JP6803241B2 - Time series data processing device and processing method - Google Patents

Description

The present invention relates to a time series data processing apparatus and a processing method for smoothing time series data collected from a monitoring target or the like.

In a multi-loop temperature controller or the like, a device equipped with a touch panel type HMI (Human Machine Interface) may be used in combination in order to perform the operation and data collection. For example, in the example of FIG. 25, a display unit 102 with a touch panel function is provided on a display unit 101 that can be separated from the main body 100 of the 4-loop temperature controller.

In recent years, it has been required to create new value by making full use of data analysis technology. A wide variety of data analysis techniques are also applied to time-series data obtained from measurement control equipment such as temperature controllers. For example, Patent Document 1 proposes a data processing method for universally applying to arbitrary data to obtain a judgment index. The technology disclosed in Patent Document 1 acquires time-series data of three indexes of the time-operated operating rate, performance operating rate, and non-defective product rate of the production equipment to be monitored, and synthesizes the time-series data of these three indexes. Time-series data of overall equipment effectiveness, which is a comprehensive index, is generated, and a point at which a significant change appears in the value of overall equipment effectiveness is detected as a change point in the state of production equipment.

In the technique disclosed in Patent Document 1, in order to calculate the operating record (hourly operating rate, performance operating rate, non-defective product rate) of the production equipment at time t and the overall equipment effectiveness, the most recent predetermined period p minutes at time t Data (data from time t-p to time p) is used. Here, the value of p is arbitrary, but if the value of p is reduced, the time series data becomes more noisy, so that false detection of change points increases, and if the value of p is increased, the time series data is smoothed. May become too strong and the sensitivity of change point detection may deteriorate. Therefore, in the technique disclosed in Patent Document 1, the predetermined period p is set to an appropriate value based on the balance between erroneous detection and sensitivity.

In this way, when smoothing time-series data, in order to properly determine the parameters of the smoothing process, trial and error based on specialized knowledge is required, which requires complicated work for the operator. There was a problem of becoming. Such a situation that requires trial and error is necessary not only for the parameters but also for deciding what kind of processing method to perform the smoothing process.

JP-A-2015-152933

The present invention has been made to solve the above problems, and provides a time-series data processing apparatus and processing method capable of reducing the complexity of trial and error for determining a processing method of smoothing processing and parameters. The purpose is.

The time-series data processing apparatus of the present invention is configured to store in advance a data storage unit configured to store time-series data to be processed and one or more processing methods for smoothing the time-series data. A configured processing method storage unit, a display unit configured to display information, an input unit configured to accept an operator's operation, and a waveform of the time-series data to be displayed on the display unit. Based on the data display control unit configured in, and the operator's operation on the input unit, the reference trajectory output unit that outputs reference trajectory data indicating the desired trajectory of the time-series data after the smoothing process, and the smoothing process. While sequentially changing at least one of the processing method and the parameters of the smoothing process, the smoothing process is executed on the time-series data stored in the data storage unit, and the smoothing process is performed on the time-series data. It is characterized by including the processing method and a processing search execution unit configured to search at least one of the parameters whose results are most suitable for the reference locus data.
Further, one configuration example of the time-series data processing device of the present invention is further provided with a search result display control unit configured to display the processing method and parameters determined by the search on the display unit. It is a thing.

Further, one configuration example of the time-series data processing apparatus of the present invention further comprises an area division processing unit configured to divide the time domain of the time-series data stored in the data storage unit according to a predetermined procedure. The processing search execution unit is characterized in that the smoothing process is executed for each divided time domain, and at least one of the processing method and the parameters is searched for each time domain. ..
Further, in one configuration example of the time-series data processing apparatus of the present invention, the area division processing unit is characterized in that the time domain of the time-series data is evenly divided.
Further, one configuration example of the time-series data processing device of the present invention further includes a data collecting unit configured to collect time-series data to be processed from the monitored device and store it in the data storage unit. The area division processing unit is characterized in that the time area of the time series data is divided each time the state of the device to be monitored is changed.
Further, one configuration example of the time-series data processing apparatus of the present invention is a search result display control unit configured to display the processing method and parameters determined by the search on the display unit for each divided time domain. It is characterized by further providing.

