JP2022032631A - Control system incorporating abnormality detection function and abnormality detection method thereof - Google Patents

Abstract

To detect abnormality of a rotary machine with a higher frequency resolution without increasing costs of computation resources and memory resources.SOLUTION: A control system 1 incorporating an abnormality detection function includes a signal preprocessing unit 121, a predictive model unit 122, and an abnormality determination unit 123. The signal preprocessing unit 121 includes a target spectrum unit 1213, a window function unit 1211, and an FFT function unit 1212. The target spectrum unit 1213 computes a range of a target frequency spectrum on the basis of a rotational speed signal ω. The window function unit 1211 and the FFT function unit 1212 compute FFT. The predictive model unit 122 predicts abnormality by AI on the basis of a frequency spectrum after FFT processing, and the abnormality determination unit 123 detects whether a motor is abnormal or not on the basis of a predictive value and so on inputted from the predictive model unit 122.SELECTED DRAWING: Figure 3

Description

The present invention relates to a control system incorporating an abnormality detection function and an abnormality detection method thereof.

In recent years, a system for determining an abnormality to be detected has been realized by using AI (Artificial Intelligence) technology and IoT (Internet of Things) technology. For example, in an abnormality detection system for a rotating machine operated by a motor that mounts an AI model, the FFT spectrum of motor control signals (current, hall, etc.) and environmental signals (vibration, temperature, etc.) that are input signals to the AI model. By analyzing the above, the presence or absence of abnormalities in the rotating machine is detected. According to such an abnormality detection system, the state of the detection target can be clearly observed, so that the labor of the operator who handles the detection target can be reduced (see, for example, Patent Document 1).

Japanese Unexamined Patent Publication No. 2018-180705

In the abnormality detection system described in Patent Document 1, the presence or absence of abnormality in the detection target signal obtained from the monitor signal to be detected is detected by using the detection algorithm. Here, when the object to be detected is a rotating machine, the monitor signal to be detected is often subjected to FFT (Fast Fourier Transform) processing.

In such an anomaly detection system using FFT (Fast Fourier Transform), there is a demand to improve the accuracy of anomaly detection by increasing the resolution of the FFT spectrum.

However, in order to increase the resolution of the entire FFT spectrum, it is necessary to increase the number of sample points (frequency bins) of the FFT, which increases the calculation load of the FFT. Further, in order to implement the FFT sample points, it is necessary to accumulate a considerable amount of data, and sufficient memory resources for calculation are required. As a result, there is a problem that the cost of the system increases.

In addition, in such an abnormality detection system, the state of the device (rotating machine, etc.) to be detected is detected and fed back to the operation (control) of the device in real time to improve the performance of the device or to fail the device. Efforts have begun to utilize it to avoid the problem. Such a method has an advantage that various signals obtained from the inside of the device can be used. On the other hand, since the processing of these various signals is performed by an internal arithmetic processing unit such as a CPU, there is also a problem that it is difficult to increase calculation resources and memory resources from the viewpoint of power consumption and cost.

Other issues and novel features will become apparent from the description and accompanying drawings herein.

The present invention has been made in view of the above, and one of the purposes thereof is to detect anomalies in rotating devices and the like with higher frequency resolution without increasing the cost of computational resources and memory resources. It is an object of the present invention to provide a control system having a built-in abnormality detection function capable of capable of detecting an abnormality and a method for detecting the abnormality. In this specification, the configuration of the present invention will be described mainly by taking a rotating machine as an example, but an abnormality of a machine or device capable of detecting an abnormality by analyzing a frequency spectrum, for example, an abnormality of a machine or device handling electric power. The present invention is also applicable to detection.

