity to the estimation model construction unit 12.

Next, the estimation model construction unit 12 calculates the absolute humidity at the installation position of each temperature/humidity sensor 2 based on the learning measurement data and the absolute humidity of the space A calculated by the absolute humidity calculation unit 11. An absolute humidity estimation model for estimation based on the absolute humidity at the installation position of the humidity sensor 2 is constructed (step S103). Hereinafter, building this absolute humidity estimation model may be referred to as "learning absolute humidity."

For example, the estimation model construction unit 12 uses a regression that calculates the absolute humidity at the installation position of a certain temperature/humidity sensor 2 as an output variable and the absolute humidity at the installation position of another temperature/humidity sensor 2 as an input variable. Build the model as an absolute humidity estimation model. For example, a model for estimating the absolute humidity at the installation position of the temperature/humidity sensor 2a can be defined as a linear regression model shown in equation (5) below.

Here, α, β, and γ represent weighting coefficients for each input variable, and δ represents a constant. Further, T_learn represents a learning period. The estimation model construction unit 12 also constructs absolute humidity estimation models for the temperature and humidity sensors 2b, 2c, and 2d other than the temperature and humidity sensor 2a using the same method. Note that the estimation model construction unit 12 constructs an absolute humidity estimation model that includes, as input variables, other measured values, estimated values, or predicted values regarding the space A, in addition to the absolute humidity at the installation position of each temperature and humidity sensor 2. It's okay. The estimated model construction unit 12 outputs the absolute humidity estimation model generated in this way to the drift score calculation unit 13. The above is the learning process for absolute humidity.

In addition, although the case where the absolute humidity estimation model is constructed as a linear regression model is explained here, the absolute humidity estimation model calculates the absolute humidity at the installation position of a certain temperature/humidity sensor 2 by calculating the absolute humidity at the installation position of another temperature/humidity sensor 2. It is not limited to a specific model as long as it is estimated based on the absolute humidity at the installation location. Therefore, a model other than a linear regression model such as a neural network may be used as the absolute humidity estimation model.

Once the absolute humidity estimation model is constructed, the absolute humidity calculation unit 11 inputs temperature and humidity measurement data at the desired timing from each temperature and humidity sensor 2 in order to diagnose drift based on the absolute humidity estimation model. (Step S201), the absolute humidity at the installation position of each temperature/humidity sensor 2 in space A is calculated based on the input measurement data (step S202). The absolute humidity calculation unit 11 outputs information indicating the calculated absolute humidity to the drift score calculation unit 13.

Hereinafter, the temperature and humidity measurement data input in step S201 will be referred to as "diagnostic measurement data", and the period for acquiring the diagnostic measurement data will be referred to as "diagnosis period". FIG. 7 is a diagram showing a specific example of the temperature change in space A during the diagnosis period, and FIG. 8 is a diagram showing a specific example of the humidity change in space A during the diagnosis period. As can be seen from FIGS. 7 and 8, in the space A during the diagnosis period, the temperature and humidity change significantly at the installation position of the temperature/humidity sensor 2d.

On the other hand, FIG. 9 is a diagram showing a specific example of a change in absolute humidity in space A during the diagnosis period. As can be seen from FIG. 9, similarly to the learning period example, the change in absolute humidity is smaller than the change in relative humidity (that is, the moisture content itself has not changed significantly).

On the other hand, the drift score calculation unit 13 estimates each absolute humidity that has changed as shown in FIG. 9 during the diagnosis period using the absolute humidity estimation model (step S203). As described above, the absolute humidity estimation model is a model that estimates the absolute humidity at one measurement point based on the absolute humidity at other measurement points. Specifically, the drift score calculation unit 13 applies each absolute humidity that has changed as shown in FIG. 9 to an absolute humidity estimation model constructed for each measurement point (installation position of each temperature and humidity sensor 2). Calculate the estimated value of absolute humidity at each measurement point. FIG. 10 is a diagram showing a specific example of the change in the absolute humidity of space A during the diagnosis period, estimated in this way. The drift score calculation unit 13 outputs the estimated value of absolute humidity calculated for each measurement point to the drift diagnosis unit 14.

