JP2023090555A - Waveform measuring device and method - Google Patents

Abstract

To facilitate shortening of the whole processing time for AI analysis using waveform data while acquiring data indicating a feature quantity.SOLUTION: A waveform measuring device 11 comprises a control section 110 that sets at least one channel of a plurality of channels to a trigger channel, sequentially compares each sample value included in digital signal data written into an area corresponding to the trigger channel with a threshold in a storage, reads at least one sample value included in the digital signal data written during a certain period including a detection time point of a trigger point in the area corresponding to each channel of the plurality of channels in the storage each time one sample value is detected as the trigger point, performs processing for extracting a feature quantity of the read sample value, and outputs data indicating the extracted feature quantity.SELECTED DRAWING: Figure 5

Description

The present disclosure relates to waveform measuring instruments and methods.

Non-Patent Literature 1 describes a technique for streamlining the setting of thresholds used for trend monitoring to detect equipment failures in advance.

Masahiko Sato, "Application of Machine Learning Technology to Trend Monitoring by Sushi Sensor," Yokogawa Technical Report, Vol.63, No.1 (2020), pp.23-26

Trend monitoring uses a waveform measuring instrument with a trigger function. A trigger function is a function of detecting a characteristic signal of digital signal data as a trigger for starting predetermined processing to be executed on the digital signal data. Detecting a characteristic signal of digital signal data is also called trigger detection. Measurement using a waveform measuring instrument with a trigger function is also called trigger measurement. Trigger measurement is effective for monitoring the behavior of a processing machine during processing in a machining production line. For example, when measuring signals from various sensors installed in a processing machine in order to observe the behavior of the processing machine, if a characteristic signal during processing is used as a trigger, the behavior of the processing machine can be monitored each time processing is performed. can be loaded into the waveform measuring instrument. By using the trigger function, it is possible to acquire data based on the same point in a series of operations to be monitored for trends. A waveform measuring instrument having a trigger function transmits digital signal data captured in response to trigger detection to an external device such as a local PC or a server each time digital signal data is captured. "PC" is an abbreviation for personal computer. On the external device side, AI analysis is performed using the received digital signal data, for example, for the purpose of failure prediction of the processing machine. "AI" is an abbreviation for artificial intelligence. Here, the digital signal data transmitted from the waveform measuring instrument is time sampling data. In order to improve the sampling accuracy, it is necessary to shorten the sampling period, which increases the amount of data. Therefore, if unprocessed digital signal data captured by the waveform measuring instrument, that is, "raw data" is directly transferred to the external device, the amount of data communication becomes enormous. Further, on the external device side, if the received raw data is analyzed as it is, the analysis time will be long due to the large number of sample values. Therefore, it is conceivable that a feature amount of sample values of a specific part of raw data is extracted and AI analysis is performed on the feature amount. A “feature amount” is a representative value that characterizes sample values included in raw data, and is, for example, an average value of sample values of a specific portion of raw data. In this case, if the external device extracts the feature amount from the raw data, the data processing load and processing time increase. As a result, the overall processing time for AI analysis using waveform data is lengthened.

An object of the present disclosure is to facilitate shortening the overall processing time for AI analysis using waveform data while acquiring data indicating feature amounts.

A waveform measuring instrument according to some embodiments converts analog signal waveforms input from a plurality of channels into digital signal data, and performs data acquisition to write the digital signal data of each channel to a corresponding area in a storage. wherein at least one of a plurality of channels is set as a trigger channel, and each sample value included in digital signal data written to an area corresponding to the trigger channel in storage, a threshold value, and are sequentially compared, and each time one sample value is detected as a trigger point, in the area corresponding to each of the plurality of channels in the storage, digital data written during a certain period including the detection point of the trigger point The control unit reads out at least one sample value included in the signal data, executes processing for extracting the feature amount of the read sample value, and outputs data indicating the extracted feature amount. In such a waveform measuring instrument, each time a trigger point is detected, processing for extracting a feature amount of at least one sample value included in digital signal data written during a certain period is executed, and the execution result is It is output as data indicating the feature amount. In other words, the waveform measuring instrument can perform the process of acquiring the data indicating the feature amount from a specific part of the raw data. By transmitting data indicating the acquired feature amount from the waveform measuring device to the terminal device, the terminal device can immediately use the data for machine learning used for AI analysis. Therefore, the amount of data communication between the waveform measuring instrument and the external device and the data processing load on the external device side are reduced compared to the case where the raw data is directly transmitted to the external device. Therefore, it becomes easy to shorten the overall processing time for AI analysis using waveform data while acquiring data indicating a feature amount.

In the waveform measuring instrument according to one embodiment, the control unit executes data acquisition and feature amount extraction processing in parallel. For example, when importing data from a press machine in order to monitor the trend of the behavior of press working by the press machine, if the data import and the process of extracting the feature amount are not executed in parallel, the feature amount is Data acquisition will not be performed while the process of extracting is being performed. In this case, it is not realistic to give priority to data acquisition, provide time for data acquisition during press working, and suspend press working until data acquisition is completed. Therefore, the data acquisition is interrupted rather than the press working. As a result, data acquisition becomes discontinuous, and data loss may occur. On the other hand, according to such an embodiment, since data can be loaded even during feature quantity extraction, dead time between data loadings does not occur, and continuity of data loading is ensured.

In the waveform measuring instrument according to one embodiment, the control unit performs a process of extracting a feature amount of sample values read from regions corresponding to two or more channels among a plurality of channels. A process of converting one or more feature amounts into one feature amount is further executed, and data representing the one feature amount is included in output data. According to such an embodiment, it is possible to acquire data indicating another feature amount by combining the execution results of the process of extracting the feature amount of a plurality of channels.