According to one embodiment, a control system incorporating an abnormality detection function for detecting an abnormality in a machine uses a control signal to the machine to set the target frequency spectrum of the control signal according to the cause of the abnormality in the machine. An abnormality is predicted by AI based on the signal preprocessing unit that performs FFT processing and the frequency spectrum after FFT processing by the signal preprocessing unit, or from the frequency spectrum after FFT processing using the signal inside the control system. It is provided with a prediction model unit for extracting an abnormality necessary for the purpose and predicting an abnormality by AI, and an abnormality determination unit for determining the presence or absence of an abnormality in the machine based on the predicted value of the prediction model unit. The signal preprocessing unit includes a target spectrum unit that calculates a range of the target frequency spectrum according to the cause of the abnormality based on a control signal to the machine or a detection signal of a sensor that detects the environment of the machine, and a target spectrum unit. It includes a filter unit that removes frequency components above the upper limit of the calculated target frequency spectrum, and an FFT function unit that performs FFT processing on the target frequency spectrum from which the high frequency band has been removed by the filter unit.

According to one embodiment, a control system having an abnormality detection function capable of detecting an abnormality in a rotating device with a higher frequency resolution without increasing the cost of calculation resources and memory resources, and an abnormality detection method thereof. Can be provided.

It is a block diagram which shows an example of the control system which incorporated the abnormality detection function which concerns on Embodiment 1. FIG. It is a block diagram which shows an example (current input) of the control system which incorporated the abnormality detection function using the AI model which concerns on Embodiment 1. FIG. It is a block diagram which shows another example (target frequency spectrum) of the control system which incorporated the abnormality detection function using the AI model which concerns on Embodiment 1. FIG. It is a flowchart which shows an example of the abnormality detection processing executed by the control system which incorporated the abnormality detection function shown in FIG. It is a block diagram which shows an example of the control system which incorporated the abnormality detection function which concerns on Embodiment 2. It is a block diagram which shows an example (target frequency spectrum) of the control system which incorporated the abnormality detection function using the AI model which concerns on Embodiment 2. FIG. It is a flowchart which shows an example of the abnormality detection processing executed by the control system which incorporated the abnormality detection function shown in FIG. It is a block diagram which shows an example (target frequency spectrum) of the control system which incorporated the abnormality detection function using the AI model which concerns on Embodiment 3. FIG. It is a flowchart which shows an example of the abnormality detection processing executed by the control system which incorporated the abnormality detection function shown in FIG.

In the following embodiments, when necessary for convenience, the description will be divided into a plurality of sections or embodiments, but unless otherwise specified, they are not unrelated to each other, one of which is the other. It is related to some or all of the modified examples, details, supplementary explanations, etc. Further, in the following embodiments, when the number of elements (including the number, numerical value, quantity, range, etc.) is referred to, when it is specified in particular, or when it is clearly limited to a specific number in principle, etc. Except for this, the number is not limited to the specific number, and may be more than or less than the specific number. Furthermore, in the following embodiments, the components (including element steps and the like) are not necessarily essential unless otherwise specified or clearly considered to be essential in principle. Needless to say. Similarly, in the following embodiments, when the shape, positional relationship, etc. of the constituent elements are referred to, the shape is substantially the same, except when it is clearly stated or when it is considered that it is not clearly the case in principle. Etc., etc. shall be included. This also applies to the above numerical values and ranges.

Hereinafter, embodiments will be described in detail with reference to the drawings. In all the drawings for explaining the embodiment, the members having the same function are designated by the same reference numerals, and the repeated description thereof will be omitted. Further, in the following embodiments, the same or similar parts will not be repeated in principle unless it is particularly necessary.

(Embodiment 1)
Conventionally, a motor control system for controlling a rotating machine that rotates by itself or a motor that operates a machine that should rotate (hereinafter collectively referred to as "rotating machine") detects an operation abnormality of the motor. It was configured to incorporate an abnormality detection system for this purpose as a separate unit. In such a configuration, the current signal from the motor control system and the current value detected from the motor driver (motor drive device) are input to the AI-based FFT spectrum analyzer. However, for such inputs, for example, it is necessary to utilize peripheral functions of the motor control system or install additional circuits to detect the current signal without the switching effect from the motor driver. There is.

On the other hand, the control system incorporating the abnormality detection function of the present embodiment is a control system equipped with the abnormality detection function for various signals in the motor control system by incorporating the abnormality detection function into the motor control system as a system. Can be utilized in. Therefore, this control system can reduce the cost for detecting a control signal (for example, current, rotation speed, etc., also called a control command) and providing an additional circuit for inputting the control signal. can.

