Information processing system, information processing method and program

Abstract

To provide an information processing system, etc. which can detect an abnormality of operation on the basis of a more general-purpose phenomenon.SOLUTION: An information processing system is provided according to an aspect of the present invention. The information processing system includes a processor. The processor is configured to execute a program for execution of the following steps. In an input processing step, a first signal which is a time-series signal is input into a neural network. In a reconstruction step, a second signal is produced which is such first signal as to be reconstructed from a plurality of output signals output by the neural network. In a detection step, an abnormality concerning the first signal is detected on the basis of an error between the first signal and the second signal and a predetermined reference standard.SELECTED DRAWING: Figure 3

Description

[Technical Field]

[0001] The present invention relates to an information processing system, an information processing method, and a program. [Background technology]

[0002] There is a technology that monitors various phenomena, such as physical phenomena, that may occur along with predetermined operations in factories, etc., and detects abnormalities in the operations based on the phenomena. For example, Patent Document 1 discloses a prior art technology that uses sound as a physical phenomenon. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Patent Publication No. 2021-163088 Summary of the Invention [Problem to be solved by the invention]

[0004] In the technology disclosed in Patent Document 1, elements excluding normal sounds that are normally expected during operation are identified as abnormal sounds, but there are cases where normal sounds and abnormal sounds are not linearly linked. In other words, there is a need for an abnormality detection technology based on more general phenomena.

[0005] In view of the above circumstances, the present invention provides an information processing system and the like that can detect operational abnormalities based on more general phenomena. [Means for solving the problem]

[0006] According to one aspect of the present invention, there is provided an information processing system. The information processing system includes a processor. The processor is configured to execute a program that executes the following steps: In the input processing step, a first signal, which is a time-series signal, is input to a neural network; In the reconstruction step, a second signal is generated so as to reconstruct the first signal from a plurality of output signals output from the neural network; and In the detection step, an anomaly related to the first signal is detected based on an error between the first signal and the second signal and a predetermined reference standard.

[0007] According to this aspect, it is possible to detect an abnormality in operation based on a more general phenomenon. [Brief explanation of the drawings]

[0008] [Figure 1] 1 is a block diagram showing a hardware configuration of an information processing device 1 (information processing system) according to a first embodiment. [Figure 2] 2 is a block diagram showing functions realized by a control unit 13 and the like in the information processing device 1. FIG. [Figure 3] FIG. 2 is a schematic diagram showing the configuration of a neural network 2. [Figure 4] 1 is an activity diagram showing the flow of information processing executed by the information processing device 1. FIG. [Figure 5] 2 is a schematic diagram illustrating a time shift between a first signal SG1 and a second signal SG2. FIG. [Figure 6] FIG. 6A shows the case where normal data is input, and FIG. 6B shows the case where abnormal data is input. [Figure 7] FIG. 7A shows the error E(n) when a normal first signal SG1 is evaluated using a non-dominant coupling with a weight w_outjj'<0.05, and FIG. 7B shows the error E(n) when a normal first signal SG1 is evaluated using a dominant coupling with a weight w_outjj'≧0.05. [Figure 8]FIG. 8A shows the error E(n) when the abnormal first signal SG1 is evaluated using a non-dominant coupling with a weight w_outjj'<0.05, and FIG. 8B shows the error E(n) when the abnormal first signal SG1 is evaluated using a dominant coupling with a weight w_outjj'≧0.05. [Figure 9] FIG. 10 is a block diagram showing a hardware configuration of an information processing system according to a second embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0009] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention will be described below with reference to the accompanying drawings. Various features shown in the following embodiments can be combined with each other.

[0010] Incidentally, the program for realizing the software appearing in this embodiment may be provided as a non-transitory computer-readable medium, or may be provided so that it can be downloaded from an external server, or may be provided so that the program is started on an external computer and its functions are realized on a client terminal (so-called cloud computing).