A method according to some embodiments converts waveforms of analog signals input from a plurality of channels into digital signal data, and writes the digital signal data of each channel to a corresponding area in storage during data acquisition. setting at least one of the plurality of channels as a trigger channel; sequentially comparing each sample value included in the digital signal data written to the area corresponding to the trigger channel in the storage with the threshold; every time one sample value is detected as a trigger point, digital signal data written during a certain period including the detection point of the trigger point is included in the area corresponding to each of the plurality of channels in the storage. reading out at least one sample value to be read, executing a process of extracting a feature amount of the read sample value, and outputting data indicating the extracted feature amount. According to this method, each time a trigger point is detected, a process of extracting a feature amount of at least one sample value included in digital signal data written during a certain period is executed, and the execution result is a feature. It is output as data indicating the amount. In other words, it is possible to obtain data indicating a feature amount from a specific portion of raw data and extract data for machine learning used in AI analysis.

The method according to one embodiment further includes performing data acquisition and feature extraction in parallel. For example, when importing data from a press machine in order to monitor the trend of the behavior of press working by the press machine, if the data import and the process of extracting the feature amount are not executed in parallel, the feature amount is Data acquisition will not be performed while the process of extracting is being performed. In this case, it is not realistic to give priority to data acquisition, provide time for data acquisition during press working, and suspend press working until data acquisition is completed. Therefore, the data acquisition is interrupted rather than the press working. As a result, data acquisition becomes discontinuous, and data loss may occur. On the other hand, according to such an embodiment, since data can be loaded even during feature quantity extraction, dead time between data loadings does not occur, and continuity of data loading is ensured.

In the method according to one embodiment, two or more feature values obtained by performing a process of extracting feature values of sample values read from regions corresponding to two or more channels out of a plurality of channels are extracted. It further includes performing a process of converting into one feature amount, and including data representing the one feature amount in the data to be output. According to such an embodiment, it is possible to acquire data indicating another feature amount by combining the execution results of the process of extracting the feature amount of a plurality of channels.

According to the present disclosure, it becomes easier to shorten the overall processing time for AI analysis using waveform data while acquiring data indicating feature amounts.

FIG. 10 is a diagram showing a data flow of a system according to a comparative example; FIG. 4 is a block diagram showing the detailed configuration of a waveform measuring instrument according to a comparative example; FIG. 10 is a diagram showing data writing and data reading in a ring buffer when a trigger is detected; It is a figure which shows the data flow of the system which concerns on this embodiment. 1 is a block diagram showing the configuration of a system according to this embodiment; FIG. 3 is a block diagram showing the detailed configuration of the waveform measuring instrument according to this embodiment; FIG. It is a flow chart which shows operation of the system concerning this embodiment. FIG. 4 is a diagram showing an example of a waveform generated by plotting sample values obtained in triggered measurements; FIG. 10 is a diagram showing areas corresponding to each channel when data is captured by four channels; FIG. 4 is a diagram showing a specific example of data reading and feature amount extraction; It is a figure which shows data acquisition, data read-out, and extraction of a feature-value in a time series. It is a figure which shows data acquisition, data read-out, and extraction of a feature-value based on a modified example.

First, a comparative example will be explained.

FIG. 1 shows the configuration of a system 20 according to a comparative example. The system 20 includes a waveform measuring instrument 21 having a trigger function, and a local PC 22 connected to the waveform measuring instrument 21 via a network such as the Internet.

The waveform measuring device 21 converts the waveform of the analog signal into digital signal data and transmits the digital signal data to the local PC 22 . The local PC 22 accumulates digital signal data and acquires data for machine learning from the accumulated data. The local PC 22 generates a model used for AI analysis for trend monitoring from the acquired data for machine learning.

The configuration of the waveform measuring device 21 will be described with reference to FIG.

In the waveform measuring instrument 21, the observed analog signal of the analog input channel CH1 is normalized by the input amplifier 201 and input to the A/D converter 202, as shown in FIG. "A/D" is an abbreviation for analog to digital. An analog signal of analog input channel CH2 observed simultaneously with analog input channel CH1 is normalized by input amplifier 204 and input to A/D converter 205 . The timing for fetching data is controlled by a sample timing generation circuit 211 . A/D converters 202 and 205 perform A/D conversion according to the timing signal generated by sample timing generation circuit 211 . The converted data are sent to the memory controller 212 through the interface circuits 203 and 206 and stored in the waveform memory 221 sequentially. The data stored in the waveform memory 221 is converted into waveform display data by the display waveform generation circuit 231 and displayed on the waveform display 232 as a waveform. The data stored in the waveform memory 221 is transmitted to the local PC 22 via the network when a certain amount of data is accumulated, is read by the local PC 22 as appropriate, is converted into waveform or numerical data, and is displayed on the local PC 22. , is observed.

FIG. 3 shows data writing and data reading in the waveform memory 221 when performing trigger detection while sequentially storing converted data in the waveform memory 221 . In the example shown in FIG. 3, the waveform memory 221 is controlled by the memory controller 212 as a ring buffer. Data fetching in the ring buffer starts from address A of the waveform memory 221, and when data fetching reaches address B, it returns to address A and data is overwritten. Trigger detection circuit 213 performs trigger detection. Specifically, the trigger detection circuit 213 determines whether or not the sample value of the digital signal data of the corresponding channel satisfies a preset trigger condition. The trigger detection circuit 213 detects a trigger point according to the determination result, and notifies the memory controller 212 of the detected trigger point. The memory controller 212 stores the address at which the sample value was written when the trigger detection circuit 213 notified the trigger point as the trigger point address nT. After the trigger point is detected, the memory controller 212 fetches a preset number of sample values, that is, when the data writing of the final address nE is completed, the data fetching is completed. As the head address for data reading, an address n0 that is a preset number of times backward from the trigger point address nT is set. The display waveform generation circuit 231 reads out data from the waveform memory 221 in the order of n0→nT→B→A→nE, creates a display waveform, and causes the waveform display 232 to display it.