Further, in one configuration example of the time series data processing apparatus of the present invention, the processing method is at least one of a processing method using a median filter and a processing method using a low-pass filter, and the parameters are the number of data of the median filter and the number of data of the median filter. At least one of the time constants of the low-pass filter.
Further, in one configuration example of the time series data processing apparatus of the present invention, the waveform of the time series data stored in the data storage unit and the waveform of the time series data after the smoothing process determined by the search are superimposed. It is characterized by further including a smoothing processing result display control unit configured to be displayed on the display unit.
Further, in one configuration example of the time series data processing device of the present invention, the display unit and the input unit are display elements with a touch panel function, and the reference locus output unit is a screen of the display element with a touch panel function. A position coordinate signal output from the display element with a touch panel function is received in response to an operator's operation, and each point on the screen indicated by the position coordinate signal has the same coordinate system as the time series data stored in the data storage unit. By converting to the above points, the reference locus data composed of a collection of the converted points is generated.

Further, the time-series data processing method of the present invention is smoothed based on the first step of displaying the waveform of the time-series data to be processed stored in the data storage unit on the display unit and the operator's operation on the input unit. Refer to the second step of generating the reference trajectory data indicating the desired trajectory of the time-series data after the conversion process, and the processing method storage unit that stores in advance one or more processing methods of the smoothing process for the time-series data. , The smoothing process is executed on the time-series data stored in the data storage unit while sequentially changing at least one of the processing method of the smoothing process and the parameters of the smoothing process, and the time-series data Is characterized by including the processing method in which the result of smoothing the data is most suitable for the reference locus data, and a third step of searching for at least one of the parameters.

According to the present invention, the waveform of the time-series data to be processed is displayed on the display unit, and the reference trajectory data indicating the desired trajectory of the time-series data after the smoothing process is generated based on the operator's operation on the input unit. Then, while sequentially changing at least one of the smoothing process processing method and the smoothing process parameters, the smoothing process is executed on the time-series data, and the result of performing the smoothing process on the time-series data is By searching for at least one of the processing methods and parameters that best match the reference trajectory data, at least one of the processing methods and parameters of the smoothing process can be appropriately determined according to the trajectory input by the operator. Therefore, it is possible to reduce the complexity of trial and error for determining the processing method and parameters.

Further, in the present invention, the time domain of the time-series data to be processed is divided by a predetermined procedure, smoothing processing is executed for each divided time domain, and at least one of the processing method and the parameter is set to the time domain. By searching each time, at least one of the processing method and the parameter of the smoothing process can be appropriately determined according to the change in the characteristics of the time-series data and the change in the trajectory input by the operator.

Further, in the present invention, by dividing the time domain of the time series data each time the state of the device to be monitored is changed, it is possible to increase the probability of adapting the smoothing process to the change in the characteristics of the time series data.

It is a figure which shows an example of the time series data collected from the monitoring target. It is a figure which shows the result of having performed the smoothing processing by the 1st order lag low-pass filter on the time series data of FIG. It is a figure which shows another result which performed the smoothing processing by the 1st order lag low-pass filter on the time series data of FIG. It is a figure which shows the result of having performed the smoothing process by the median filter on the time series data of FIG. It is a figure which shows another result which performed the smoothing process by the median filter on the time series data of FIG. It is a block diagram which shows the structure of the time series data processing apparatus which concerns on 1st Example of this invention. It is a flowchart explaining the operation of the time series data processing apparatus which concerns on 1st Embodiment of this invention. It is a figure which shows the example which an operator inputs a desirable locus of time series data after a smoothing process in 1st Example of this invention. It is a figure which shows another example which an operator inputs a desirable locus of time series data after a smoothing process in 1st Example of this invention. It is a figure which shows the display example by the smoothing processing result display control part of the time-series data processing apparatus which concerns on 1st Example of this invention. It is a figure which shows another display example by the smoothing processing result display control part of the time series data processing apparatus which concerns on 1st Example of this invention. It is a figure which shows another display example by the smoothing processing result display control part of the time series data processing apparatus which concerns on 1st Example of this invention. It is a figure which shows another display example by the smoothing processing result display control part of the time series data processing apparatus which concerns on 1st Example of this invention. It is a figure which shows the difference value of the reference locus data based on the locus shown in FIG. 8 and the difference value of the time series data by the conventional smoothing process shown in FIG. It is a figure which shows the difference value of the reference locus data based on the locus shown in FIG. 9 and the difference value of the time series data by the conventional smoothing process shown in FIG. It is a figure which shows the difference value of the reference locus data based on the locus shown in FIG. 8 and the difference value of the time series data by the smoothing process which concerns on 1st Example of this invention. It is a figure which shows the difference value of the reference locus data based on the locus shown in FIG. 9 and the difference value of the time series data by the smoothing process which concerns on 1st Example of this invention. It is a block diagram which shows the structure of the time series data processing apparatus which concerns on 2nd Example of this invention. It is a flowchart explaining the operation of the time series data processing apparatus which concerns on 2nd Embodiment of this invention. It is a figure which shows another example of the time series data collected from the monitoring target. It is a figure which shows the example which an operator inputs a desirable locus of time series data after a smoothing process in the 2nd Example of this invention. It is a figure which shows another example which an operator inputs a desirable locus of time series data after a smoothing process in the 2nd Example of this invention. It is a figure which shows another example which an operator inputs a desirable locus of time series data after a smoothing process in the 2nd Example of this invention. It is a figure which shows the display example by the smoothing processing result display control part of the time series data processing apparatus which concerns on 2nd Embodiment of this invention. It is an external view of a multi-loop temperature controller.

