JP2024046277A - Encoding method - Google Patents

Abstract

[Problem] It is possible to generate an image that can learn and infer correlation between multiple signals. [Solution] The encoding method is a computer-executed encoding method for encoding the output of multiple output devices, and includes an acquisition step of acquiring a collected data group that is the output of the multiple output devices, a verification step of generating a verified data group that is verified data for each time period and for each output device based on the collected data group, and an image generation step of creating an encoded image that encodes the verified data of two or more output devices included in the multiple output devices based on the verified data group. [Selected Figure] Figure 2

Description

The present invention relates to an encoding method.

A variety of sensors have traditionally been used in manufacturing facilities to automate production and control. In recent years, a variety of sensors are also being used in a variety of places other than manufacturing facilities, such as offices and ordinary homes. Measurement values from sensors are used not only for control but also for status monitoring and anomaly detection. Patent Document 1 discloses a system that detects patterns from time-series data.

US Patent Application Publication No. 2019/379589

In the invention described in Patent Document 1, only one signal can be processed.

The encoding method according to the first aspect of the present invention is a computer-executed encoding method for encoding the output of a plurality of output devices, and includes an acquisition step of acquiring a collected data group that is the output of the plurality of output devices, a verification step of generating a verified data group that is verified data for each time period and for each of the output devices based on the collected data group, and an image generation step of creating an encoded image that encodes the verified data of two or more of the output devices included in the plurality of output devices based on the verified data group.

According to the present invention, it is possible to generate an image that allows learning and inference of correlation between a plurality of signals.

Diagram showing usage scenes of arithmetic device Functional configuration diagram of arithmetic unit Hardware configuration diagram of the computing device Diagram showing an example of a device list A diagram showing an example of collected data Diagram showing an example of preprocessed data A diagram showing an example of verified data Diagram showing an example of image parameters Diagram showing an example of required specifications FIG. 13 is a diagram showing an example of a graph image. Diagram showing an example of a pixel image Flowchart showing the process of the drawing determination unit Flowchart showing the processing of the image generation unit FIG. 13 shows variations of a graph image in Modification 1. A diagram showing another variation of the graph image in Modification Example 1 FIG. 13 is a diagram showing another variation of a pixel image in Modification 2.

-First embodiment-
A first embodiment of the encoding method will be described below with reference to Fig. 1 to Fig. 13. Fig. 1 is a diagram showing a usage scene of a calculation device 1 according to the present invention. The calculation device 1 generates a graph image 48 and a pixel image 49 using a large number of output values output by a plurality of output devices 9. The graph image 48 and the pixel image 49 are input to a learning model 1000 and used for learning and inference.

The learning model 1000 may employ various known configurations. There are no particular limitations on the type of model employed by the learning model 1000, and it may be any of supervised learning, unsupervised learning, and reinforcement learning. The learning model 1000 may be executed by the computing device 1, or may be executed by a device different from the computing device 1. The learning model 1000 may be executed by different devices in the learning phase and the inference phase, or the computing device 1 that generates the graph image 48 and the pixel image 49 in the learning phase and the inference phase may be physically different devices.

The output device 9 is a sensor or various electronically controlled electronic control devices. The output value is sensing data from a sensor, a calculation result from an electronic control device, a value set from the outside of the electronic control device, etc. Specific examples of output values are the temperature measured by a temperature sensor, the number of proximity events per unit time measured by a proximity sensor, the speed measured by a speed sensor, the illuminance set by a light with an automatic dimming function, the valve opening calculated by a positioner that controls a control valve, etc. Furthermore, the output value may be the set temperature of an air conditioning device set by a user, the speed of a conveyor belt set by a user, etc.

The output relationship between the output device 9 and the learning model 1000 is, for example, as follows. In a first example, the output device 9 is a sensor or electronic control device installed in a room. The calculation device 1 installed in the room generates a graph image 48 and inputs it to the learning model 1000 running on a cloud server. The learning model 1000 infers the state of the room, for example, normal, flooded, fire, etc., based on the graph image 48. In a second example, the output device 9 is an acceleration sensor and gyro built into a smartphone. The calculation device 1 is the same smartphone, and generates a pixel image 49 and inputs it to the learning model 1000 to infer the movement of the person holding the smartphone, for example, running, walking, moving on an escalator, etc.