Subsequently, the drift diagnosis section 14 determines whether each temperature/humidity sensor A drift score of 2 is calculated (step S204). Here, as an example, a case will be described in which the drift amount of absolute humidity is used as the drift score. In this case, the drift diagnosis unit 14 can calculate the drift score using the calculation formulas shown in equations (6) to (9) below.

Here, AHDa, AHDb, AHDc, and AHDd represent drift scores for the temperature and humidity sensors 2a, 2b, 2c, and 2d, respectively. In addition, AHa ^~ [t], AHb ^~ [t], AHc ^~ [t], and AHd ^~ [t] are the estimated values of the absolute humidity at time t at the installation positions of the temperature and humidity sensors 2a, 2b, 2c, and 2d, respectively. represents. AHa ^~ [t], AHb ^~ [t], AHc ^~ [t], and AHd ^~ [t] are calculated based on the respective absolute humidity estimation models. Furthermore, in equations (6) to (9), tεT_test, and N is the data length (the number of data acquired for each measurement point). T_test represents the diagnosis period. That is, according to equations (6) to (9), the average value of the difference between the estimated value and the actual value of absolute humidity during the diagnosis period is output as a drift score.

Next, the drift diagnosis unit 14 calculates the maximum temperature drift amount, the maximum humidity value, and the lift amount for each temperature/humidity sensor 2 based on the calculated drift score (here, the absolute humidity drift amount) (step S205 ). Here, the maximum temperature drift amount is the temperature drift amount when it is assumed that the absolute humidity matches the estimated value on the premise that the humidity is not drifting. Specifically, when calculating the maximum temperature drift amount TDa for the temperature/humidity sensor 2a, a value of TDa that satisfies the following equation (10) may be calculated.

The drift diagnosis unit 14 calculates the maximum temperature drift amounts TDb, TDc, and TDd for the temperature and humidity sensors 2b, 2c, and 2d in the same manner as for the temperature and humidity sensor 2a.

On the other hand, the maximum humidity drift amount is the humidity drift amount when it is assumed that the absolute humidity matches the estimated value on the premise that the temperature is not drifting. Specifically, when calculating the maximum humidity drift amount RHDa for the temperature/humidity sensor 2a, it is sufficient to calculate RHDa that satisfies the following equation (11).

The drift diagnosis unit 14 calculates the maximum humidity drift amounts RHDb, RHDc, and RHDd for the temperature and humidity sensors 2b, 2c, and 2d using the same method as for the temperature and humidity sensor 2a.

The drift diagnosis unit 14 compares the maximum drift amount of temperature and humidity calculated in this way with a predetermined threshold value to determine whether or not calibration by maintenance is necessary for each temperature and humidity sensor 2 (step S206). The drift diagnosis section 14 generates and outputs diagnosis result information based on this determination result. Note that the presence or absence of drift may be diagnosed based on the maximum drift amount of either temperature or humidity, or may be diagnosed based on the maximum drift amount of both.

The sensor drift diagnosis device 1 of the first embodiment configured as described above constructs an estimation model for absolute humidity at each measurement point based on past temperature and humidity measurement data, and performs calculations using the constructed estimation model. The temperature/humidity sensor 2 is configured to diagnose drift of each temperature/humidity sensor 2 by comparing the estimated absolute humidity value with the actual measured value. That is, the sensor drift diagnosis device 1 of the first embodiment detects the drift of a sensor that measures the temperature or humidity of a space in which the amount of moisture in the air is controlled, based on the absolute humidity of the controlled space. It is used for diagnosis. By having such a configuration, the sensor drift diagnosis device 1 of the first embodiment can detect the temperature and humidity of each temperature and humidity sensor 2 even when the temperature and humidity change greatly as shown in FIGS. 7 and 8. It becomes possible to diagnose drift more accurately.

Although the drift diagnosis unit 14 has calculated the maximum drift amount here, the drift score calculation unit 13 may calculate the maximum drift amount and provide this to the drift diagnosis unit 14 as a drift score.