In such a system 20, the waveform measuring instrument 21 converts the waveform data captured in response to trigger detection into digital signal data, and transmits the digital signal data to the local PC 22 each time data is captured. As described above, if the raw data is passed from the waveform measuring instrument 21 to the local PC 22 as it is, the amount of data communication becomes enormous. Further, if the local PC 22 side extracts the feature amount from the raw data, the data processing load and processing time become excessive. As a result, the overall processing time for AI analysis using waveform data is lengthened.

Therefore, it is necessary to make it easier to shorten the overall processing time for AI analysis using waveform data while acquiring data indicating feature amounts.

Several embodiments of the present disclosure are described below with reference to the drawings. In each figure, the same reference numerals are given to the same or corresponding parts. In the description of each embodiment, the description of the same or corresponding parts will be omitted or simplified as appropriate.

A first embodiment, which is one embodiment of the present disclosure, will be described.

The configuration of the system 10 according to this embodiment will be described with reference to FIG.

The system 10 includes a waveform measuring instrument 11 and a terminal device 12, as shown in FIG.

The waveform measuring device 11 is a device that converts the waveform of an input analog signal into digital signal data. The waveform measuring instrument 11 includes an input section 100, a control section 110, a storage section 120, a display section 130, and a communication section 140, as shown in FIG. The waveform measuring instrument 11 can communicate with the terminal device 12 via a network 15 such as the Internet.

The terminal device 12 is a desktop computer, a laptop computer, or the like installed on-site to display waveforms. The terminal device 12 includes an input unit 150, a control unit 160, a storage unit 170, a display unit 180, and a communication unit 190, as shown in FIG.

A part or all of the functions of the waveform measuring instrument 11 are implemented by a programmable circuit or a dedicated circuit included in the control section 110 . That is, part or all of the functions of the waveform measuring instrument 11 are realized by hardware.

When a part of the functions of the waveform measuring instrument 11 is realized by hardware, the rest of the functions of the waveform measuring instrument 11 are realized by executing the program according to the present embodiment with a processor included in the control unit 110. may be implemented. That is, the functions of the rest of the waveform measuring instrument 11 may be realized by software. In this case, the program is a program for causing the computer to execute the processing of steps included in the operation of the waveform measuring instrument 11, thereby causing the computer to implement the functions corresponding to the processing of the steps. That is, the program is a program for causing the computer to function as the waveform measuring instrument 11 .

The program can be recorded in a computer-readable recording medium. A computer-readable recording medium is, for example, a magnetic recording device, an optical disk, a magneto-optical recording medium, or a semiconductor memory. Program distribution is performed, for example, by selling, assigning, or lending a portable recording medium such as a DVD or CD-ROM on which the program is recorded. "DVD" is an abbreviation for digital versatile disc. "CD-ROM" is an abbreviation for compact disc read only memory. The program may be distributed by storing the program in the storage of the server and transferring the program from the server to another computer via the network 15 . A program may be provided as a program product.

The computer, for example, once stores in memory a program recorded on a portable recording medium or a program transferred from a server. Then, the computer reads the program stored in the memory with the processor, and executes processing according to the read program with the processor. The computer may read the program directly from the portable recording medium and execute processing according to the program. The computer may execute processing according to the received program every time the program is transferred from the server to the computer. The processing may be executed by a so-called ASP type service that realizes the function only by executing the execution instruction and obtaining the result without transferring the program from the server to the computer. The program includes information that is used for processing by a computer and that conforms to the program. For example, data that is not a direct instruction to a computer but that has the property of prescribing the processing of the computer corresponds to "things equivalent to a program."

An outline of the present embodiment will be described with reference to FIGS. 4 and 5. FIG.

The waveform measuring instrument 11 is a waveform measuring instrument that converts analog signal waveforms input from a plurality of channels into digital signal data and writes the digital signal data of each channel into the corresponding area in the storage. . Waveform measuring instrument 11 sets at least one of the plurality of channels as a trigger channel. A “trigger channel” is a channel for detecting a sample value that serves as a trigger for starting execution of processing for extracting a feature amount. The waveform measuring instrument 11 sequentially compares each sample value included in the digital signal data written in the area corresponding to the trigger channel in the storage with the threshold, and detects one sample value as the trigger point TP. A “trigger point TP” is a sample value that serves as a trigger for starting the execution of the processing for extracting the feature amount. Each time the waveform measuring instrument 11 detects the trigger point TP, the waveform measuring instrument 11 detects the digital signal data written during a certain period including the trigger point detection time in the area corresponding to each of the plurality of channels in the storage. At least one sample value is read out and a process of extracting the feature amount of the read sample value is executed. The waveform measuring device 11 outputs data indicating the extracted feature amount. In this embodiment, storage is managed as a ring buffer. Data fetching is executed by overwriting data in a ring form in the section from the start address to the end address of the area corresponding to the trigger channel in the storage until the trigger point TP is detected. Therefore, when the number of sample values included in the digital signal data is L, L is not a fixed value but an indefinite value until the trigger point TP is detected. For example, depending on the time until the trigger point TP is detected, the ring-shaped data overwriting process in the section may be repeated several hundred times. That is, the number L of sample values varies depending on the measurement conditions at each time. On the other hand, the number of addresses of the area where sample values are written is a fixed value. This fixed value is greater than the number of sample values for which the feature extraction process is performed, ie, the number of sample values to be observed for trigger measurement. Trigger detection begins simultaneously with data acquisition and is performed in real time for all sample values contained in the acquired digital signal data. The waveform measuring instrument 11 outputs data indicating the extracted feature quantity to the terminal device 12 .

According to the present embodiment, as shown in FIG. 4, a process of acquiring data indicating a feature amount from a specific portion of raw data and extracting data for machine learning used in AI analysis is performed by a waveform measuring device. 11 can be executed. That is, the waveform measuring instrument 11 side can execute processing for extracting the feature amount, and can transfer the data indicating the acquired feature amount to the terminal device 12 . On the terminal device 12 side, all that is required is to accumulate data indicating the transferred feature amounts and generate a model used for AI analysis for trend monitoring from the accumulated data. Therefore, the amount of data communication between the waveform measuring instrument 11 and the terminal device 12 and the data processing load on the terminal device 12 side are reduced compared to the case where the raw data is transmitted to the terminal device 12 as it is. Therefore, it becomes easy to shorten the overall processing time for AI analysis using waveform data while acquiring data indicating a feature amount.