In the following, generation of the graph image 48 and pixel image 49 by the arithmetic device 1 in the configuration shown in FIG. 1 will be explained in detail. The method of generating graph images 48 and pixel images 49 by arithmetic device 1 is constant regardless of the operation phase of learning model 1000. Although the calculation device 1 will be described below as being able to create both the graph image 48 and the pixel image 49, it is sufficient that the calculation device 1 can create at least one of the graph image 48 and the pixel image 49.

Figure 2 is a functional block diagram of the calculation device 1. The calculation device 1 includes an acquisition unit 11, a preprocessing unit 12, a verification unit 13, an image generation unit 14, and a drawing determination unit 15. However, for ease of explanation, Figure 2 also shows data stored in the storage device 4 included in the calculation device 1. The storage device 4 stores a device list 41, a collected data group 42, a preprocessed data group 43, a verified data group 44, image parameters 45, required specifications 46, a graph image 48, and a pixel image 49.

The acquisition unit 11 acquires output values from multiple output devices 9 and stores them in the storage device 4 as a collected data group 42. Communication between the acquisition unit 11 and each output device 9 may be wired or wireless. The acquisition unit 11 may communicate directly with the output devices 9 or indirectly via a repeater or network. Furthermore, instead of communicating directly with the output devices 9, the acquisition unit 11 may acquire output values offline via a storage medium.

The pre-processing unit 12 processes the collected data group 42 to generate a pre-processed data group 43. The pre-processing unit 12 arranges all data included in the collected data group 42 in chronological order and fills in blanks using an interpolation method described in the device list 41. The pre-processing unit 12 also converts non-numeric data included in the collected data group 42 into numeric values using a predetermined method. For example, the pre-processing unit 12 replaces ON in an ON-OFF value with "1" and OFF with "0".

The verification unit 13 processes the preprocessed data group 43 to generate a verified data group 44 . The verification unit 13 integrates data and supplements missing data so that the sensing data is at predetermined time intervals to generate verified data. In other words, the verification unit 13 generates verified data that is an output value representative of each time period for each output device 9. In this embodiment, an example will be explained in which this time period is 1 minute, but it may be shorter than 1 minute such as 1 second or 0.5 seconds, or longer than 1 minute such as 30 minutes, 1 hour, 1 day, etc. But that's fine.

The image generation unit 14 generates a graph image 48 and a pixel image 49. The drawing judgment unit 15 determines the drawing type and layer number in the device list 41. In other words, the drawing judgment unit 15 determines the drawing type and layer number of each output value and writes them to the device list 41. However, the drawing judgment unit 15 is not an essential component of the calculation device 1, and the calculation device 1 does not need to have the drawing judgment unit 15 if the device list 41 is completed in advance or if the operator rewrites the device list 41.

The device list 41 lists data related to the output values output by the output device 9. The data included in the device list 41 is created in advance, except for the drawing type and layer number. The collected data group 42 is a collection of sensing information acquired by the acquisition unit 11 from the output device 9. This sensing information is time-stamped. The time-stamp may be assigned by each output device 9 or by the acquisition unit 11. Figure 2 merely conceptually illustrates the collected data group 42, and in reality it may be composed of multiple files.

The preprocessed data group 43 is a collection of data generated by the preprocessing unit 12 using the collected data group 42. The verified data group 44 is a collection of data generated by the verification unit 13 using the preprocessed data group 43. The image parameters 45 are a set of parameters created in advance, and are referenced when the image generation unit 14 generates the graph image 48 and the pixel image 49. The requirement specifications 46 are requirements regarding images generated by the arithmetic device 1, and are created in advance. The graph image 48 is an image obtained by removing the axes and scale values from a graph created using the verified data forming the verified data group 44. The pixel image 49 is an image in which each piece of verified data included in the verified data group 44 is expressed as brightness, shade, or color. Specific examples of the graph image 48 and pixel image 49 will be described later.

Figure 3 is a hardware configuration diagram of the calculation device 1. The calculation device 1 includes a CPU 81, which is a central processing unit, a ROM 82, which is a read-only storage device, a RAM 83, which is a readable and writable storage device, an input/output device 84, which is a user interface, a communication device 85, and a storage device 4. The CPU 81 loads a program stored in the ROM 82 into the RAM 83 and executes it to perform the various calculations described above.