(Second embodiment)
FIG. 11 is a block diagram showing a specific example of the configuration of the sensor drift diagnosis device 1a in the second embodiment. The sensor drift diagnosis device 1a differs from the sensor drift diagnosis of the first embodiment in that it includes a probability distribution generation section 15 instead of the estimation model construction section 12, and a drift score calculation section 13a instead of the drift score calculation section 13. Unlike the device 1, the other configurations are the same as the sensor drift diagnosis device 1 of the first embodiment. Therefore, the same reference numerals as in FIG. 1 are given to the same components as in the first embodiment, and the explanation thereof will be omitted.

The estimation model construction unit 12 of the first embodiment constructed an absolute humidity estimation model based on the measurement data of temperature and humidity. In contrast, the probability distribution generation unit 15 of the second embodiment generates a probability distribution in which a deviation from the representative value occurs for each measured value of absolute humidity. Here, a case will be described in which the representative value is the median value, but the representative value may be a statistical value other than this. Specifically, first, the probability distribution generation unit 15 calculates the median value of absolute humidity in the learning period and the diagnosis period for each time. In this case, the median value of absolute humidity during the learning period and the diagnosis period is calculated by the following equations (12) and (13).

Here, median is a function that calculates the median value of the elements of the four series at each time t. Subsequently, the probability distribution generation unit 15 generates correction data by subtracting the median value from each value of absolute humidity for each period of the learning period and the diagnosis period. Equations (14) to (17) below represent equations for calculating correction data for the learning period, and equations (18) to (21) represent equations for calculating correction data for the diagnosis period.

Next, the probability distribution generation unit 15 generates probability distributions of the correction data AHa', AHb', AHc', AHd' for the learning period and the correction data AHa", AHb", AHc", AHd" for the diagnosis period, respectively. . For example, if this probability distribution is expressed as a histogram, the probability of occurrence for each bin can be determined by dividing the fluctuation range of absolute humidity into a predetermined number and dividing the number of data included in each fluctuation range by the total number of data. . 12 and 13 are histograms showing specific examples of probability distributions generated for the learning period and the diagnosis period. The probability distribution generation unit 15 outputs data indicating the probability distribution generated in this way (hereinafter referred to as “probability distribution data”) to the drift score calculation unit 13a.

The drift score calculation unit 13a calculates a drift score based on the probability distribution data generated by the probability distribution generation unit 15. Specifically, the drift score calculation unit 13a calculates a drift score for each temperature/humidity sensor 2 by comparing the probability distributions of the learning period and the diagnosis period. For example, the drift score calculation unit 13a may calculate the difference between the modes of the respective probability distributions as the drift score. In other words, the drift score in this case can be said to be the amount of drift that is considered to be highly likely to have occurred from the perspective of probability, and its calculation formula can be expressed as in the following equations (22) to (25). I can do it.

The drift score calculation unit 13a outputs the drift score calculated in this way to the drift diagnosis unit 14. For example, in the examples of FIGS. 12 and 13, the drift scores of the temperature and humidity sensors 2c and 2d are large values, so it is diagnosed that the temperature and humidity sensors 2c and 2d require calibration.

Note that in this case, instead of the mode, it is also possible to use statistical values such as an average value or a median value that can be representative values. In addition, to create a probability distribution, it is possible to use not only a histogram but also a fitting of a normal distribution or a Gaussian mixture model. In this case, the drift score calculation unit 13a may convert the difference between the representative values (for example, the mode) calculated above into a temperature drift amount or a humidity drift amount using a predetermined relational expression, and use the value as the drift score. .

The sensor drift diagnosis device 1a of the second embodiment configured in this way can diagnose the drift of each temperature/humidity sensor 2 from changes in the distribution of the probability that a deviation from a representative value (for example, the median value) will occur. can. In the sensor drift diagnosis device 1a of the second embodiment, there is no need to construct an absolute humidity estimation model, so the processing load is reduced or the accuracy of diagnosis is increased compared to the sensor drift diagnosis device 1 of the first embodiment. There is a possibility that it can be done.

(Third embodiment)
FIG. 14 is a block diagram showing a specific example of the functional configuration of the sensor drift diagnosis device 1b in the third embodiment. The sensor drift diagnosis device 1b differs from the sensor drift diagnosis device 1a of the second embodiment in that it includes a drift score calculation unit 13b instead of the drift score calculation unit 13a, and the other configurations are the same as the sensor drift of the second embodiment. This is similar to the diagnostic device 1a. Therefore, the same components as those in the second embodiment are designated by the same reference numerals as in FIG. 11, and the description thereof will be omitted.