[0011] In this embodiment, the term "unit" may also include, for example, a combination of hardware resources implemented by a circuit in the broad sense and software information processing that can be specifically realized by these hardware resources. In addition, various types of information are handled in this embodiment, and this information may be represented by, for example, physical values ​​of signal values ​​representing voltages and currents, high and low signal values ​​as a binary bit set consisting of 0 or 1, or quantum superposition (so-called quantum bits), and communication and calculations may be performed on a circuit in the broad sense.

[0012] In addition, a circuit in the broad sense is a circuit realized by at least appropriately combining a circuit, circuitry, a processor, a memory, etc. That is, it includes an application specific integrated circuit (ASIC), a programmable logic device (e.g., a simple programmable logic device (SPLD), a complex programmable logic device (CPLD), and a field programmable gate array (FPGA)), etc.

[0013] [First embodiment] 1. Hardware Configuration In this section, the hardware configuration of the first embodiment will be described. In this embodiment, an information processing system is comprised of one or more devices or components. Therefore, for example, even an information processing device 1 alone is an example of an information processing system. Below, the hardware configuration of the information processing device 1, which is an example of an information processing system, will be described.

[0014] 1 is a block diagram showing the hardware configuration of an information processing device 1 (information processing system) according to the first embodiment. The information processing device 1 has a communication unit 11, a storage unit 12, a control unit 13, a display unit 14, and an input unit 15, and these components are electrically connected within the information processing device 1 via a communication bus 10. Each component will be further described below.

[0015] The communication unit 11 is preferably a wired communication means such as USB, IEEE1394, Thunderbolt, or wired LAN network communication, but may also include wireless LAN network communication, mobile communication such as LTE / 5G, Bluetooth (registered trademark) communication, or the like, as necessary. That is, it is more preferable to implement it as a collection of multiple communication means. That is, the information processing device 1 may communicate various information from the outside via the communication unit 11 and a network. Specifically, the information processing device 1 is connected to a sensor 51 via the communication unit 11. The type of sensor 51 is not particularly limited, and a microphone (a broadly defined audio sensor), a camera (a broadly defined vision sensor), a temperature sensor, an infrared sensor, a pressure sensor, a photoelectric sensor, a color sensor, an ultrasonic sensor, a displacement sensor, or the like may be used as appropriate.

[0016] The memory unit 12 stores various pieces of information defined above. This can be implemented, for example, as a storage device such as a solid state drive (SSD) that stores various programs and the like related to the information processing device 1 executed by the control unit 13, or as a memory such as a random access memory (RAM) that stores temporarily required information (arguments, arrays, etc.) related to the program operations. The memory unit 12 stores various programs, variables, etc. related to the information processing device 1 executed by the control unit 13.

[0017] The control unit 13 (processor) processes and controls the overall operations related to the information processing device 1. The control unit 13 is, for example, a central processing unit (CPU) not shown. The control unit 13 realizes various functions related to the information processing device 1 by reading out predetermined programs stored in the storage unit 12. In other words, information processing by software stored in the storage unit 12 is specifically realized by the control unit 13, which is an example of hardware, and can be executed as each functional unit included in the control unit 13. These will be described in more detail in the next section. Note that the control unit 13 is not limited to being single, and multiple control units 13 may be provided for each function. A combination of these may also be used.

[0018] The display unit 14 may be, for example, included in the housing of the information processing device 1 or may be externally attached. The display unit 14 displays a graphical user interface (GUI) screen that can be operated by the user. This is preferably implemented by selectively using display devices such as a CRT display, a liquid crystal display, an organic EL display, and a plasma display.

[0019] The input unit 15 may be included in the housing of the information processing device 1 or may be externally attached. For example, the input unit 15 may be implemented as a touch panel integrated with the display unit 14. The touch panel allows the user to input tapping, swiping, and the like. Of course, a switch button, a mouse, a QWERTY keyboard, or the like may be used instead of the touch panel. That is, the input unit 15 accepts an operation input made by the user. The input is transferred as a command signal to the control unit 13 via the communication bus 10, and the control unit 13 can execute predetermined control or calculation as necessary.