[Principle 1 of the invention]
As one of the typical examples of the analysis process for time series data, there is a smoothing process for removing noise components. Also in this smoothing process, it is important to determine the parameter value (number of data of interest or data time) corresponding to the predetermined period of Patent Document 1 described above. One of the purposes of the smoothing process is to grasp the essential characteristics of the measurement target and the control target from which the time series data is collected, and to monitor the characteristic change. In many cases, an operator who is skilled in monitoring the target can intuitively understand how the time-series data before smoothing should be smoothed to be convenient for monitoring the target. ..

Therefore, the inventor displays the time-series data before the smoothing process on a display with a touch panel function, and causes the operator to input the trajectory image of the time-series data after the smoothing process by tracing the touch panel. , I came up with a method of obtaining a reference trajectory for automatically determining the processing method and parameters of the smoothing process. According to this method, the processing method and the parameter values that match the intuition of the operator who is skilled in monitoring can be automatically determined, so that the complexity of trial and error can be reduced.

[Principle 2 of the invention]
When the operator inputs a desired trajectory image of the time-series data after the smoothing process, the trajectory input operation is not always performed uniformly for the entire area of the time-series data with the same smoothing feeling. The time series data itself does not always have the same characteristics over the entire area. Therefore, it is impossible to obtain a uniform processing method and parameter values for the entire time series data. On the other hand, it is reasonable to think that the smoothing feeling of the operator and the characteristics of time series data are unlikely to change frequently.

Therefore, it is preferable to simply divide the entire time series data into two, four, and eight divisions, and perform automatic determination processing separately in each area. As a result, the same processing method and the same parameter value may be obtained in a plurality of areas, but it is a good idea to divide them once. In addition, the area division is not limited to simple equal division, but the state information (mode information, etc.) of the device to be monitored is acquired from the MES (Manufacturing Execution System), etc., and each state change (each area with a different mode). A method of automatically dividing into is also conceivable. In this case, it can be expected that the probability of matching with the change in the characteristics of the time series data is high.

[First Example]
Hereinafter, examples of the present invention will be described with reference to the drawings. This embodiment is an example corresponding to the principle 1 of the above invention.

[Comparative example for verifying the effect of this example]
First, a comparative example for verifying the effect of this embodiment will be described. The example of FIG. 1 shows time-series data D1 (temperature data) collected from a monitoring target at a cycle of 0.1 seconds. In this example, the measurement noise, the high frequency fluctuation having a cycle of about 1 second, and the low frequency fluctuation having a cycle of about 12 seconds are superimposed on the time series data D1. In this embodiment, a case where the operation is limited to a single type of smoothing process is described assuming a work that is relatively not complicated for the operator. If it is a single type of smoothing process, the number of parameters to be tried and errored can be narrowed down to one, so that it can be carried out as a simple operation such as gradually increasing from a small numerical value.

The time-series data of the present invention is discrete data collected at predetermined sampling cycles, but in FIG. 1, the time-series data is represented by a continuous waveform. Similarly, in the following figures, the time series data and the data obtained by smoothing the time series data are represented by continuous waveforms.

D2 of FIG. 2 shows the result of smoothing the time-series data D1 with a first-order lag low-pass filter in order to extract the high-frequency fluctuation of the time-series data D1 of FIG. In the example of FIG. 2, the time constant of the first-order lag low-pass filter is set to 0.13 seconds.
D3 of FIG. 3 shows the result of performing smoothing processing on the time series data D1 by another first-order lag low-pass filter in order to extract the low frequency fluctuation of the time series data D1. In the example of FIG. 3, the time constant of the first-order lag low-pass filter is set to 1.8 seconds.

D4 of FIG. 4 shows the result of smoothing the time-series data D1 with a median filter in order to extract the high-frequency fluctuation of the time-series data D1. In the example of FIG. 4, the result of the median filter on a total of three data including the data of interest to be processed and the data in the vicinity thereof is used as the smoothing processing result of the data of interest.
D5 of FIG. 5 shows the result of smoothing the time-series data D1 with another median filter in order to extract the low-frequency fluctuation of the time-series data D1. In the example of FIG. 5, the result of the median filter on a total of 11 data including the data of interest to be processed and the data in the vicinity thereof is used as the smoothing processing result of the data of interest.