The drift score calculation unit 13 of the second embodiment calculated the drift score based on changes in the probability distribution during the learning period and the diagnosis period. On the other hand, the drift score calculation unit 13b of the third embodiment quantifies the difference in tendency of the measurement data between the learning period and the diagnosis period using the distance between the probability distributions, and calculates the value as the drift score. Calculated as

The Kullback-Leibler distance and the Jensen-Shannon distance are known methods for determining the distance of a probability distribution. For example, the Kullback-Leibler distance when the probability distribution is discrete is expressed by the following equation (26).

Here, P(i) and Q(i) are the probabilities that the value selected according to the probability distributions P and Q is i. Further, in this case, the Jensen-Shannon distance is expressed by the following equation (27).

From the above equation, it is assumed that the Jensen-Shannon distance takes a large value as the amount of drift increases, but it does not directly indicate the amount of drift. Therefore, when determining whether or not the temperature/humidity sensor 2 needs to be calibrated, it is necessary to set a threshold value in advance and compare it with the threshold value. This threshold value can be determined using, for example, past analysis results. For example, the Jensen-Shannon distance is calculated using the above method in a period T_prelearn in which no drift occurs and in a period T_pretest in which the amount of drift is known. FIG. 15 is a diagram showing a specific example of the relationship between the Jensen-Shannon distance and the amount of drift.

According to the example in FIG. 15, drift occurs in temperature and humidity sensors 2c and 2d whose Jensen-Shannon distance is equal to or greater than the threshold value 0.22, and temperature and humidity sensor 2a whose Jensen-Shannon distance is equal to or less than 0.12. No drift occurred in and 2b. In such a case, for example, drift can be diagnosed using a threshold value of 0.17, which is between 0.22 and 0.12.

(Fourth embodiment)
FIG. 16 is a block diagram showing a specific example of the functional configuration of the sensor drift diagnosis device 1c in the fourth embodiment. The sensor drift diagnosis device 1c differs from the sensor drift diagnosis device 1 of the first embodiment in that it includes an estimated model construction section 12c instead of the estimated model construction section 12, and the other configuration is the same as the sensor drift diagnosis device 1 of the first embodiment. This is similar to the diagnostic device 1. Therefore, the same reference numerals as in FIG. 1 are given to the same components as in the first embodiment, and the explanation thereof will be omitted.

The estimation model construction unit 12 of the first embodiment diagnosed the drift of the temperature and humidity sensor 2 using the absolute humidity estimation model. However, in this case, if the temperature sensor and humidity sensor simultaneously drift, the drift may not be detected. FIGS. 17 to 20 are diagrams showing specific examples of temperature/humidity data and absolute humidity acquired in such a situation. 17 to 20 assume a situation in which a similar temperature and humidity sensor 2e is added in addition to the temperature and humidity sensors 2a to 2d described in the first embodiment.

Moreover, FIG. 21 is a diagram showing changes in absolute humidity estimated for the diagnosis period by constructing an absolute humidity estimation model similar to that in the first embodiment for the temperature/humidity sensor 2e. In this example, there appears to be no difference between the measured value and the estimated value (that is, no drift). However, this situation is one in which the temperature sensor and the humidity sensor are drifting at the same time, so that the difference in absolute humidity coincidentally shows the correct value. As described above, with the diagnostic method of the first embodiment, it may not be possible to correctly diagnose that both the temperature sensor and the humidity sensor are drifting.

Therefore, the estimation model construction unit 12c of the fourth embodiment determines the presence or absence of drift in such a situation using the following method. First, the estimated model construction unit 12c generates time difference data dAHa[t] and dTa[t] for AH[t] and T[t], respectively. For example, time difference data regarding the measurement data AHa[t] and Ta[t] of the temperature and humidity sensor 2a can be generated using the following calculation formulas (28) and (29). The estimated model construction unit 12c generates time difference data for the other temperature and humidity sensors 2b to 2e using similar calculation formulas (30) to (37).