[0020] 2. Functional configuration In this section, the functional configuration of this embodiment will be described. As described above, information processing by software stored in the storage unit 12 is specifically realized by the control unit 13, which is an example of hardware, and each functional unit included in the control unit 13 can be executed. In other words, the information processing device 1, which is an example of an information processing system, includes the control unit 13 (processor). The control unit 13 is configured to execute a program that causes each functional unit to execute steps.

[0021] 2 is a block diagram showing functions realized by the control unit 13 and the like in the information processing device 1. Specifically, the information processing device 1 includes a reception unit 131, a conversion unit 132, an input processing unit 133, an output processing unit 134, a reconstruction unit 135, a determination unit 136, a detection unit 137, and a transmission unit 138.

[0022] The receiving unit 131 is configured to receive various pieces of information received from the outside via the communication unit 11. For example, as an example of the receiving step, the receiving unit 131 may receive the original signal SG0, which is a time-series signal.

[0023] The conversion unit 132 is configured to convert various pieces of information handled by the information processing device 1. For example, as an example of the conversion step, the conversion unit 132 may convert the received original signal SG0 into a spectrogram decomposed into each frequency component.

[0024] The input processing unit 133 is configured to input necessary information as input to the neural network 2 (see FIG. 3) pre-stored in the storage unit 12. For example, as an example of the input processing step, the input processing unit 133 may input a first signal SG1, which is a time-series signal, to the neural network 2.

[0025] The output processing unit 134 is configured to output various information from the neural network 2 (see FIG. 3) stored in advance in the storage unit 12.

[0026] The reconstruction unit 135 is configured to reconstruct desired information from the information output by the output processing unit 134. For example, as an example of the reconstruction step, the reconstruction unit 135 may generate a second signal SG2 so as to reconstruct a first signal SG1 from a plurality of output signals output by the neural network 2.

[0027] The determination unit 136 is configured to determine various pieces of information. For example, the determination unit 136 may use a reference standard RC stored in the storage unit 12 to determine whether or not an abnormality exists.

[0028] As an example of the detecting step, the detecting unit 137 may detect an abnormality in the first signal SG1 based on the determination result by the determining unit 136.

[0029] The transmitter 138 is configured to output or transmit various types of information. The various types of information may be various types of signals SG, audio data related to an abnormality, or display information. The display information may be visual information itself generated in a manner that is visible to the user, such as a screen, an image, an icon, or text, or may be rendering information for displaying a screen, an image, an icon, or text on the display unit 14, for example.

[0030] 3. Construction of Neural Network 2 In this section, the configuration of the neural network 2 in the information processing device 1 will be described.

[0031] 3 is a schematic diagram showing the configuration of the neural network 2. The neural network 2 includes an input layer 21, a processing layer 22, and an output layer .

[0032] A first signal SG1 is input as an input signal to the input layer 21. The input layer 21 has input terminals 211 to 213, for example, but this is merely an example and is not limiting.

[0033] The processing layer 22 is the main layer of the neural network 2, and is connected to a plurality of neurons. Specifically, various neurons 31 to 37 are shown in FIG. 3. The calculation results passing through the connection lines are transmitted from the side without the open circle to the side with the open circle. For example, neuron 31 receives input from input terminal 211, assigns weights to neurons 32 and 34, and transmits the calculation results. Neuron 32 receives input from neuron 31, assigns weights to output neuron 231 (described later), and transmits the calculation results.

[0034] Particularly preferably, the processing layer 22 includes a reservoir 3 having a recurrent connection. As an example of a recurrent connection, a signal SG is input from an input terminal 212 to a neuron 34, and then passes through neurons 37, 36, and 33 before being re-input to neuron 34 again. According to this embodiment, reservoir computing using a time-series signal as an input can be realized in the field of reservoir computing, which generally does not use time-series information.