In the smoothing process by the low-pass filter, the attenuation of the amplitude and the phase shift of the processed data D2 and D3 with respect to the original time series data D1 are conspicuous. On the other hand, in the smoothing process by the median filter, the processed data D4 and D5 do not follow the change of the original time series data D1, and the same value is frequently generated continuously. As described above, it can be said that the results shown in FIGS. 2 to 5 leave room for improvement for the purpose of grasping the essential characteristics of the object or monitoring the changes in the characteristics.

FIG. 6 is a block diagram showing a configuration of a time series data processing apparatus according to this embodiment. The time-series data processing device includes a data collection unit 1 that collects time-series data from a monitoring target, a data storage unit 2 that stores the collected time-series data, and one or more processing methods for smoothing the time-series data. (Calculation procedure) is stored in a processing method storage unit 3 in advance, a display element 4 with a touch panel function which is a display unit for transmitting information to an operator and an input unit for receiving an operation from an operator, and a data storage unit 2. Based on the data display control unit 5 that displays the waveform of the time-series data being displayed on the display element 4 with a touch panel function and the operation of the operator, the reference trajectory data indicating the desired trajectory of the time-series data after the smoothing process is output. The smoothing process is executed for the time-series data stored in the data storage unit 2 while sequentially changing at least one of the reference locus output unit 6 and the processing method of the smoothing process and the parameters of the smoothing process. , A processing search execution unit 7 that searches for at least one of the processing methods and parameters in which the result of smoothing the time-series data best matches the reference trajectory data, and a search result display control unit 8 that displays the search results. And a smoothing process result display control unit 9 for displaying the smoothing process result.

Next, the operation of the time-series data processing apparatus of this embodiment will be described with reference to FIG. 7. The data collection unit 1 collects time-series data (for example, temperature data) from the monitoring target (step S100 in FIG. 7). The time-series data collected by the data collection unit 1 is stored in the data storage unit 2 (step S101 in FIG. 7).
Next, the data display control unit 5 causes the display element 4 with the touch panel function to display the waveform of the time series data stored in the data storage unit 2 (step S102 in FIG. 7).

The operator looks at the waveform of the time-series data displayed on the display element 4 with the touch panel function, and inputs a desired locus (waveform) of the time-series data after the smoothing process. As an input method, a method of drawing a desirable locus of time-series data after smoothing processing by moving a finger or a writing instrument or the like on the screen of the display element 4 with a touch panel function is preferable. In response to the operator's operation, the display element 4 with a touch panel function detects the position on the screen in which a finger or a writing instrument touches one after another, and outputs a position coordinate signal indicating the detected position (step S103 in FIG. 7). ..

8 and 9 are views showing how the operator moves the finger 400 on the screen 40 of the display element 4 with the touch panel function to input a desired locus of time-series data after the smoothing process. FIG. 8 shows an example in which the locus L1 is input to the time series data D1 displayed on the screen 40 with the operator conscious of extracting high frequency fluctuations. FIG. 9 shows an example in which the locus L2 is input in consideration of the operator extracting the low frequency fluctuation.

When the reference locus output unit 6 receives the position coordinate signal output from the display element 4 with the touch panel function, each point on the screen indicated by the position coordinate signal has the same coordinates as the time series data stored in the data storage unit 2. By converting to points on the system, reference locus data consisting of a collection of converted points is generated and output (step S104 in FIG. 7). Needless to say, the horizontal axis of the coordinate system of the time series data is time, and the vertical axis is the data value. The data display control unit 5 converts the time series data into points on the coordinate system of the screen and displays them on the display element 4 with a touch panel function. The reference locus output unit 6 may perform the reverse processing of the data display control unit 5.

The time series data is discrete data arranged along the time axis. Therefore, the data display control unit 5 needs to interpolate each discrete data and display the time series data as a continuous waveform. Since such interpolation processing is a well-known technique, detailed description thereof will be omitted.
Further, as will be described later, since the time series data and the reference locus data are compared by the processing search execution unit 7, it is desirable that the reference locus data is the same discrete type data as the time series data. That is, it is desirable that the reference locus output unit 6 sets the time interval of each point of the reference locus data to the same value as the sampling period of the time series data.