Next, the configuration of the system 10 according to the first embodiment will be described in detail with reference to FIGS. 5 and 6. FIG.

The system 10 includes the waveform measuring instrument 11 and the terminal device 12 as described above.

Waveform measuring instrument 11 includes input unit 100, control unit 110, storage unit 120, display unit 130, and communication unit 140, as described above.

The input unit 100 is an interface that takes in data of each channel to the waveform measuring instrument 11 . 6, the input unit 100 includes input amplifiers 101-1, 101-2, . . . , A/D converters 102-1, 102-2, . , 103-2, .

As shown in FIG. 6, the observed analog signal of channel CH1 is normalized by input amplifier 101-1 and input to A/D converter 102-1. Here, normalization means adjusting the amplitude of the analog input signal so that it falls within an appropriate range for the input specifications of the A/D converter. Analog signals of channels CH2, CH3, and CH4 observed simultaneously are normalized by input amplifiers 101-2, . . . and input to A/D converters 102-2, . As shown in FIG. 6, when there are a plurality of channels, input amplifiers, A/D converters, and interface circuits are provided for the number of channels.

Control unit 110 includes at least one processor, at least one programmable circuit, at least one dedicated circuit, or any combination thereof. The processor is a general-purpose processor such as a CPU or GPU (graphics processing unit), or a dedicated processor specialized for specific processing. The programmable circuit is, for example, an FPGA (field-programmable gate array). The dedicated circuit is, for example, an ASIC (application specific integrated circuit). The control unit 110 executes processing related to the operation of the waveform measuring instrument 11 while controlling each part of the waveform measuring instrument 11 . As shown in FIG. 6, the control unit 110 includes a sample timing generation circuit 111, a memory controller 112, a trigger detection circuit 113, a memory controller 114, an arithmetic unit 115, a data number counting circuit 116, and a processing indicator. 117. A sample timing generation circuit 111 controls the timing of fetching data. A/D conversion is performed in the A/D converters 102 and 105 of the input section 100 according to the timing signal generated by the sample timing generation circuit 111 . The memory controller 112 receives data converted by the A/D converters 102 and 105 through the interface circuits 103 and 106 of the input section 100 . The trigger detection circuit 113 executes trigger detection, which will be described later. The memory controller 114 receives from the memory controller 112 a notification indicating the completion of data acquisition after trigger detection, and sequentially reads data from corresponding addresses in the waveform memory 121 of the storage unit 120, which will be described later. The data number counting circuit 116 counts the number of sample values sequentially read out from the waveform memory 121 with a counter. Sample values sequentially read out from the waveform memory 121 are input to the calculator 115 . The arithmetic unit 115 executes processing for extracting feature amounts from the input data. The processing indicator 117 determines the type of processing to be executed by the arithmetic unit 115 on the sample values input to the arithmetic unit 115 according to the number of sample values indicated by the count value of the counter of the data number counting circuit 116. It also controls the start and stop of processing.

Storage unit 120 includes at least one semiconductor memory, at least one magnetic memory, at least one optical memory, or any combination thereof. A semiconductor memory is, for example, a RAM (random access memory). RAM is, for example, SRAM (static random access memory) or DRAM (dynamic random access memory). The storage unit 120 may be an electrically erasable programmable read only memory (EEPROM). The storage unit 120 includes a waveform memory 121 and a computation result memory 122, as shown in FIG.

In the waveform memory 121, the memory controller 112 of the control unit 110 writes the digital signal data acquired in each sampling period according to the timing signal generated by the sample timing generation circuit 111. FIG. Waveform memory 121 is controlled by memory controller 112 as a ring buffer as described above with reference to FIG.

The calculation result memory 122 stores the execution result of the processing executed by the calculator 115 of the control unit 110 .

The display unit 130 includes a display waveform generation circuit 131 and a waveform display 132, as shown in FIG.

The display waveform generation circuit 131 is a general-purpose processor such as GPU, a dedicated processor specialized for waveform display processing, a programmable circuit such as FPGA, or a dedicated circuit such as ASIC. The display waveform generation circuit 131 converts the digital signal data stored in the waveform memory 121 into waveform display data upon command from the calculator 115 as required. The display waveform generation circuit 131 displays the waveform display data as a waveform on the waveform display 132 as required.

Waveform display 132 includes at least one output interface. The output interface is, for example, an LCD (liquid crystal display) or an organic EL (electro luminescent) display. A waveform display 132 outputs data obtained by the operation of the waveform measuring device 11 . The waveform display 132 may be provided integrally with a touch panel capable of accepting input from the user. The waveform display 132 displays waveform display data under the control of the display waveform generation circuit 131 .

Communication unit 140 includes a communication circuit. A communication circuit is one or more communication ICs. "IC" is an abbreviation for integrated circuit. The communication IC is, for example, a LAN communication IC. "LAN" is an abbreviation for local area network. The communication unit 140 may be a PCI network adapter, a USB network adapter, or an Ethernet (registered trademark) adapter. "PCI" is an abbreviation for Peripheral Component Interconnect. "USB" is an abbreviation for Universal Serial Bus. The communication unit 140 receives information used for the operation of the waveform measuring instrument 11 via the network 15 and transmits information obtained by the operation of the waveform measuring instrument 11 via the network 15 .

The terminal device 12 includes the input unit 150, the control unit 160, the storage unit 170, the display unit 180, and the communication unit 190, as described above.

The input unit 150 includes at least one input interface. Input interfaces are, for example, physical keys, capacitive keys, pointing devices, touch screens, or microphones. The input unit 150 receives an operation of inputting data used for operating the terminal device 12 . The input unit 150 may be connected to the terminal device 12 as an external input device instead of being provided in the terminal device 12 .