Next, the estimated model construction unit 12c creates a scatter diagram of the time difference data calculated for each temperature/humidity sensor 2. FIG. 22 is a diagram showing a specific example of a scatter diagram created for the time difference data of each temperature/humidity sensor 2. The numerical value written in the upper right corner of each scatter diagram is the value of the correlation coefficient between dAH and dT for each distribution shown, and represents the slope of the distribution shown in the scatter diagram. Here, in space A where the amount of moisture in the air is controlled, there should be no (or small) change in absolute humidity with respect to temperature change. However, a correlation occurs in the temperature and humidity sensors 2 (temperature and humidity sensors 2d and de in the example of FIG. 22) where drift occurs. FIG. 23 is an psychrometric chart for explaining this reason. FIG. 23 is an psychrometric diagram showing the relationship between temperature, absolute humidity, and humidity.

The example in FIG. 23 shows that in state P11 where the temperature is 23°C and the humidity is 50%, even if the temperature fluctuates by ±1°C, only the temperature changes while the absolute humidity remains constant (P12-P11- (see curve P13). In this case, if the temperature sensor drifts by +2°C, if the relative humidity is constant, the state change of P12-P11-P13 (hereinafter referred to as "state change 1") will change to P22-P21-P23 (hereinafter referred to as "state change 2"). ). Therefore, in this case, absolute humidity has a negative correlation with temperature.

Similarly, when the humidity sensor drifts by -5.4%, if the temperature is constant, the state change of P12-P11-P13 changes to P32-P31-P33 (hereinafter referred to as "state change 3"). Therefore, in this case as well, absolute humidity has a negative correlation with temperature. Further, when the temperature sensor drifts by +2° C. and the humidity sensor drifts by -5.4%, the state change of P12-P11-P13 changes to P42-P41-P43 (hereinafter referred to as "state change 4"). Therefore, in this case as well, absolute humidity has a negative correlation with temperature.

In this way, when the temperature sensor and the humidity sensor are drifting at the same time, for example, when the temperature and humidity change as shown in state change 4 in the example of FIG. 23, the difference between the estimated value and actual value of absolute humidity during the diagnosis period Even if there is no difference between the two, drift may actually occur in both the temperature sensor and the humidity sensor (hereinafter referred to as "multiple drift"). However, even in such a case, it has been found that the presence or absence of drift appears as a correlation between absolute humidity and temperature, as described above.

In the sensor drift diagnosis device 1c of the fourth embodiment, the estimation model construction unit 12c has a function of checking the presence or absence of such a correlation, so that it is possible to detect the occurrence of multiple drift. Further, such a multiple drift detection function is applicable not only to the sensor drift diagnostic device 1 of the first embodiment, but also to the sensor drift diagnostic device 1a of the second embodiment or the sensor drift diagnostic device 1b of the third embodiment. may be prepared for.

Note that as an index value for determining whether there is a correlation between absolute humidity and temperature, a coefficient of a regression model or a correlation coefficient of dAH and dT may be calculated. In addition, the threshold value for determining the presence or absence of drift may be determined based on, for example, the regression coefficient of the temperature/humidity sensor 2 in which no drift has occurred, and its representative value (e.g., average value) and standard deviation. good.

In addition, here, the estimation model construction unit 12c analyzed the correlation between absolute humidity and temperature, but instead of this, the drift score calculation unit 13 analyzes the correlation, and uses the analysis result as a drift score for drift diagnosis. 14.

It has been explained that the sensor drift diagnostic device 1c of the fourth embodiment differs from the sensor drift diagnostic device 1 of the first embodiment in that it includes the estimated model construction section 12c instead of the estimated model construction section 12. Here, the estimation model construction unit 12c is configured to have a function of diagnosing drift using the sensor drift diagnosing method of the first embodiment in addition to a function of diagnosing drift using the sensor drift diagnosing method of the fourth embodiment. It's okay. With such a configuration, the estimation model construction unit 12c first diagnoses the presence or absence of drift using the sensor drift diagnosis method of the first embodiment, and when it is diagnosed that there is no drift, the estimation model construction unit 12c performs the fourth implementation. The presence or absence of multiple drift can be diagnosed using the above sensor drift diagnosis method.