Next, the processing search execution unit 7 executes a smoothing process on the time-series data stored in the data storage unit 2 according to the processing method stored in the processing method storage unit 3 (step S105 in FIG. 7). In this embodiment, it is assumed that the processing method (calculation procedure) of executing the smoothing process by the first-order lag low-pass filter after executing the smoothing process by the median filter is stored in advance in the processing method storage unit 3. ..

The process search execution unit 7 calculates an error between the time-series data after the smoothing process and the reference locus data, and determines whether or not the error is equal to or less than a predetermined allowable value (step S106 of FIG. 7). When the error between the time series data after the smoothing process and the reference locus data exceeds the permissible value, the process search execution unit 7 changes at least one of the smoothing process processing method and the smoothing process parameter. (FIG. 7, step S107), the process returns to step S105, and the smoothing process for the time series data is executed again.

In this way, the process search execution unit 7 repeatedly executes the processes of steps S105 to S107 until the error between the time series data after the smoothing process and the reference locus data becomes equal to or less than the permissible value, thereby performing the smoothing process. Search for at least one of the processing methods and parameters in which the time series data best matches the reference trajectory data. As such a search method, a well-known method such as the simplex method can be applied.

In this embodiment, since one type of processing method stored in advance in the processing method storage unit 3 is fixed, the parameters of the smoothing process (for example, the number of data of the median filter and the time constant of the first-order lag low-pass filter). Will search for the optimal solution for. It is preferable that the processing method storage unit 3 stores in advance the upper and lower limit values of the parameters corresponding to the processing method of the smoothing processing. As a result, the processing search execution unit 7 changes the parameters within the range of the upper and lower limit values.

When the error between the time series data after the smoothing process and the reference locus data becomes equal to or less than the allowable value and the search ends (YES in step S106), the search result display control unit 8 names and parameters of the search result processing method. Is displayed on the display element 4 with a touch panel function (step S108 in FIG. 7).

The smoothing processing result display control unit 9 has a touch panel function so that the waveform of the time-series data after the smoothing processing determined by searching for the processing method and parameters is superimposed on the waveform of the time-series data already displayed. It is displayed on the display element 4 (step S109 in FIG. 7). Similar to the case of the data display control unit 5, the smoothing process result display control unit 9 interpolates each discrete data and displays the time-series data after the smoothing process as a continuous waveform.

Further, the smoothing processing result display control unit 9 superimposes the waveform of the reference locus data and the waveform of the time-series data after the smoothing processing determined by searching for the processing method and parameters, so that the display element with the touch panel function is used. It may be displayed in No. 4 (step S110 in FIG. 7). In this way, the processing of the time series data processing device is completed.

10 to 13 are views showing a display example by the smoothing process result display control unit 9. FIG. 10 shows the waveform of the time-series data D1 before the smoothing process and the waveform of the time-series data D6 after the smoothing process determined by searching for the processing method and parameters on the screen 40 of the display element 4 with the touch panel function. An example of displaying them in an overlapping manner is shown. The time-series data D6 after the smoothing process shows the result of searching the parameters of the smoothing process according to the locus L1 (FIG. 8) input in consideration of the extraction of the high-frequency fluctuation of the time-series data D1. As described above, in the present embodiment, the processing method of executing the smoothing process by the first-order lag low-pass filter after executing the smoothing process by the median filter is stored in advance in the processing method storage unit 3. The optimum solution of the parameters obtained as a result of the search was that the median filter had three data and the first-order lag low-pass filter had a time constant of 0.05 seconds.

FIG. 11 shows an example in which the waveform of the time-series data D1 before the smoothing process and the waveform of the time-series data D7 after the smoothing process are superimposed and displayed on the screen 40. The time-series data D7 after the smoothing process shows the result of searching the parameters of the smoothing process according to the locus L2 (FIG. 9) input in consideration of the extraction of the low frequency fluctuation of the time-series data D1. .. The optimum solution of the parameters obtained as a result of the search was that the number of data of the median filter was 11, and the time constant of the first-order lag low-pass filter was 0.35 seconds.

FIG. 12 shows an example in which the waveform of the reference locus data RD1 based on the locus L1 (FIG. 8) and the waveform of the time series data D6 after the smoothing process are superimposed and displayed on the screen 40.
FIG. 13 shows an example in which the waveform of the reference locus data RD2 based on the locus L2 (FIG. 9) and the waveform of the time series data D7 after the smoothing process are superposed on the screen 40.

It should be noted that which of the form shown in FIGS. 10 and 11 (step S109) and the form shown in FIGS. 12 and 13 (step S110) is displayed can be selected by the operation of the operator. It is also possible to select another display form. For example, although a specific example is not shown, the name of the searched processing method (in this embodiment, the median filter and the first-order lag low-pass filter) and the parameter values are displayed on the screen 40 shown in FIGS. 10 to 13. You may try to do it.