The arithmetic unit 1 may be realized by an FPGA (Field Programmable Gate Array), which is a rewritable logic circuit, or an ASIC (Application Specific Integrated Circuit), which is a rewritable logic circuit, instead of the combination of the CPU 81, ROM 82, and RAM 83. good. Furthermore, instead of the combination of the CPU 81, ROM 82, and RAM 83, the arithmetic device 1 may be realized by a combination of different configurations, for example, a combination of the CPU 81, ROM 82, RAM 83, and FPGA.

The input/output device 84 is a keyboard, a mouse, an LCD display, etc. The computing device 1 accepts input from a user using the input/output device 84. The user can use the input/output device 84 to input and modify the device list 41 and image parameters 45. The communication device 85 is a communication module that enables communication with other devices, and may be independent hardware such as a network interface card, or may be realized as a function of a SoC or a CPU. The communication device 85 may realize either wireless communication or wired communication. The storage device 4 is a non-volatile storage device, such as a hard disk drive.

Figure 4 is a diagram showing an example of the device list 41. The device list 41 lists information about all output values contained in the output devices 9. In the example shown in Figure 4, the output type, minimum value, maximum value, storage method, and drawing type are shown for the output values of four output devices 9, D001 to D004. The output type is a distinction between whether the sensing data is a numerical value, i.e., a continuous value, or a discrete value. In the example shown in Figure 4, in the case of a discrete value, the number of values is also listed. In the example shown in Figure 4, a drawing type is specified for each output device 9, but a common drawing type may be specified for all output devices 9.

The minimum and maximum values are written only when the output type is a continuous value, and the minimum and maximum of the continuous value are specified. The interpolation method indicates the interpolation method used by the pre-processing unit 12. The drawing type is information on whether the image is a graph image 48 or a pixel image 49, and on which layer it is to be drawn. Of the information written in the device list 41, only this drawing type is set by the drawing determination unit 15.

Figure 5 is a diagram showing an example of a collected data group 42. In the example shown in Figure 5, output values of four output devices 9, D001 to D004, are shown, and columns with no values are marked with a sharp (#). For ease of explanation, in Figure 5, all data is compiled into one table and sorted by timestamp value. The values in Figures 6 and 7 explained below are based on the values shown in Figure 5.

FIG. 6 is a diagram showing an example of the preprocessed data group 43. In FIG. 6, asterisks are attached to columns that have different values from the collected data group 42 shown in FIG. 5. The changes in FIG. 6 from FIG. 5 are as follows. First, the "ON" value of the second device D002 is replaced with "1", and the "OFF" value is replaced with "0". Second, some values are set by interpolation in the five columns where no values existed. As will be described in detail later, the interpolation process is performed by the preprocessing unit 12 based on the value of the interpolation method listed in the device list 41.

FIG. 7 is a diagram showing an example of the verified data group 44. The difference from FIG. 6 in FIG. 7 is that the data of "10:24:30" is deleted so that a constant time interval is maintained. As a result, representative values of the output of the output device 9 for each one-minute time period are listed in FIG. 7 as verified data. Note that although the figure shows an example in which data is deleted, data may be added at fixed time intervals in the same way. As a method for adding data, a known interpolation technique such as linear interpolation from previous and subsequent data may be used.

Figure 8 is a diagram showing an example of image parameters 45. The image parameters 45 are referred to when the image generating unit 14 generates a graph image 48 and a pixel image 49. The image parameters 45 are broadly divided into two: for graph images and for pixel images. The image parameters 45 for graph images are three: the number of horizontal pixels, the number of vertical pixels, and the time corresponding to the entire width. The "number of horizontal pixels" and "number of vertical pixels" are the horizontal and vertical sizes of the graph image 48 to be generated. The "time corresponding to the entire width" is the length of the period represented by the entire width of the graph image 48, and can also be said to be the length of the period to be contained in one graph image 48 to be generated. The graph image 48 generated based on the example shown in Figure 8 is 1440 pixels wide and 1000 pixels long, and stores data for 24 hours, i.e., one day.

FIG. 9 is a diagram showing an example of the required specifications 46. The required specification 46 is referred to by the drawing determination unit 15. The requirement specifications 46 are requirements regarding images generated by the arithmetic device 1, and include whether or not visibility is required and whether or not conciseness is required. In the example shown in FIG. 9, "YES" indicating that visibility is required for the image generated by the computing device 1 and "YES" indicating that conciseness is required for the image generated by the computing device 1 is written. .