(Fifth embodiment)
FIG. 24 is a block diagram showing a specific example of the functional configuration of the sensor drift diagnosis device 1d of the fifth embodiment. The sensor drift diagnosis device 1d differs from the sensor drift diagnosis device 1c of the fourth embodiment in that it includes a drift score calculation unit 13d instead of the drift score calculation unit 13, and the other configuration is the same as the sensor drift of the fourth embodiment. This is similar to the diagnostic device 1c. Therefore, the same reference numerals as in FIG. 16 are given to the same configurations as those in the fourth embodiment, and the explanation thereof will be omitted here.

The drift score calculation unit 13d calculates the respective drift amounts of the thermometer and the hygrometer using AH and dT calculated by the estimation model construction unit 12c. Specifically, in a situation where no drift occurs, the correlation between dAH and dT is zero as described above, so by changing the temperature drift amount TD and the humidity drift amount RHD, the correlation becomes zero. Find the TD and RHD.

FIG. 25 is a flowchart showing a specific example of a process in which the drift score calculation unit 13d calculates the drift amounts of the temperature sensor and humidity sensor when multiple drifts occur in the sensor drift diagnosis device 1d of the fifth embodiment. First, the drift score calculation unit 13d sets TD=0 (step S301). Next, the drift score calculation unit 13d calculates T[t]=T[t]+TD, and calculates RHD where AH ^~ [t]-mixr(T[t],RH[t]+RHD) is the minimum. (Step S302).

Next, the drift score calculation unit 13d calculates dT' and dRH' from T'[t]=T[t]+TD and RH'[t]=RH[t]+RHD to obtain a correlation coefficient ( Step S303), and whether the correlation is positive or negative is determined (Step S304).

If it is determined in step S304 that the correlation is negative (step S304: negative), the drift score calculation unit 13d adds a minute amount ΔTD to TD (step S305). For example, ΔTD=0.1°C. On the other hand, if it is determined in step S304 that the correlation is positive (step S304: positive), the drift score calculation unit 13d subtracts the minute amount ΔTD from TD (step S306).

Subsequently, the drift score calculation unit 13d determines whether the termination condition is satisfied (step S307). For example, the termination condition may be that the correlation has become close to zero, or that the number of loops from steps S302 to S307 has exceeded the maximum number. If it is determined in step S307 that the termination condition is not satisfied (step S307-NO), the drift score calculation unit 13d returns the process to step S302. On the other hand, if it is determined in step S307 that the termination condition is satisfied (step S307-YES), the drift score calculation unit 13d outputs the TD and RHD at that time as temperature and humidity drift amounts (step S308), and processes end.

The sensor drift diagnostic device 1d of the fifth embodiment configured as described above can detect multiple drifts of the temperature/humidity sensor 2 and calculate the amount of drift of each of the temperature sensor and the humidity sensor. In the first embodiment, the maximum drift amount of each temperature/humidity sensor 2 was determined, but in this case, even if the drift amount is actually less than the maximum drift amount, it may be misdiagnosed that calibration is required. There was sex. On the other hand, according to the sensor drift diagnosis device 1d of the fifth embodiment, the amount of drift can be calculated for each of the temperature sensor and the humidity sensor, so it is possible to reduce the possibility of the above diagnosis. becomes.

According to at least one embodiment described above, the sensor drift diagnosis device of the embodiment includes an absolute humidity calculation unit that calculates the actual measured value of absolute humidity of the target space based on the measured values of the temperature and relative humidity of the target space. and a reference data generation unit that generates reference data that is data used for diagnosing drift of a sensor that measures the temperature and relative humidity of the space, and that provides a comparison target for the measured values of the sensor. By having a drift score calculation unit that calculates an index value regarding the drift based on the reference data, and a drift diagnosis unit that diagnoses the drift of the sensor based on the index value, the temperature or humidity of the space can be determined. sensor drift can be diagnosed more accurately. Here, the estimated model construction section and the probability distribution generation section in the above embodiment are examples of the reference data generation section. Furthermore, the absolute humidity estimation model and probability distribution data in the above embodiments are examples of reference data. Further, the learning period in the above embodiment is an example of a reference time. Further, the diagnosis period in the above embodiment is an example of the time of diagnosis.