[Verification of the effect of this example]
Next, the effect of this embodiment will be verified. Here, we focus on the detection of the change point, and check the point where the sign of the differential value (strictly speaking, the difference value because it is discrete data) of the time series data becomes zero or shifts from plus to minus.

FIG. 14 shows the difference value ΔRD1 (differential value) of the reference locus data RD1 based on the locus L1 input in consideration of the extraction of the high frequency fluctuation of the time series data D1 and the time by the conventional smoothing process shown in FIG. It is a figure which shows the difference value ΔD2 (differential value) of the series data D2. RP1 is a change point of the difference value ΔRD1 (a point where the sign of the difference value becomes zero), and DP2 is a change point of the difference value ΔD2. It can be seen that even in the enlarged range as shown in FIG. 14, four more change points appear in the conventional smoothing processing result than the image of the locus L1 conscious of by the operator.

FIG. 15 shows the difference value ΔRD2 (differential value) of the reference locus data RD2 based on the locus L2 input in consideration of the extraction of the low frequency fluctuation of the time series data D1 and the conventional smoothing process shown in FIG. It is a figure which shows the difference value ΔD3 (differential value) of time series data D3. RP2 is a change point of the difference value ΔRD2, and DP3 is a change point of the difference value ΔD3. It can be seen that, even in the enlarged range as shown in FIG. 15, in the conventional smoothing processing result, 15 more change points appear than the image of the locus L2 conscious of by the operator.

FIG. 16 shows the difference value ΔRD1 (differential value) of the reference locus data RD1 based on the locus L1 input in consideration of the extraction of the high frequency fluctuation of the time series data D1, and the smoothing process of this embodiment shown in FIG. It is a figure which shows the difference value ΔD6 (differential value) of the later time series data D6. RP1 is a change point of the difference value ΔRD1, and DP6 is a change point of the difference value ΔD6. According to the smoothing processing result of this embodiment, it can be seen that the same number of change points as the image of the locus L1 conscious of the operator appears.

FIG. 17 shows the difference value ΔRD2 (differential value) of the reference locus data RD2 based on the locus L2 input in consideration of the extraction of the low frequency fluctuation of the time series data D1, and the smoothing of the present embodiment shown in FIG. It is a figure which shows the difference value ΔD7 (differential value) of the time series data D7 after processing. RP2 is a change point of the difference value ΔRD2, and DP7 is a change point of the difference value ΔD7. Similar to the case of FIG. 16, according to the smoothing processing result of this embodiment, it can be seen that the same number of change points as the image of the locus L2 conscious of the operator appears.

As described above, in this embodiment, the parameters of the smoothing process (the number of data to be focused on, the data time, the time constant) can be appropriately determined according to the locus input by the operator, so that the parameters can be determined. The complexity of trial and error can be reduced.

The processing method for obtaining the smoothing effect is not limited to the median filter and the low-pass filter, and other processing methods may be used. Other processing methods include a moving average method, a data smoothing method disclosed in JP-A-04-121621, and the like.

Further, in this embodiment, one type of processing method is stored in the processing method storage unit 3, but a plurality of selectable processing methods may be stored. When a plurality of processing methods are stored in the processing method storage unit 3, the processing search execution unit 7 changes both the processing method and the parameters within the range of the contents stored in the processing method storage unit 3. Either one of the processing method and the parameter is changed to search for the optimum solution of the processing method and the parameter. As a result, not only the parameters but also the processing method of the smoothing process can be appropriately determined, so that the complexity of trial and error for determining the processing method can be reduced.

[Second Example]
Next, a second embodiment of the present invention will be described. This embodiment is an example corresponding to the principle 2 of the above invention. FIG. 18 is a block diagram showing a configuration of a time-series data processing apparatus according to the present embodiment, and the same configuration as that of FIG. 6 is designated by the same reference numerals. The time-series data processing device of this embodiment includes a data collection unit 1, a data storage unit 2, a processing method storage unit 3, a display element 4 with a touch panel function, a data display control unit 5, and a reference trajectory output unit 6. A process search execution unit 7a that executes a time-series data smoothing process for each divided time domain and searches for at least one of the smoothing process processing method and parameters for each time domain, and the divided time. A search result display control unit 8a for displaying search results for each area, a smoothing processing result display control unit 9, and an area division processing unit 10 for dividing a time domain of time-series data according to a predetermined procedure are provided. There is.