[0035] The output layer 23 includes output neurons 231 to 233. That is, the output signals processed through the reservoir 3 are output via the output neurons 231 to 233, and these element signals are reconstructed by the above-mentioned reconstruction unit 135 to obtain the second signal SG2.

[0036] 4. Information Processing Method In this section, an information processing method of the information processing device 1 described above will be explained. The information processing method according to this embodiment includes each step of the information processing device 1, which is an example of the information processing system described above. FIG. 4 is an activity diagram showing the flow of information processing executed by the information processing device 1. The following explanation will be given along with each activity in this activity diagram.

[0037] In this embodiment, as an example, the information processing device 1 is a device related to dedicated devices such as robots and machine tools installed in a factory, and the sensor 51 is a microphone capable of collecting (detecting) the operating sounds of these devices. That is, the sensor 51 records an audio signal as a time-series signal, and this is transmitted to the information processing device 1 as an original signal SG0 via the communication unit 11. That is, the sensor 51 is assumed to have a one-to-one correspondence with the information processing device 1 (information processing system). According to this embodiment, it is possible to easily determine the location of the source of the abnormality, thereby improving usability.

[0038] First, the sensor 51 continuously collects the operating sound of the device as a time-series signal SG. Then, as an example of a receiving step, the receiving unit 131 receives the operating sound as an original signal SG0, which is a time-series signal (activity A101). In other words, the receiving unit 131 receives the original signal SG0 (time-series signal) from the sensor 51, which has a one-to-one correspondence with the information processing system. The original signal SG0 may be received in an uncompressed format such as wav or aiff, or in various compressed formats such as aac, wmv, or mp3. The received original signal SG0 is stored in a temporary storage area of ​​the storage unit 12.

[0039] Next, the conversion unit 132 performs a Fourier transform on the original signal SG0 stored in the temporary storage area of ​​the storage unit 12 using a dedicated program previously stored in the storage unit 12 (activity A102). In other words, as an example of a conversion step, the conversion unit 132 converts the received original signal SG0 into a spectrogram resolved into each frequency component.

[0040] Next, as an example of an input processing step, the input processing unit 133 inputs the first signal SG1, which is a time-series signal, to the neural network 2. More preferably, the input processing unit 133 inputs a spectrogram as the first signal SG1 to the neural network 2 (activity A103). That is, the input first signal SG1 is input to the input terminals 211 to 213. In particular, it is preferable that the frequency components contained in the first signal SG1 are classified, and then different frequency components are input to the input terminals 211 to 213, respectively. According to this embodiment, the input signal SG can be analyzed more quantitatively, including in the frequency domain, compared to when the original signal SG0 is directly input to the neural network 2. As a result, more accurate anomaly detection can be achieved.

[0041] Next, in reservoir 3 included in processing layer 22, an operation including a recurrent operation is performed on the input first signal SG1 (activity A104). That is, processing layer 22 is configured to hold the first signal SG1 previously input to input layer 21 for a finite time. This makes it possible to use components of the first signal SG1 input before a certain time in order to output a second signal SG2 corresponding to the first signal SG1 corresponding to the certain time.

[0042] Next, the output processing unit 134 outputs the signal SG that has passed through and been processed in the processing layer 22 as multiple output signals to the output layer 23 (activity A105). After that, the components of the output signals output from the neurons 32, 35, and 37 of the reservoir 3 are distributed and input to the output neurons 231, 232, and 233.

[0043] Subsequently, as an example of a reconstruction step, the reconstruction unit 135 generates a second signal SG2 so as to reconstruct the first signal SG1 from a plurality of output signals output from the neural network 2 (activity A106). More specifically, the reconstruction unit 135 generates the second signal SG2 based on the output signals and a trained model that has undergone machine learning of the weighting in the output layer 23. According to this aspect, it is possible to realize the generation of a highly accurate second signal SG2 while reducing the processing load required for machine learning.