The image parameters 45 for a pixel image are the number of horizontal pixels, the time corresponding to the entire width, the period of one data in one image, and the pixel size of one data point. "Number of horizontal pixels" has the same meaning as for graph images. “The time corresponding to the entire width” is the length of the period expressed by the entire width of the pixel image 49. “Period of one data in one image” is the length of a period contained in one pixel image 49 to be generated. In the example shown in FIG. 8, the "time corresponding to the entire width" is "24 hours", that is, one day, and the "period of one data in one image" is "7 days", so one pixel image 49 has one There are seven columns of sensing data of different types. “Pixel size of one data point” is the number of horizontal and vertical pixels of one sensing data in the pixel image 49.

FIG. 10 is a diagram showing an example of the graph image 48. In the example shown in FIG. 10, the graph image 48 is a graph obtained by plotting one day's data of output values of five output devices 9, for example, the first device D001 to the fifth device D005, as a line graph. However, to be more precise, the graph image 48 is created using verified data that has been processed by the preprocessing unit 12 and the verification unit 13, rather than the values directly output by the output device 9 included in the collected data group 42. The graph image 48 extends in the X direction and the Y direction, which are orthogonal to each other. The graph image 48 shown in FIG. 10 is an image obtained by drawing the verified data constituting the verified data group 44, with time taken in the X direction (horizontal axis) and output value in the Y direction (vertical axis). . Specifically, the left end of the horizontal axis is the value of 0:00, and the right end is the value of 23:59. However, the time axis of the graph image 48 does not have to be one day, but may be one hour, several days, one year, etc. Furthermore, the time at the left end of the horizontal axis is not limited to the above, and may be any time depending on the characteristics and purpose of the data.

In the plot for the first device, the bottom end of the vertical axis is the minimum output value of the first device, and the top end of the vertical axis is the maximum output value of the first device. In the plot for the second device, the bottom end of the vertical axis is the minimum output value of the second device, and the top end of the vertical axis is the maximum output value of the second device. The same is true for the plots for the third to fifth devices. However, the minimum output value of the first device is not the minimum value of the first device used to generate a specific graph image 48, but the minimum output value of the first device throughout the entire period. In this embodiment, the minimum value of the first device listed in the device list 41 is used. Similarly, the maximum value of the first device listed in the device list 41 is used as the maximum value of the first device. The same is true for the minimum and maximum values of the second to fifth devices. In FIG. 10, the type of line is changed for each device for convenience of drawing, but the difference between the devices may be expressed by using different colors.

FIG. 11 is a diagram showing an example of the pixel image 49. The pixel image 49 shown in FIG. 11 shows output values for one week of a total of ten output devices 9, the first device D001 to the tenth device D010. The pixel image 49 shown in FIG. 11 has 1440 pixels horizontally and 70 pixels vertically. One pixel constituting this pixel image 49 represents any output value every minute. Since there are 1440 minutes in a day, one horizontal line indicates the output value of the output device 9 for one day. Since FIG. 11 shows output values for one week, that is, seven days, for each output device 9, seven pixels are used in the vertical direction for each output device 9.

The enlarged view of the pixel image 49 shown at the bottom of FIG. 11 shows a range of 5 pixels horizontally and 15 pixels vertically. In FIG. 11, the output value is represented by the type of hatching for each pixel. The first column at the left end represents the output value from 0:00 to 0:01, and the second column from the left end represents the output value from 0:01 to 0:2. The top row shows the output value of the first device on the first day, and the second row shows the output value of the first device on the second day. That is, the plots are plotted in the horizontal direction at time intervals of 1 minute, and in the vertical direction at time intervals of 1 day. In FIG. 11, plot values are expressed by hatching types for convenience of drawing, but plot values may also be expressed by colors.

Figure 12 is a flowchart showing the processing of the drawing judgment unit 15. The drawing judgment unit 15 performs the processing described below when the drawing type and layer number columns in the device list 41 are blank. The processing of the drawing judgment unit 15 is performed prior to the generation of the graph image 48 and pixel image 49 by the image generation unit 14. The drawing judgment unit 15 first refers to the requirement specifications 46 and determines whether or not visibility is required. If the drawing judgment unit 15 determines that visibility is required, in other words that the visibility value in the requirement specifications 46 is "YES", it makes a positive judgment at this step and proceeds to step S304, and if it determines that visibility is not required, it proceeds to step S302.