Next, the operation of the time-series data processing apparatus of this embodiment will be described with reference to FIG. The operations of the data collecting unit 1, the data storage unit 2, the display element 4 with the touch panel function, and the reference locus output unit 6 (steps S100 to S104 in FIG. 19) are as described in the first embodiment.

FIG. 20 shows the time series data D8 (temperature data) collected from the monitored object in this embodiment. In this example, the measurement noise, the high frequency fluctuation of about 1 second cycle, and the low frequency fluctuation of about 12 seconds cycle are superimposed on the time series data D8, and the amplitude of the high frequency fluctuation becomes larger around 7 seconds. There is.

21 to 23 are diagrams showing how the operator moves a finger 400 on the screen 40 of the display element 4 with a touch panel function to input a desired locus of time-series data after smoothing processing in this embodiment. .. FIG. 21 shows an example in which the locus L8 is input to the time series data D8 displayed on the screen 40 with the operator conscious of extracting high frequency fluctuations. FIG. 22 shows an example in which the locus L9 is input in consideration of the operator extracting the low frequency fluctuation. In FIG. 23, the operator inputs the locus L10 in consideration of extracting the low frequency fluctuation without being aware of the high frequency fluctuation of the time series data D8 for the first half portion up to the time 7 seconds, and the latter half after the time 7 seconds. For the portion, an example in which the locus L11 is input with an awareness of extracting high-frequency fluctuations having a large amplitude is shown.

Next, the area division processing unit 10 divides the time domain of the time-series data stored in the data storage unit 2 according to a predetermined procedure (step S111 in FIG. 19). In this embodiment, the time domain of the time series data is evenly divided into four. As described above, the time is obtained by acquiring state information (for example, mode information of the heating device to be monitored) from MES or the like, and using the switching of the state of the device to be monitored (the point at which the mode is switched) as the time domain boundary. The time domain of the series data may be divided.

The processes of steps S105a, S106a, and S107a of the process search execution unit 7a are the same as those of steps S105, S106, and S107 of the first embodiment, respectively. However, the process search execution unit 7a performs the processes of steps S105a, S106a, and S107a for each time domain divided in step S111.

The search result display control unit 8a is divided when the error between the time-series data after the smoothing process and the reference locus data for the entire time domain becomes equal to or less than the allowable value and the search ends (YES in step S112 of FIG. 19). The name of the search result processing method and the value of the parameter are displayed on the display element 4 with the touch panel function for each time domain (step S108a in FIG. 19).

The processing of steps S109 and S110 of the smoothing processing result display control unit 9 is as described in the first embodiment. In this way, the processing of the time series data processing apparatus of this embodiment is completed.

FIG. 24 is a diagram showing a display example by the smoothing processing result display control unit 9 of this embodiment. FIG. 24 shows the waveform of the time-series data D8 before the smoothing process and the waveform of the time-series data D9 after the smoothing process determined by searching for the processing method and parameters on the screen 40 of the display element 4 with the touch panel function. An example of displaying them in an overlapping manner is shown. The time-series data D9 after the smoothing process shows the result of searching the parameters of the smoothing process according to the loci L10 and L11 (FIG. 23) input by the operator.

According to the example of FIG. 24, in the two time domains of Area1 and Area2 in the first half of the time domain divided into four, the smoothing process is performed because the operator inputs the locus L10 conscious of extracting the low frequency fluctuation. The waveform of the later time-series data D9 has a shape that substantially matches the locus L10. On the other hand, in the two time domains of Area 3 and Area 4 in the latter half, since the operator inputs the locus L11 conscious of extracting high frequency fluctuations, the smoothing process of the time series data D8 is performed using parameters different from those in the first half. Therefore, the waveform of the time-series data D9 after the smoothing process has a shape that substantially matches the locus L11.

Similar to the first embodiment, which of the embodiment described in step S109 (FIG. 24) and the embodiment described in step S110 is to be displayed can be selected by the operation of the operator. It is also possible to select another display form. For example, although a specific example is not shown, the name of the searched processing method and the value of the parameter may be displayed for each divided time domain on the screen 40 shown in FIG. 24.

Further, the search result display control unit 8a and the smoothing processing result display control unit 9 have the time domain division lines TL1, TL2, and TL3 as shown in FIG. 24, for example, in order to make the operator recognize the divided time domain. The area numbers Area1, Area2, Area3, and Area4 may or may not be displayed.

The time-series data processing device described in the first and second embodiments may be mounted inside the controller, or may be provided separately from the controller. Further, it is a general matter to output the time-series data after the smoothing process from the time-series data processing device and use the time-series data. Needless to say, the present invention can be applied to various usage forms of time series data after smoothing.
Further, in the first and second embodiments, the temperature data is described as an example of the time series data, but it goes without saying that the time series data is not limited to the temperature data.