[0044] The second signal SG2 output from the output layer 23 is preferably a signal generated to reconstruct the first signal SG1 at or before the time of input. In other words, the teacher signal for machine learning the weighting in the output layer 23 is preferably the first signal SG1 at or before the time of input to the input layer 21. According to this embodiment, the accuracy of the reconstruction itself is improved compared to reconstructing a future predicted signal, which is often performed in the neural network 2, and as a result, more accurate anomaly detection can be achieved.

[0045] Next, the determination unit 136 performs various determinations on the generated second signal SG2. Specifically, the determination unit 136 extracts the difference between the first signal SG1 and the second signal SG2 (activity A107). Next, the determination unit 136 refers to a reference standard RC pre-stored in the storage unit 12 or the like (activity A108). By comparing these differences with the reference standard RC, the presence or absence of an abnormality in the equipment installed in the factory (an abnormality related to the first signal SG1) is determined. In other words, the detection unit 137 detects an abnormality related to the first signal SG1 based on the error between the first signal SG1 and the second signal SG2 and the predetermined reference standard RC (activity A109). If an abnormality is detected, visual information including a notification notifying the user of the abnormality may be displayed on the display unit 14. Furthermore, the transmission unit 138 may transmit a notification notifying the user of the abnormality to a separate terminal (not shown) accessible to the user. The information processing described above may be repeated until an abnormality in operation is detected in activity A109.

[0046] It should be noted that the second signal SG2 is a signal generated to reconstruct the first signal SG1 at or before the time of input, and therefore when extracting the difference, the difference from the first signal SG1 at an appropriate time is extracted. For example, if the second signal SG2 is a signal generated to reconstruct the first signal SG1 one unit time ago, the second signal SG2 and the first signal SG1 one unit time ago may be compared to extract the difference therebetween.

[0047] Furthermore, the reference standard RC may be a lookup table that compiles information on thresholds related to the difference (error) between the first signal SG1 and the second signal SG2. If the difference is equal to or exceeds the threshold, the determination unit 136 may determine that there is an abnormality in a factory device related to the information processing device 1.

[0048] More preferably, the reference standard RC includes information regarding the conditions for adoption. Furthermore, any signal among the first signals SG1 that does not satisfy the conditions for adoption is preferably not used in the detection step by the detector 137. According to this embodiment, signals SG that are not desirable for anomaly detection can be eliminated, thereby achieving more accurate anomaly detection. In particular, it is preferable not to use the signal SG at the beginning (within a predetermined time), since it is likely to depend on the initial internal state of the reservoir 3.

[0049] In other words, the adoption condition relates to the time from the start of the input processing unit 133. That is, it is preferable that the first signal SG1 within a predetermined time from the start of the input processing unit 133 is not used in the detection step by the detection unit 137.

[0050] The predetermined times (also understood as numbers) that are not adopted are, for example, 10, 20, 30, 40, 50, 60, 70, 80, 90, 100, 110, 120, 130, 140, 150, 160, 170, 180, 190, 200, 210, 220, 230, 240, 250, 260, 270, 280, 290, 300, 310, 320, 330, 340, 350, 360, 370, 380, 390, 400, 410, 420, 430, 440, 450, 460, 470, 480, 490, 500, 510, 520, 530, 540, 550, 560, 570, 580, 590, 600, 610, 620, 630, 640, 650, 660, 670, 680, 690, 700, 710, 720, 730, 740, 750, 760, 770, 780, 790, 800, 810, 820, 830, 840, 850, 860, 870, 880, 890, 900, 910, 920, 930, 940, 950, 960, 970, 980, 990, 1000, 20, 530, 540, 550, 560, 570, 580, 590, 600, 610, 620, 630, 640, 650, 660, 670, 680, 690, 700, 710, 720, 730, 740, 750, 760, 770, 780, 790, 800, 810, 820, 830, 840, 850, 860, 870, 880, 890, 900, 910, 920, 930, 940, 950, 960, 970, 980, 990, 1000, and may be within a range between any two of the numerical values ​​exemplified herein.