In step S302, the drawing judgment unit 15 judges the relationship between the total number of output values listed in the device list 41 and the threshold values of "20" and "50". If the drawing judgment unit 15 judges that the total number of output values is less than 20, the process proceeds to step S304. If the drawing judgment unit 15 judges that the total number of output values is 20 or more but less than 50, the process proceeds to step S303. If the drawing judgment unit 15 judges that the total number of output values is 50 or more, the process proceeds to step S305. In this embodiment, the drawing judgment unit 15 makes the above judgment using the device numbers "20" and "50" as threshold values, but the threshold values are not necessarily limited to these values, and any value may be set. In step S303, the drawing judgment unit 15 refers to the required specifications 46 and judges whether or not conciseness is necessary. If the drawing judgment unit 15 judges that conciseness is necessary, in other words, that the conciseness value in the required specifications 46 is "YES", the drawing judgment unit 15 judges this step as positive and proceeds to step S305. If the drawing judgment unit 15 judges that conciseness is unnecessary, the process proceeds to step S304.

In step S304, the drawing determination unit 15 determines the graph image 48 as the image to be generated, and proceeds to step S311. In step S305, the drawing determination unit 15 determines the pixel image 49 as the image to be generated, and proceeds to step S311. In step S311, the drawing determination unit 15 uses the huge amount of output values obtained so far to calculate a correlation coefficient that indicates the correlation between output values. As this correlation coefficient, Pearson's product moment correlation coefficient or Spearman's rank correlation coefficient can be used, and other correlation coefficients may also be used.

In the following step S312, the drawing determination unit 15 groups output values that are strongly related to each other based on the value of the correlation coefficient calculated in step S311. However, there is a limit to the number of output values that can be included in one group, and an example of the upper limit is about 5 to 10 for the graph image 48 and about 20 for the pixel image 49. Then, the drawing determination unit 15 assigns a serial number starting from "1" to each group. Note that there is no priority in group numbers, and this group number is used as a layer number.

In the following step S313, the drawing judgment unit 15 records the image type and group number determined in step S304 or step S305 in the device list 41, and ends the process shown in FIG. 12. The drawing judgment unit 15 writes the image type in the drawing type column of the device list 41, and writes the group number in the layer number column. Note that hereinafter, steps S301 to S305 in FIG. 12 are also referred to as drawing decision steps, and steps S311 to S312 are also referred to as drawing judgment steps.

Figure 13 is a flowchart showing the processing of the image generation unit 14. First, in step S321, the image generation unit 14 selects an undrawn layer as the layer to be processed. An undrawn layer is a layer that has not yet been selected as the layer to be processed among all the combinations of drawing types and layer numbers listed in the device list 41. For example, in the example of the device list 41 shown in Figure 4, there are three layers: layer number 1 of graph image 48, layer number 2 of graph image 48, and layer number 1 of pixel image 49. When step S321 is executed for the first time, the image generation unit 14 selects one of these three as the layer to be processed and proceeds to step S322.

In the following step S322, the image generation unit 14 reads all verified data belonging to the processing target layer from the verified data group 44, and proceeds to step S323. In the following step S323, the image generation unit 14 determines the type of the layer to be processed, that is, the graph image 48 or the pixel image 49. If the image generation unit 14 determines that the layer to be processed is the layer of the graph image 48, the process proceeds to step S324, and if it determines that the layer to be processed is the layer of the pixel image 49, the process proceeds to step S326.

In step S324, the image generation unit 14 reads the minimum and maximum values for each data read in step S322 from the device list 41. These minimum and maximum values correspond to the bottom and top of the graph drawing area as shown in FIG. 10. In the following step S325, the image generation unit 14 draws a graph of the data read in step S322 according to the image parameters 45 and the minimum and maximum values read in step S324. This graph becomes an image of the specified layer of the graph image 48. Next, the image generation unit 14 proceeds to step S328.

In step S326, the image generation unit 14 reads the minimum value and maximum value for each data read in step S322 from the device list 41, and determines the correspondence between the output value and color. For example, if the minimum value of a certain output value is "0" and the maximum value is "100", the image generation unit 14 sets the weight of the output value "1" to the value obtained by dividing "0xffffff" by "100", The top two digits of the hexadecimal notation are each the R, G, and B values. In the following step S327, the image generation unit 14 draws on the specified layer of the pixel image 49 based on the value read in step S322, the correspondence between the value and color determined in step S326, and the value of the image parameter 45, and performs the step Proceed to S328.