Note that the sensor drift diagnosis device of the embodiment does not necessarily need to be configured as one device. The sensor drift diagnosis device of the embodiment may be configured by a plurality of devices that can communicate via a network. In this case, each functional unit included in the sensor drift diagnosis device of the embodiment may be configured to be distributed among a plurality of devices. For example, the absolute humidity calculation section, the estimation model construction section, the probability distribution generation section, the drift score calculation section, and the drift diagnosis section may each be implemented in different devices. For example, the sensor drift diagnosis device of the embodiment includes a learning device that learns absolute humidity (that is, constructs an absolute humidity estimation model), and a learning device that diagnoses the drift of a temperature/humidity sensor based on the estimated value and actual measured value of absolute humidity. The diagnostic device may also be configured to be distributed between the diagnostic device and the diagnostic device.

Although several embodiments of the invention have been described, these embodiments are presented by way of example and are not intended to limit the scope of the invention. These embodiments can be implemented in various other forms, and various omissions, substitutions, and changes can be made without departing from the gist of the invention. These embodiments and their modifications are included within the scope and gist of the invention as well as within the scope of the invention described in the claims and its equivalents.

1, 1a, 1b, 1c, 1d...Sensor drift diagnostic device, 2, 2a, 2b, 2c, 2d, 2e...Other temperature and humidity sensors, 3...Outside conditioner, 4, 4a, 4b, 4c, 4d...Air conditioning 11... Absolute humidity calculation section, 12, 12c... Estimation model construction section, 13, 13a, 13b, 13d... Drift score calculation section, 14... Drift diagnosis section, 15... Probability distribution generation section.

Claims (11)