[0051] 4, if the adoption conditions related to the predetermined time, etc. are not met, the process transitions to the signal processing of the next unit time before the process of comparing the difference with the threshold (skipping the comparison with the threshold). This is merely an example, and such skip processing may be performed before activities A101 to A103.

[0052] According to this aspect, it is possible to detect an abnormality in operation based on a more general phenomenon.

[0053] 5. Theory and Experimental Results This section describes the theory using mathematical formulas and the experimental results of anomaly detection using neural network 2. Figure 5 is a schematic diagram showing the time shift between the first signal SG1 and the second signal SG2.

[0054] A first signal SG1, which is an input time-series signal of normal data, is successively input to the reservoir 3. At the same time, the first signal SG1, which is appropriately time-shifted (denoted as n_shift in FIG. 5), is presented to the output layer 23. Then, E(n), which is the norm2 (root mean square error (RMSE) averaged over k) of the difference between the second signal SG2, which is a temporary output signal during learning, and the first signal SG1, which is a teacher signal, is calculated for each time (see Equation 1).

number

[0055] This can be considered an instantaneous error, and it is preferable to perform learning so that this value becomes small. For example, the weights can be updated every time using the gradient method to obtain the optimal weights W{^opt}_out (see Equation 4). Alternatively, if the signal is statistically in a steady state, the optimal weights W{^opt}_out (see Equation 4) can be obtained using a matrix X (see Equation 2, where x(n) is the output signal from each neuron) defined by arranging the time-series signals obtained from the reservoir 3 and a matrix S_out (see Equation 3) defined by arranging the teacher signals.

number

number

number

[0056] where X_dagger is the pseudoinverse matrix of X, N' is the number of x(n) (excluding unused elements), and f_inverse is the inverse function of f, acting on each element. If n_shift≧0, the reservoir 3 learns to output the past first signal SG1 (input waveform) by making good use of the internal delay, and this task is called reconstructing the input signal. On the other hand, if n_shift<0, the future input waveform is predicted. In this embodiment, n_shift≧0 is adopted as the preferred mode.

[0057] An experiment was conducted in which normal or abnormal data was input to a trained neural network with n_shift=1. The error E(n) at this time is shown in Figure 6. Figure 6A shows the case where normal data was input, and Figure 6B shows the case where abnormal data was input. In particular, the floor of the error E(n) is higher when the abnormal data in Figure 6B is input than when the normal data in Figure 6A is input. It is also confirmed that the fluctuation is larger when abnormal data is used. This shows that by setting an appropriate threshold for E(n), it is possible to detect anomalies by reconstructing the input signal.

[0058] 7A shows the error E(n) when a normal first signal SG1 is evaluated using a non-dominant coupling with a weight w_outjj'<0.05, and FIG. 7B shows the error E(n) when a normal first signal SG1 is evaluated using a dominant coupling with a weight w_outjj'≧0.05. FIG. 8A shows the error E(n) when an abnormal first signal SG1 is evaluated using a non-dominant coupling with a weight w_outjj'<0.05, and FIG. 8B shows the error E(n) when an abnormal first signal SG1 is evaluated using a dominant coupling with a weight w_outjj'≧0.05.

[0059] When evaluating the abnormal first signal SG1, the error values ​​are large regardless of the weight w_outjj', but the trends differ. When only non-dominant connections are used, the error floors rise significantly in all cases. The values ​​show similar responses to both normal and abnormal data, indicating that anomaly detection is not performed well. When only dominant connections are used, the error floor rises slightly, but not by a large margin. However, the error floor for normal data is as high as that for abnormal data, indicating that it is again impossible to distinguish between normal and abnormal data. The fluctuations are also large. In other words, it can be said that the neurons in reservoir 3 include neurons that determine the general trend and neurons that adjust the details.