In step S328, which is executed when the processing of step S325 or step S327 is completed, the image generation unit 14 determines whether or not drawing has been completed for all combinations of drawing types and layer numbers listed in the device list 41. If the image generation unit 14 determines that all drawing has been completed, it ends the processing shown in FIG. 13, and if it determines that there are any layers that have not been drawn, it returns to step S321.

According to the first embodiment described above, the following effects can be obtained.
(1) The arithmetic device 1, which is a computer, encodes the outputs of the plurality of output devices 9 using the following encoding method, that is, converts them into images. The encoding method executed by the arithmetic device 1 includes an acquisition step executed by the acquisition unit 11 to acquire the collected data group 42 that is the output of the plurality of output devices 9, and an acquisition step executed by the verification unit 13 to acquire the collected data group 42. A verification step of generating a verified data group 44 that is verified data for each time period and each output device 9 based on the verified data group 44 performed by the image generation unit 14; The process includes an image generation step of creating an encoded image, that is, a graph image 48 and a pixel image 49, by encoding the verified data of two or more output devices 9 included in the process. Therefore, it is possible to generate an image in which correlations between multiple signals can be learned and inferred.

(2) The graph image 48 is a two-dimensional image that extends in the X direction and the Y direction. In the image generation step, verified data in different time zones are arranged at different positions in the X direction, and verified data with different values are arranged in different positions in the Y direction. Therefore, the correlation between multiple signals can be expressed as a two-dimensional graph and used for learning machine learning models and inference using machine learning models.

(3) In the image generation step, the verified data originating from the same output device 9 is further connected with a line in chronological order and expressed as a line graph.

(4) In the image generation step, the pixel image 49 is divided into a plurality of regions, and for each region, the verified data corresponding to the output of the output device to be processed, which is one output device 9, is assigned as brightness, shade, or color. encode. Therefore, the correlation between a plurality of signals can be expressed as a pixel image 49.

(5) The pixel image 49 is a two-dimensional image that spreads in a second direction perpendicular to the X and Y directions. The image generation unit 14 targets the output device to be processed for each region, and arranges verified data corresponding to time periods of one minute in the X direction and verified data corresponding to time periods of one day in the Y direction. Therefore, the calculation device 1 can generate a pixel image 49 in which the output of each output device 9 is arranged for each rectangular region.

(6) The drawing determination unit 15 executes a drawing determination step (S311 to S312 in FIG. 12) of determining a combination of output devices 9 to draw on the same layer based on the correlation of the outputs of the plurality of output devices 9. Therefore, the outputs of the output devices 9 that have a strong correlation can be arranged in the same layer.

(7) The rendering determination unit 15 executes a rendering determination step (S301 to S305 in FIG. 12) for determining the type of encoded image based on the required specifications 46. In the image generation step executed by the image generation unit 14, an encoded image of the type described in the drawing type column of the device list 41, that is, a graph image 48 or a pixel image 49, is generated.

(Modification 1)
The operator may register different forms of the graph image 48, for example, "graph 2" or "graph 3" as the drawing type in the device list 41. In this case, the image generation unit 14 generates a graph image 48 in a format different from that in the first embodiment.

Figure 14 shows a second graph image 48-2, which is another variation of graph image 48. The example shown in Figure 14 is a different type of graph from the example shown in Figure 10. That is, Figure 14 is a scatter plot in which only markers are plotted, and no lines exist connecting the markers. In Figure 14, the type of plot is changed for each output device 9 for convenience of drawing, but the differences between output devices 9 may be expressed by using different colors.

Figure 15 shows third graph image 48-3, which is yet another variation of graph image 48. However, for comparison, Figure 15 also shows graph image 48, which plots the same data. Also, to clarify the differences in expression, Figure 15 shows only the output of one output device 9. In third graph image 48-3, specific values contained in the output values are not plotted, but are substituted with other immediately preceding values. Third graph image 48-3 is effective for output values that change from one value to another value, via zero, such as currents used in motor control.

(Modification 2)
The operator may register different forms of the pixel image 49, such as "pixel 2" and "pixel 3", as the drawing type in the device list 41. In this case, the image generation unit 14 generates a pixel image 49 in a format different from that in the first embodiment.