an absolute humidity calculation unit that calculates an actual measured value of absolute humidity of the target space based on measured values of temperature and relative humidity of the target space;
The data used for diagnosing the drift of a sensor that measures the temperature and relative humidity of the target space, the data for providing a comparison target for the measured values of the sensor , is the absolute humidity of the target space. a reference data generation unit that generates reference data based on actual measured values ;
a drift score calculation unit that calculates an index value regarding the drift based on the reference data;
a drift diagnosis unit that diagnoses drift of the sensor based on the index value;
Equipped with
The reference data generation unit constructs an absolute humidity estimation model that indicates the relationship between absolute humidity at the installation positions of the plurality of sensors ,
The drift diagnosis unit diagnoses the drift of the sensor by comparing the absolute humidity estimated by the absolute humidity estimation model with the actual measured value of absolute humidity calculated based on the measured values of temperature and relative humidity. do,
Sensor drift diagnostic device. an absolute humidity calculation unit that calculates an actual measured value of absolute humidity of the target space based on measured values of temperature and relative humidity of the target space ;
The data used for diagnosing the drift of a sensor that measures the temperature and relative humidity of the target space , the data for providing a comparison target for the measured values of the sensor , is the absolute humidity of the target space. a reference data generation unit that generates reference data based on actual measured values ;
a drift score calculation unit that calculates an index value regarding the drift based on the reference data ;
a drift diagnosis unit that diagnoses drift of the sensor based on the index value;
Equipped with
The reference data generation unit generates a first probability distribution regarding absolute humidity at the reference time and a second probability distribution regarding absolute humidity at the time of diagnosis,
The drift score calculation unit calculates a drift amount of the temperature or humidity of the sensor as the index value by comparing the first probability distribution and the second probability distribution.
Sensor drift diagnostic device. The drift score calculation unit calculates a statistical distance between the first probability distribution and the second probability distribution as the index value.
The sensor drift diagnostic device according to claim 2 . The drift diagnosis unit calculates a maximum drift amount of temperature or humidity for the sensor based on the actual measured value and estimated value of the absolute humidity,
The drift diagnosis unit diagnoses the drift of the sensor based on the maximum drift amount of the temperature or humidity.
The sensor drift diagnostic device according to any one of claims 1 to 3 . The reference data generation unit analyzes the correlation between temperature and absolute humidity based on the distribution of time differences between temperature and absolute humidity , and diagnoses multiple drift of the sensor based on the analysis result.
The sensor drift diagnostic device according to any one of claims 1 to 4 . The drift score calculation unit calculates a drift amount of each of the temperature and relative humidity based on the correlation analysis result.
The sensor drift diagnostic device according to claim 5 . The target space is a space where the amount of moisture in the air is controlled to a predetermined amount;
The sensor drift diagnostic device according to any one of claims 1 to 6 . an absolute humidity calculation step of calculating an actual measured value of the absolute humidity of the target space based on the measured values of the temperature and relative humidity of the target space;
The data used for diagnosing the drift of a sensor that measures the temperature and relative humidity of the target space, the data for providing a comparison target for the measured values of the sensor , is the absolute humidity of the target space. a reference data generation step of generating reference data based on actual measured values ;
a drift score calculation step of calculating an index value regarding the drift based on the reference data;
a drift diagnosis step of diagnosing drift of the sensor based on the index value;
has
In the reference data generation step, constructing an absolute humidity estimation model that indicates the relationship between absolute humidity at the installation positions of the plurality of sensors ;
In the drift diagnosis step, the drift of the sensor is diagnosed by comparing the absolute humidity estimated by the absolute humidity estimation model with the actual measured value of absolute humidity calculated based on the measured values of temperature and relative humidity. ,
Sensor drift diagnosis method. Based on the measured values of the temperature and relative humidity of the target space, the actual value of the absolute humidity of the target space is determined. an absolute humidity calculation step;
Used to diagnose drift of sensors measuring temperature and relative humidity of the target space. the data for providing a comparison target for the measured value of the sensor; Standard data generation that generates standard data based on the actual measured value of absolute humidity in the target space. step and
Drift score calculation for calculating an index value regarding the drift based on the reference data step and
a drift diagnosis step of diagnosing drift of the sensor based on the index value;
has
In the reference data generation step, a first probability distribution regarding absolute humidity at the reference time; generating a second probability distribution regarding absolute humidity at the time of diagnosis;
In the drift score calculation step, the first probability distribution and the second probability distribution By comparing with cloth, the amount of drift of temperature or humidity of the sensor can be determined as the index value. calculate,
Sensor drift diagnosis method. an absolute humidity calculation step of calculating an actual measured value of the absolute humidity of the target space based on the measured values of the temperature and relative humidity of the target space;
The data used for diagnosing the drift of a sensor that measures the temperature and relative humidity of the target space, the data for providing a comparison target for the measured values of the sensor , is the absolute humidity of the target space. a reference data generation step of generating reference data based on actual measured values ;
a drift score calculation step of calculating an index value regarding the drift based on the reference data;
a drift diagnosis step of diagnosing drift of the sensor based on the index value;
A program for causing a computer to execute
The reference data generation step includes constructing an absolute humidity estimation model that indicates the relationship between absolute humidity at the installation positions of the plurality of sensors,
The drift diagnosis step diagnoses the drift of the sensor by comparing the absolute humidity estimated by the absolute humidity estimation model with the actual measured value of absolute humidity calculated based on the measured values of temperature and relative humidity. including doing;
Sensor drift diagnostic program. Based on the measured values of the temperature and relative humidity of the target space, the actual value of the absolute humidity of the target space is determined. an absolute humidity calculation step;
Used to diagnose drift of sensors measuring temperature and relative humidity of the target space. the data for providing a comparison target for the measured value of the sensor; Standard data generation that generates standard data based on the actual measured value of absolute humidity in the target space. step and
Drift score calculation for calculating an index value regarding the drift based on the reference data step and
a drift diagnosis step of diagnosing drift of the sensor based on the index value;
A program for causing a computer to execute
The reference data generation step includes a first probability distribution regarding absolute humidity at the reference time and a first probability distribution regarding absolute humidity at the time of diagnosis. and generating a second probability distribution for the absolute humidity of
The drift score calculation step includes calculating the first probability distribution and the second probability distribution. By comparing, the amount of drift of the temperature or humidity of the sensor is calculated as the index value. including
Sensor drift diagnostic program.

Images