[0060] This result shows that although reservoir 3 performs some processing as an autoencoder, unlike an autoencoder, anomaly detection does not simply require narrowing down the information using a small number of neurons, but rather requires skillfully combining small signals. This can also be seen as the fact that time series signals, unlike static signals, vary over time, and therefore require many neurons to reconstruct them.

[0061] [Second embodiment] Next, an information processing device 1 according to a second embodiment will be described. Note that descriptions of configurations and functions that are substantially the same as those of the information processing device 1 according to the first embodiment will be omitted. FIG. 9 is a block diagram showing the hardware configuration of an information processing system according to the second embodiment. As shown in the figure, the information processing system may include components other than the information processing device 1. Specifically, the information processing system according to the second embodiment includes the information processing device 1 and an external device 5 including a sensor 51 and the like, and these components are connected via a communication network 4.

[0062] As described above, in the second embodiment, the information processing device 1 is a server (cloud type), and the information processing shown in FIG. 4 can be adopted in the form of, for example, SaaS (Software as a Service) or cloud computing. In such a case, as an example of a receiving step, the receiving unit 131 may receive the original signal SG0 (time-series signal) from the plurality of sensors 51 or the external device 5 via the communication network 4 (activity A101). Even with this aspect, it is possible to detect an abnormality in operation based on a more general phenomenon. In particular, it is possible to collectively use information transmitted from the external device 5 or the plurality of sensors 51, etc., and apply this to more advanced data utilization (so-called big data) and machine learning such as deep learning.

[0063] [others] The information processing device system described above may adopt the following aspects.

[0064] In the above-described embodiments, various functions have been described as components of the information processing device 1, but a program may be provided to cause a computer to execute each step of the information processing device 1, which is an example of an information processing system.

[0065] In the above-described embodiment, the receiving unit 131 receives the original signal SG0, which is a time-series signal, from the sensor 51, the conversion unit 132 performs a Fourier transform to generate a spectrogram, and the input processing unit 133 inputs the first signal SG1, which is the spectrogram, to the neural network 2. However, the time-series signal received by the receiving unit 131 from the sensor 51 may be input to the neural network 2 as the first signal SG1 without being converted.

[0066] In the above embodiment, the reference standard RC is described as a lookup table, but the reference standard RC is not limited to a lookup table and may be a database, a mathematical model in which a plurality of pieces of information are mathematically related, or a trained model in which the correlation of a plurality of pieces of information is previously learned by machine learning. Specifically, the reference standard RC may be a trained model in which the error between the first signal SG1 and the second signal SG2 and the presence or absence of an abnormality in operation are learned.

[0067] In this embodiment, the reception unit 131, the conversion unit 132, the input processing unit 133, the output processing unit 134, the reconstruction unit 135, the judgment unit 136, the detection unit 137, and the transmission unit 138 are described as functional units realized by the control unit 13 of the information processing device 1, but an information processing system may be realized by implementing at least a part of these in another device.