Among the time-series data processing units described in the first and second embodiments, the data collection unit 1, the data storage unit 2, the processing method storage unit 3, the data display control unit 5, the reference trajectory output unit 6, and the processing search execution. The units 7, 7a, the search result display control units 8, 8a, the smoothing processing result display control unit 9, and the area division processing unit 10 are a computer provided with a CPU (Central Processing Unit), a storage device, and an interface, and these. It can be realized by a program that controls hardware resources. The CPU executes the processes described in the first and second embodiments according to the program stored in the storage device.

The present invention can be applied to a technique for smoothing time series data.

1 ... Data collection unit, 2 ... Data storage unit, 3 ... Processing method storage unit, 4 ... Display element with touch panel function, 5 ... Data display control unit, 6 ... Reference trajectory output unit, 7, 7a ... Processing search execution unit, 8, 8a ... Search result display control unit, 9 ... Smoothing processing result display control unit, 10 ... Area division processing unit.

Claims (10)

A data storage unit configured to store time-series data to be processed,
A processing method storage unit configured to store one or more processing methods for smoothing the time-series data in advance, and a processing method storage unit.
A display unit configured to display information and
An input unit configured to accept operator operations,
A data display control unit configured to display the waveform of the time series data on the display unit, and
A reference locus output unit that outputs reference locus data indicating a desired locus of time-series data after smoothing processing based on an operator's operation on the input unit.
While sequentially changing at least one of the processing method of the smoothing process and the parameters of the smoothing process, the smoothing process is executed on the time series data stored in the data storage unit, and the time series data is converted into the time series data. A time-series data process comprising the process method in which the result of the smoothing process best matches the reference locus data, and a process search execution unit configured to search at least one of the parameters. apparatus. In the time series data processing apparatus according to claim 1,
A time-series data processing apparatus further comprising a search result display control unit configured to display the processing method and parameters determined by the search on the display unit. In the time series data processing apparatus according to claim 1,
Further, an area division processing unit configured to divide the time domain of the time series data stored in the data storage unit according to a predetermined procedure is further provided.
The processing search execution unit executes the smoothing process for each divided time domain, and searches for at least one of the processing method and the parameters for each time domain. .. In the time series data processing apparatus according to claim 3,
The area division processing unit is a time-series data processing device characterized in that the time domain of the time-series data is evenly divided. In the time series data processing apparatus according to claim 3,
A data collection unit configured to collect time-series data to be processed from a device to be monitored and store it in the data storage unit is further provided.
The area division processing unit is a time-series data processing device that divides the time domain of the time-series data each time the state of the device to be monitored is switched. In the time series data processing apparatus according to any one of claims 3 to 5,
A time-series data processing apparatus further comprising a search result display control unit configured to display the processing method and parameters determined by the search on the display unit for each divided time domain. In the time series data processing apparatus according to any one of claims 1 to 6.
The processing method is at least one of a processing method using a median filter and a processing method using a low-pass filter.
The time-series data processing apparatus, wherein the parameter is at least one of the number of data of the median filter and the time constant of the low-pass filter. In the time series data processing apparatus according to any one of claims 1 to 7.
A smoothing process configured to display on the display unit by superimposing the waveform of the time series data stored in the data storage unit and the waveform of the time series data after the smoothing process determined by the search. A time-series data processing device including a result display control unit. In the time series data processing apparatus according to any one of claims 1 to 8.
The display unit and the input unit are display elements with a touch panel function.
The reference locus output unit receives a position coordinate signal output from the display element with a touch panel function in response to an operator's operation on the screen of the display element with a touch panel function, and outputs each point on the screen indicated by the position coordinate signal. , Time-series data processing characterized in that the reference locus data consisting of a collection of converted points is generated by converting to points on the same coordinate system as the time-series data stored in the data storage unit. apparatus. The first step of displaying the waveform of the time-series data to be processed stored in the data storage unit on the display unit, and
Based on the operator's operation on the input unit, the second step of generating reference trajectory data indicating the desired trajectory of the time-series data after the smoothing process, and
With reference to the processing method storage unit that stores one or more processing methods of the smoothing process for the time series data in advance, while sequentially changing at least one of the processing method of the smoothing process and the parameters of the smoothing process. , The processing method and the parameter whose time series data stored in the data storage unit is subjected to smoothing processing and the result of performing the smoothing processing on the time series data is most suitable for the reference locus data. A time-series data processing method including a third step of searching at least one of them.