Control unit 160 includes at least one processor, at least one programmable circuit, at least one dedicated circuit, or any combination thereof. A processor may be a general-purpose processor such as a CPU or GPU, or a dedicated processor specialized for a particular process. A programmable circuit is, for example, an FPGA. A dedicated circuit is, for example, an ASIC. The control unit 160 executes processing related to the operation of the terminal device 12 while controlling each unit of the terminal device 12 .

Storage unit 170 includes at least one semiconductor memory, at least one magnetic memory, at least one optical memory, or any combination thereof. A semiconductor memory is, for example, a RAM. RAM is, for example, SRAM or DRAM. Storage unit 170 may be an EEPROM.

The display unit 180 includes at least one output interface. The output interface is, for example, an LCD or organic EL display. The display unit 180 outputs data obtained by operating the terminal device 12 . The display unit 180 may be provided integrally with a touch panel that can accept input from the user. The display unit 180 displays waveform display data under the control of the control unit 160 .

The communication unit 190 is one or more communication ICs. The communication IC is, for example, a LAN communication IC. The communication unit 190 may be a PCI network adapter, a USB network adapter, or an Ethernet (registered trademark) adapter. The communication unit 190 receives information used for the operation of the terminal device 12 via the network 15 and transmits information obtained by the operation of the terminal device 12 via the network 15 .

Next, operation of the system 10 according to the present embodiment will be described with reference to FIGS. 7 to 11. FIG. FIG. 7 illustrates the operation of system 10 . The operation of FIG. 7 corresponds to the method according to this embodiment.

In step S101, the control section 110 of the waveform measuring instrument 11 sets at least one of a plurality of channels as a trigger channel. Specifically, control unit 110 converts waveforms of analog signals input from a plurality of channels into digital signal data. The control unit 110 sets at least one of the plurality of channels as a trigger channel when performing data acquisition for writing the digital signal data of each channel to the corresponding area in the storage. The storage is the storage unit 120 of the waveform measuring instrument 11 in this embodiment, specifically the waveform memory 121 of the storage unit 120 , but may be a storage device external to the waveform measuring instrument 11 . In that case, waveform measuring instrument 11 may upload digital signal data to an external storage device via network 15 . In the example shown in FIG. 8, four channels CH1 to CH4 are used for trigger measurement in order to monitor the trend of the behavior of the press, which is the processing machine. Channel CH1 is the drive AC voltage of the press, channel CH2 is the displacement of the press obtained from the displacement sensor, channel CH3 is the vibration displacement of the press obtained from the vibration sensor, and channel CH4 is the pressure sensor. An analog signal indicating the amount of change in hydraulic pressure of the press is input. The four waveforms shown in FIG. 8 are obtained by plotting the sample values included in the digital signal data of each channel, with the horizontal axis representing time and the vertical axis representing signal intensity. Of these four channels, one channel CH2 is set as the trigger channel.

Next, the control unit 110 sequentially compares each sample value included in the digital signal data written in the area corresponding to the trigger channel in the storage with the threshold, and detects one sample value as the trigger point. Each time, at least one sample value included in the digital signal data written during a certain period of time including the detection time of the trigger point is read from a region corresponding to each of the plurality of channels in the storage. Specifically, the following steps S102 to S105 are executed.

In step S102, control unit 110 executes trigger point detection for the trigger channel set in step S101. Trigger point detection is data in which digital signal data including a plurality of sample values for each trigger channel is sequentially written in a ring shape in an area having a predetermined number of addresses corresponding to the trigger channel in the storage unit 120 corresponding to storage. Executed during capture. The predetermined number is an integer of 2 or more, and is set in advance as the number of addresses capable of storing sample values for each channel including the trigger channel in storage unit 120 . The control unit 110 performs trigger point detection by comparing each sample value of the plurality of sample values with a threshold value and detecting one sample value from among the plurality of sample values as the trigger point TP. . Specifically, the trigger detection circuit 113 of the control unit 110 determines whether or not the sample value of the digital signal data of the trigger channel satisfies a preset trigger condition. A trigger condition is that a sample value above or below a preset threshold is detected from the sample values contained in the digital signal data acquired for the trigger channel. In the example shown in FIG. 8, the trigger detection circuit 113 sequentially writes digital signal data including a plurality of sample values indicating the amount of displacement of the press for the channel CH2 into the area corresponding to the channel CH2 in the storage unit 120. Perform trigger detection during data acquisition. The trigger detection circuit 113 compares each sample value with a threshold for the amount of displacement of the press. As a result of the comparison, if one sample value exceeding the threshold is detected, the trigger detection circuit 113 detects the detected sample value as the trigger point TP. For example, by setting the threshold near the maximum displacement of the press during the press operation, it is possible to record changes in the signal associated with each series of press operations. The trigger detection circuit 113 notifies the memory controller 112 of the detected trigger point TP. In this way, the control unit 110 sequentially compares each sample value included in the digital signal data written in the area corresponding to the trigger channel in the storage with the threshold, and uses one sample value as a trigger point. To detect.

In step S103, the control unit 110 sets, as a trigger point address nT, the address where the sample value was written when the trigger point TP was detected in step S102 in the area corresponding to each of the plurality of channels. Specifically, the trigger detection circuit 113 of the control unit 110 notifies the memory controller 112 of the trigger point TP. When the trigger point TP is notified, the memory controller 112 stores the address to which the sample value was written at the time of detection of the trigger point TP as the trigger point address nT for the trigger channel.