FIG. 16 is a diagram showing another variation of the pixel image 49. The upper row of FIG. 16 shows the pixel image 49 in the first embodiment, the middle row of FIG. 16 shows the second pixel image 49-2 of the second format, and the lower row of FIG. The third pixel image 49-3 is shown in the form of . The pixel image 49 is divided into regions only in the vertical direction in the figure, and only the output values of the same output device 9 are arranged in the horizontal direction in the figure.

In contrast, the second pixel image 49-2 is divided into two not only in the vertical direction but also in the horizontal direction. The third pixel image 49-3 is divided into three not only in the vertical direction but also in the horizontal direction. Furthermore, although the divided areas are shown here as rectangles, they may be divided into areas of shapes other than rectangles. In this way, the pixel image 49 can be divided into any shape, as long as the shape remains the same throughout the learning phase and inference phase of the learning model 1000.

(Modification 3)
In the embodiment described above, it was assumed that the graph image 48 and the pixel image 49 had a plurality of layers. However, the graph image 48 and the pixel image 49 may be image files that do not have a layer structure, such as JPG or BMP. In this case, the graph image 48 and pixel image 49 become multiple files.

In each of the embodiments and modifications described above, the configuration of the functional blocks is merely an example. Several functional configurations shown as separate functional blocks may be integrated, or a configuration shown in one functional block diagram may be divided into two or more functions. Further, a configuration may be adopted in which some of the functions of each functional block are provided in other functional blocks.

In the above-described embodiments and variations, the program is stored in ROM 82, but the program may be stored in storage device 4. Furthermore, the arithmetic device 1 may be provided with an input/output interface (not shown), and the program may be read from another device when necessary via the input/output interface and a medium available to the arithmetic device 1. Here, the medium refers to, for example, a storage medium that is detachable from the input/output interface, or a communication medium, i.e., a network such as a wired, wireless, or optical network, or a carrier wave or digital signal that propagates through the network. Furthermore, some or all of the functions realized by the program may be realized by a hardware circuit or FPGA.

Each of the embodiments and modifications described above may be combined. Although various embodiments and modifications have been described above, the present invention is not limited to these. Other embodiments considered within the technical spirit of the present invention are also included within the scope of the present invention.

1: Arithmetic device 9: Output device 11: Acquisition unit 12: Preprocessing unit 13: Verification unit 14: Image generation unit 15: Drawing judgment unit 41: Device list 42: Collection data group 43: Preprocessed data group 44: Verification Completed data group 45: Image parameters 46: Required specifications 48: Graph image 49: Pixel image

Claims (7)

An encoding method executed by a computer for encoding outputs of a plurality of output devices, the method comprising:
an acquisition step of acquiring a collection data group that is the output of the plurality of output devices;
a verification step of generating a verified data group that is verified data for each time period and each output device based on the collected data group;
An encoding method comprising: creating an encoded image by encoding the verified data of two or more output devices included in the plurality of output devices, based on the verified data group. 2. The encoding method according to claim 1,
the encoded image is a two-dimensional image having an extent in a first direction and a second direction perpendicular to the first direction,
The encoding method, in the image generating step, the verified data having different time periods are arranged at different positions in the first direction, and the verified data having different values are arranged at different positions in the second direction. The encoding method according to claim 2,
In the image generation step, further,
An encoding method in which the verified data originating from the same output device are connected in chronological order by a line. 2. The encoding method according to claim 1,
In the image generating step,
An encoding method comprising: dividing the encoded image into a plurality of regions; and encoding, for each region, the verified data corresponding to the output of a processing target output device, which is one of the output devices, as brightness, shading, or color. 5. The encoding method according to claim 4,
the encoded image is a two-dimensional image having an extent in a first direction and a second direction perpendicular to the first direction,
In the image generating step,
An encoding method, for each area, targeting the output device to be processed, arranging the verified data corresponding to the output at a first time interval in the first direction, and arranging the verified data corresponding to the output at a second time interval, which is a time interval longer than the first time interval, in the second direction. The encoding method according to claim 1,
the encoded image includes multiple layers,
The encoding method further includes a rendering determination step of determining a combination of the output devices to be rendered on the same layer based on a correlation between the outputs of the plurality of output devices. The encoding method according to claim 1,
further comprising a drawing determination step of determining the type of the encoded image based on required specifications,
In the image generation step, the encoding method generates the encoded image of the type determined in the drawing determination step.

Images