[0068] Furthermore, it may be provided in the following aspects. In the information processing system, the neural network includes an input layer, a processing layer, and an output layer, the processing layer is configured to hold the first signal that was previously input to the input layer for a finite period of time, and the second signal output from the output layer is a signal generated to reconstruct the first signal at or before the time of input. In the information processing system, the processing layer includes a reservoir having a recurrent connection. In the information processing system, the reconstruction step generates the second signal based on the output signal and a trained model in which weighting in the output layer is machine-learned. In the information processing system, the teacher signal for the machine learning is the first signal at or before the time it was input to the input layer. In the information processing system, further, in the receiving step, an original signal which is a time series signal is received, and further, in the converting step, the received original signal is converted into a spectrogram resolved into each frequency component, and in the input processing step, the spectrogram is input to the neural network as the first signal. In the information processing system, the receiving step further includes receiving a time-series signal from a sensor having a one-to-one correspondence with the information processing system. In the information processing system, the receiving step further receives time-series signals from a plurality of sensors or external devices via a communication network. In the information processing system, the reference standard has information regarding an adoption condition, and a signal among the first signals that does not satisfy the adoption condition is not used in the detection step. In the information processing system, the adoption condition relates to the time from the start of the input processing step, and any of the first signals that are within a predetermined time from the start of the input processing step are not used in the detection step. An information processing method comprising the steps of the information processing system. A program that causes a computer to execute each step of the information processing system. Of course, this is not the case.

[0069] Finally, while various embodiments of the present invention have been described, these are presented by way of example only and are not intended to limit the scope of the invention. The novel embodiments may be embodied in various other forms, and various omissions, substitutions, and modifications may be made without departing from the spirit of the invention. Such embodiments and modifications are intended to be included within the scope and spirit of the invention, as well as within the scope of the inventions and their equivalents as defined in the accompanying claims.

[0070] 1: Information processing equipment 10: Communication bus 11: Communications Department 12: Storage section 13: Control section 131: Reception 132: Conversion section 133: Input processing section 134: Output processing section 135:Reconstruction section 136: Judgment section 137:Detection unit 138: Transmission unit 14: Display section 15: Input section 2: Neural Networks 21: Input layer 211: Input terminal 212: Input terminal 213: Input terminal 22: Processing layer 23: Output layer 231: Output neuron 232: Output neuron 233: Output neuron 3: Reservoir 31: Neuron 32: Neuron 33: Neuron 34: Neuron 35: Neuron 36: Neuron 37: Neuron 4: Communication network 5: External equipment 51: Sensor RC: Reference standard SG: Signal SG0: Original signal SG1: First signal SG2: Second signal

Claims

1. An information processing system, a processor; The processor is configured to execute a program that performs the following steps: In the input processing step, a first signal, which is a time series signal, is input to the neural network, where: the neural network includes an input layer, a processing layer, and an output layer; the processing layer includes a reservoir having a recurrent connection, whereby the reservoir is configured to hold the first signal previously input to the input layer for a finite time; In the reconstruction step, a second signal is generated based on a plurality of output signals output from the neural network and a trained model in which weighting in the output layer has been machine-learned, so as to reconstruct the first signal at or before the time of input, wherein a teacher signal for machine learning is the first signal at or before the time of input to the input layer; In the detecting step, an abnormality related to the first signal is detected based on an error between the first signal and the second signal and a predetermined reference standard.

2. 2. The information processing system according to claim 1, Furthermore, in the receiving step, the original signal, which is a time-series signal, is received, Furthermore, in the conversion step, the received original signal is converted into a spectrogram decomposed into each frequency component; In the input processing step, the spectrogram is input to the neural network as the first signal.

3. 3. The information processing system according to claim 1, Furthermore, the receiving step receives a time series signal from a sensor that has a one-to-one correspondence with the information processing system.

4. In the information processing system according to any one of claims 1 to 3, Furthermore, in the receiving step, time-series signals are received from a plurality of sensors or external devices via a communication network.

5. In the information processing system according to any one of claims 1 to 4, the reference standard comprises information regarding employment conditions; Any of the first signals that does not satisfy the adoption condition is not used in the detection step.

6. 6. The information processing system according to claim 5, the adoption condition relates to a time period from the start of the input processing step, Any of the first signals that are present within a predetermined time period after the start of the input processing step is not used in the detection step.

7. An information processing method, comprising: A method comprising the steps of the information processing system according to any one of claims 1 to 6.

8. A program, A computer that executes each step of the information processing system according to any one of claims 1 to 6.

Images