In step S104, the control unit 110 sets K sample values including the trigger point address nT set in step S103 in the region corresponding to each channel, where K is an integer equal to or greater than 1 and less than the predetermined number. is written as the target address. In the example shown in FIG. 8, the control unit 110 sets the address one or more before the corresponding trigger point address nT to the first of the K sample values in the area corresponding to each of the channels CH1 to CH4. is set as the start address n0 in which the sample value of is written. The top address n0 is defined as an address that is traced back from the trigger point address nT by a number that is preset for each channel. Hereinafter, the sample value written to the address before the trigger point address nT is also called pre-trigger data. The read end address nE is defined as the number of sample values from the start address n0, which is preset for each channel. Below, the number of sample values is also referred to as the number of data points. The start address n0 and the end address nE for reading are notified from the memory controller 112 of the control unit 110 to the memory controller 114 together with data indicating the end of data acquisition when data acquisition after trigger detection is completed. . The number of sample values sequentially read out from the waveform memory 121 is counted by the data number counting circuit 116 . For example, for channel CH1, assume that trigger point address nT is set to address "n1T". Also, assume that the start address n0 is set to an address "n1T-500" that is 500 backwards from the address "n1T". Assume that the number of sample values represented by K is set to 1000. In this case, the control unit 110 sets the 1000th address "n1T+500" counted from the address "n1T-500" as the final read address nE. The addresses from the read start address n0 to the end address nE of each channel correspond to the target address of this embodiment.

In this way, control unit 110 assigns an address one or more before the corresponding trigger point address in the area corresponding to each channel to the leading address where the first sample value of K sample values is written. , and the K-th address counted from the set start address is set as the final read address, thereby specifying the address in which the K sample values are written as the target address.

In step S105, the control unit 110 sequentially reads K sample values from the specified target address in the area corresponding to each channel. Specifically, the memory controller 112 of the control unit 110 notifies the memory controller 114 that data acquisition after trigger detection is completed. At the same time, the memory controller 112 notifies the memory controller 114 of the start address n0 and the end address nE of data write. In response to this, the memory controller 114 specifies the target address from the start address n0 and the end address nE in the area corresponding to each channel in the waveform memory 121 of the storage unit 120, and sequentially samples from the target address of each channel. Read value.

FIG. 9 shows each area in the storage unit 120 corresponding to each channel when data is taken in four channels CH1 to CH4. A sample value of each channel is stored in an area in storage section 120 corresponding to each channel. In the example shown in FIG. 9, in the first to fourth areas corresponding to channels CH1 to CH4, A1, A2, A3, and A4 correspond to the start addresses of the ring buffers, and B1, B2, B3, and B4 correspond to , corresponding to the last address of the ring buffer. n1T, n2T, n3T, and n4T are set as trigger point addresses corresponding to each area. Data acquisition of each channel is synchronized with the same sampling period. Therefore, the read start address (n10, n20, n30, n40) and the end address (n1E, n2E, n3E, n4E) of each channel are relatively the same, and the number of sample values stored in the target address is are also equal. Note that FIG. 3 described above shows how data is overwritten in a ring shape over two round trips to continuous addresses arranged one-dimensionally from the top address A to the end address B of the ring buffer. On the other hand, in FIG. 9, the addresses of each channel are shown as consecutive addresses arranged two-dimensionally. For example, in CH1, there are m+1 addresses from the start address A1 to the end address B1 of the ring buffer. The m+1 addresses correspond to the predetermined number of addresses. The same is true for CH2 to CH4.

In step S106, when J is an integer of 1 or more and K or less, in the region corresponding to each channel, for J sample values out of the K sample values sequentially read, , executes the process of extracting the feature amount. The control unit 110 acquires the execution result as data indicating the feature amount. As a result, the waveform measuring instrument 11 side executes a process of acquiring data indicating a feature amount from a specific part of the raw data and extracting data for machine learning used for AI analysis. Specifically, in the area corresponding to each channel, control unit 110 selects J sample values read from addresses within the range indicated by the set value, out of the K sample values that are sequentially read. Identify. A "set value" is a pair of values indicating the start address and the end address of J sample values, and is set in advance for each channel according to the process to be executed. J may be set to the same or different numbers between channels. In the example shown in FIG. 8, p11 and p12 are set as the set values for channel CH1. The control unit 110 specifies sample values read from the address range from p11 to p12 in the area corresponding to the channel CH1. Similarly, p21 and p22, p23 and p24 are set as the set values for channel CH2, p31 and p32 are set as set values for channel CH3, and p41 and p42 are set as set values for channel CH4. Similarly to channel CH1, control unit 110 specifies sample values read from addresses within ranges indicated by respective set values in regions corresponding to channels CH2 to CH4. Then, control unit 110 executes a process of extracting a feature amount for the identified J sample values. The processing for extracting feature amounts may be one type of processing or two or more types of processing for each channel. In the example shown in FIG. 8, in channel CH1, the root mean square (RMS) of the sample values read from the address range p11 to p12 (dashed line) is performed. Thereby, the feature quantity of the driving AC voltage of the press is obtained. In the channel CH2, a process of obtaining a slope is executed for the sample values sequentially read out from the address range from p21 to p22 (broken line). In channel CH2, processing for obtaining the maximum and minimum values is also executed for the sample values sequentially read out from the address range from p23 to p24 (broken line). Thereby, the feature amount of the displacement amount of the press is acquired. In channel CH3, a process of obtaining an average value is executed for the sample values sequentially read out from the address range from p31 to p32 (broken line). Thereby, the feature quantity of the vibration displacement of the press is obtained. In channel CH4, a process of obtaining the maximum value is executed for the sample values sequentially read out from the address range from p41 to p42 (broken line). As a result, the feature amount of the hydraulic pressure change amount of the press is acquired. Further, the control unit 110 extracts the feature amounts of the sample values read from the regions corresponding to the two or more channels out of the plurality of channels. A process of converting into one feature amount may be executed, and data representing one feature amount may be included in output data. For example, the control unit 110 executes a process of obtaining the inclination of the displacement amount of the press on the channel CH2. The control unit 110 selects the press machine from a combination of the obtained inclination of the displacement amount and the maximum value indicating the characteristic amount of the hydraulic pressure change amount obtained in the channel CH4 (for example, the maximum pressure value/slope of the displacement amount). Acquire the feature value of work efficiency. Then, the control unit 110 can include the acquired data indicating the feature amount of the work efficiency of the press in the data to be output.

FIG. 10 shows how J sample values out of 1000 sample values sequentially read in the regions corresponding to the channels CH1 to CH4 in the storage unit 120 shown in FIG. A specific example of executing different processing for each is shown. 1000 sample values correspond to K sample values in this embodiment. In the example shown in FIG. 10, data reading is performed by reading addresses of each area one by one while circulating areas corresponding to channels CH1 to CH4. That is, data reading is executed in the order of n10→n20→n30→n40→n11→...→n41→n12→...n42→.... 1000 sample values are read for each channel, and 4000 sample values are read in total for the four channels from n10 to n4E. The data number counting circuit 116 of the control unit 110 increments the count value of the counter by 1 each time the sample value of the channel CH1 is read. In the example shown in FIG. 10, the count value pE is 1,000. As shown in FIG. 10, the timing of executing the feature extraction process may overlap between channels. In this embodiment, the timings of executing two different processes within one channel CH2 do not overlap, but the timings of executing processes for extracting one or more feature quantities within the same channel overlap. You may For example, the timing of executing the process of obtaining the slope in channel CH2 and the timing of executing the process of obtaining the maximum and minimum values may overlap. In the channel CH2, the same process for extracting the feature quantity, for example, the process for obtaining the slope is executed on the sample values sequentially read from the address ranges p21 to p22 and p23 to p24. Timing may overlap.

As described above, control unit 110 specifies J sample values read from addresses within a range indicated by preset values for each channel. The settings for channel CH1 are a pair of values p11 and p12. For example, p11 indicates the first address counted from the head address "n10" of CH1, and p12 indicates the 1000th address counted from "n10". In this case, the address corresponding to p11 is "n10" and the address corresponding to p12 is "n1E". Therefore, the control unit 110 identifies 1000 sample values read from the address range from 'n10' to 'n1E'. The set values for channel CH2 are then a pair of values of p21 and p22 and a pair of values of p23 and p24. For example, p21 indicates the 451st address counted from the head address "n20" of CH2, and p22 indicates the 550th address counted from "n20". In this case, the control unit 110 identifies 100 sample values read from the address range from the address corresponding to "p21" to the address corresponding to "p22". For example, p23 indicates the 601st address counted from the head address "n20" of CH2, and p24 indicates the 800th address counted from "n20". In this case, the control unit 110 identifies 200 sample values read from the address range from the address corresponding to p23 to the address corresponding to p24. The set values for channel CH3 are then the pair of values p31 and p32. For example, p31 indicates the first address counted from the head address "n30" of CH3, and p32 indicates the 1000th address counted from "n30". In this case, the control unit 110 identifies 1000 sample values read from the address range from "n30" to "n3E". Finally, the settings for channel CH4 are a pair of values p41 and p42. For example, p41 indicates the 651st address counted from the head address "n40" of CH4, and p42 indicates the 850th address counted from "n40". In this case, the control unit 110 identifies 200 sample values read from the address range from the address corresponding to p41 to the address corresponding to p42. After specifying J sample values for each channel, processing indicator 117 of control unit 110 determines whether the count value of data number counting circuit 116 has become equal to the set value for each channel. As a result of the determination, when a channel whose count value is equal to the set value is specified, the process indicator 117 starts the process of extracting the feature quantity to be executed for the J sample values specified for that channel. The operator 115 is instructed of the type, and the start timing and end timing of the processing. The arithmetic unit 115 executes processing for extracting feature amounts according to instructions from the processing instruction unit 117, and acquires execution results.

In this way, every time one sample value is detected as a trigger point, the control unit 110 controls the area corresponding to each of the plurality of channels in the storage for a certain period of time including the detection time of the trigger point. At least one sample value included in the written digital signal data is read, and a process of extracting a feature amount of the read sample value is executed.

In this embodiment, the control unit 110 executes data acquisition and feature amount extraction processing in parallel. FIG. 11 shows an example in which data acquisition by the above-described trigger measurement is repeatedly performed, and feature amounts are extracted for sample values of digital signal data acquired in each trigger measurement. FIG. 11 shows data loading, data reading, and feature amount extraction in chronological order from the left side to the right side of the paper. As shown in FIG. 11, for the first data acquisition, data readout and feature quantity extraction are started for the sample values written in target addresses from n0 to nE specified through trigger point detection. . The second data acquisition and trigger point detection are started without waiting for the completion of the first data readout and feature amount extraction. The same applies to the third and subsequent data acquisitions and trigger point detections. Thus, in this embodiment, data acquisition and feature amount extraction processing are executed in parallel. In this embodiment, data loading into the waveform memory 121 and data reading from the waveform memory 121 can be executed simultaneously in parallel by pipeline processing by the memory controllers 112 and 114 . That is, it is possible to simultaneously execute data acquisition, data readout, and feature amount extraction in real time. In this case, there is no need to secure an operation time for extracting the feature amount between data acquisition and the next data acquisition. Therefore, even when the trigger measurement is repeatedly performed, it becomes easy to continue data acquisition without causing dead time due to the extraction of the feature amount. Note that the processing speed of the calculator 115 may be the same as the sampling period, and does not need to be faster than the sampling period. This is because pipeline processing is possible if the arithmetic unit 115 can process the fetched data at the same processing speed as the sampling period.

In step S107, control unit 110 outputs data indicating the feature amount extracted in step S106. Specifically, the communication unit 140 of the waveform measuring instrument 11 transmits data indicating the feature amounts of the K sample values that have been acquired to the terminal device 12 via the network 15 . In the example shown in FIG. 8, the communication unit 140 transmits the feature amount acquired for each channel CH1 to CH4 to the terminal device 12 as data indicating the feature amount of K sample values. The communication unit 190 of the terminal device 12 receives, via the network 15 , the data indicating the feature quantity of the K sample values transmitted from the waveform measuring instrument 11 . The control unit 160 of the terminal device 12 accumulates in the storage unit 170 data indicating the feature amounts of the K sample values received by the communication unit 190 . The control unit 160 of the terminal device 12 reads the data indicating the feature amount of the K sample values accumulated in the storage unit 170, and executes AI analysis for the purpose of failure prediction of the processing machine, for example. In this embodiment, the communication unit 140 transmits the data indicating the feature amount of the K sample values to the terminal device 12. Together with the data indicating the feature amount, the K samples read for each channel are sent to the terminal device 12. The value may be sent to terminal device 12 . In that case, the terminal device 12 causes the display unit 180 to display a waveform as shown in FIG. 8 obtained by plotting K sample values for each channel. In the example shown in FIG. 8, the left end of each display waveform is the plot point of the sample value stored at the start address n0 for the corresponding channel, and the right end is the plot point of the sample value stored at the end address nE for the corresponding channel. Corresponds to plot points.

According to the present embodiment, the waveform measuring instrument 11 side can perform the process of acquiring the data indicating the feature amount from a specific portion of the raw data and extracting the data for machine learning used in the AI analysis. Since feature extraction is implemented by hardware and can be processed at the same speed as data acquisition, dead time between data acquisitions does not occur. Therefore, the amount of data communication between the waveform measuring instrument 11 and the terminal device 12 and the data processing load on the terminal device 12 side are reduced compared to the case where the raw data is transmitted to the terminal device 12 as it is. Therefore, it becomes easy to shorten the overall processing time for AI analysis using waveform data while acquiring data indicating a feature amount.

However, the present disclosure is not limited to the embodiments described above. For example, multiple blocks depicted in the block diagrams may be integrated or one block may be divided. Instead of executing the steps described in the flowchart in chronological order according to the description, the steps may be executed in parallel or in a different order depending on the processing power of the device executing each step or as required. Other modifications are possible without departing from the scope of the present disclosure.

For example, the above-described embodiment is based on the assumption that pretrigger data is also acquired, but if pretrigger data is not acquired, data readout and feature extraction may be started at the same time data acquisition is started. In that case, the processing by memory controller 112 and memory controller 114 does not necessarily have to be pipeline processing because there is no dead time between data fetches. FIG. 12 shows the operation of system 10 when pre-trigger data is not acquired as a modification of the present embodiment. As shown in FIG. 12, reading of data and extraction of feature quantities are started simultaneously with the start of data acquisition. That is, the processing by the memory controller 112 and the memory controller 114 becomes sequential processing instead of pipeline processing. Also in this case, similarly to the above method, the number of data points from the start of data acquisition can be set in advance, and the feature amount can be extracted from a specific portion of the raw data.

In the embodiment described above, the sampling period is the same for each channel, as shown in FIG. 8, but the sampling period may be different for each channel. In this case, reading from the waveform memory 121 is performed in accordance with the sampling period of each channel, and the calculation rate of the calculator 115 is adjusted to the sampling period of each channel, so that processing can be performed even if the sampling period differs between channels. be able to.

REFERENCE SIGNS LIST 10 system 11 waveform measuring instrument 12 terminal device 15 network 20 system 21 waveform measuring instrument 22 local PC
100 input section 101-1, 101-2 input amplifier 102-1, 102-2 A/D converter 103-1, 103-2 interface circuit 110 control section 111 sample timing generation circuit 112 memory controller 113 trigger detection circuit 114 memory Controller 115 Arithmetic unit 116 Data counting circuit 117 Processing indicator 120 Storage unit 121 Waveform memory 122 Calculation result memory 130 Display unit 131 Display waveform generation circuit 132 Waveform display unit 140 Communication unit 150 Input unit 160 Control unit 170 Storage unit 180 Display unit 190 communication unit 201, 204 input amplifier 202, 205 A/D converter 203, 206 interface circuit 211 sample timing generation circuit 212 memory controller 221 waveform memory 231 display waveform generation circuit 232 waveform display

Claims (6)

A waveform measuring instrument for capturing data by converting waveforms of analog signals input from a plurality of channels into digital signal data and writing the digital signal data of each channel to corresponding areas in a storage,
setting at least one of the plurality of channels as a trigger channel, and sequentially comparing each sample value included in digital signal data written to an area corresponding to the trigger channel in the storage with a threshold; , every time one sample value is detected as a trigger point, digital signal data written during a certain period including the detection point of the trigger point in an area corresponding to each of the plurality of channels in the storage. A waveform measuring instrument comprising a control unit that reads out at least one sample value included in the , executes a process of extracting a feature amount of the read sample value, and outputs data indicating the extracted feature amount. 2. The waveform measuring instrument according to claim 1, wherein said control unit concurrently executes said data acquisition and the process of extracting said feature amount. The control unit extracts two or more feature amounts obtained by performing a process of extracting feature amounts of sample values read from regions corresponding to two or more channels among the plurality of channels into one 3. The waveform measuring instrument according to claim 1, further performing a process of converting into a feature quantity, and including data representing said one feature quantity in data to be output. At least one of the plurality of channels is converted to digital signal data, and the digital signal data of each channel is written to a corresponding area in the storage. setting one channel as the trigger channel;
Each sample value included in the digital signal data written in the area corresponding to the trigger channel in the storage is sequentially compared with a threshold, and each time one sample value is detected as a trigger point, and reading out at least one sample value included in digital signal data written during a certain period including the detection time of the trigger point, in a region corresponding to each channel of the plurality of channels;
executing a process of extracting a feature quantity of the read sample value;
outputting data indicating the extracted feature amount;
method including. 5. The method of claim 4, further comprising performing the data acquisition and the feature extraction process in parallel. Two or more feature values obtained by performing a process of extracting feature values of sample values read from regions corresponding to two or more channels among the plurality of channels are converted into one feature value. 6. The method of claim 4 or 5, further comprising performing further processing to include data indicative of said one feature quantity in data to be output.

Images