<Configuration of control system with built-in abnormality detection function>
First, an example of the configuration of the control system 1 incorporating the abnormality detection function will be described. FIG. 1 is a configuration diagram showing an example of a control system incorporating an abnormality detection function according to the first embodiment. In the present embodiment, the control system 1 having an abnormality detection function (hereinafter, also simply referred to as “this control system 1”) has an abnormality detection function for detecting an operation abnormality in a rotating machine such as a motor 3. It has a configuration that is integrated into the control system.

As shown in FIG. 1, the control system 1 incorporating the abnormality detection function according to the present embodiment is a system for controlling the drive of the motor 3 and detecting the abnormality of the motor 3. The operation of the motor 3 is controlled by the motor driver 2. The motor driver 2 controls the rotation of the motor 3 based on a control signal input from the control system 1 as a motor control system.

In the vicinity of the motor 3, a position / speed sensor 5 for detecting the position (rotational position) of a rotor (not shown) in the motor 3 and detecting the rotation speed of the rotor is provided. Further, an environment sensor 4 for detecting the environment of the motor 3 (for example, vibration, temperature, humidity, etc.) is provided in the vicinity of the housing of the motor 3. Further, the electric wire between the motor driver 2 and the motor 3 is provided with a current sensor 6 for detecting the current supplied from the motor driver 2 to the motor 3.

The control system 1 controls the entire system including the motor 3, and also has a microcomputer (MCU) 10 for executing a function of detecting an abnormality of a rotating machine and executing a control of the motor, and motor control and abnormality detection. It is provided with a memory 20 for storing various applications and data necessary for the system.

The MCU 10 may be realized by a general-purpose integrated circuit (ASSP) or the like for a specific application. Further, the MCU 10 includes a CPU 11 and a peripheral function unit 14. Further, the CPU 11 includes an abnormality detection unit 12 and a motor control unit 13. The peripheral function unit 14 includes a data input unit 15.

Various data detected by the environment sensor 4, the position / speed sensor 5, and the current sensor 6 are input to the data input unit 15. The data input unit 15 outputs these detection data to the CPU 11.

The abnormality detection unit 12 detects the presence or absence of an abnormality in the motor 3 based on the detection data input from the data input unit 15. Specifically, it will be described later with reference to FIGS. 2 and 3. The motor control unit 13 has the same function as that of the conventional motor control system. The motor control unit 13 has, for example, a control signal (current value) to be output via the peripheral function unit 14, a current value detected by the current sensor 6, and a rotation speed of the motor 3 detected by the position / speed sensor 5. Feedback control is performed based on.

For example, the preprocessing data 21, the prediction model data 22, and the motor control data 23 are stored in the memory 20. The preprocessing data 21 and the prediction model data 22 will be described later. The motor control unit 13 controls the current value (current signal) to be input to the motor 3 based on the motor control data 23.

<A configuration example of a control system with a built-in abnormality detection function>
FIG. 2 is a block diagram showing an example (current input) of a part that realizes an abnormality detection function of a control system incorporating an abnormality detection function using the AI model according to the first embodiment. As shown in FIG. 2, the control system 1 of this example includes a data input unit 15 and an abnormality detection unit 12. The control signal (current signal) input to the motor 3 via the motor driver 2 is detected by the current sensor 6, and the detected current signal i is input to the data input unit 15 of the control system 1. Since this current signal i is used for feedback control of the motor 3, it is also referred to as a feedback phase current i.

Here, the abnormality detection unit 12 includes a signal preprocessing unit 121, a prediction model unit 122, and an abnormality determination unit 123. Further, the signal preprocessing unit 121 includes a window function unit 1211 and an FFT function unit 1212. The signal preprocessing unit 121 realizes the FFT method of the target frequency spectrum based on the preprocessing data 21 stored in the memory 20 as preprocessing for abnormality detection. As the preprocessing method, for example, the method described in Patent Document 1 may be adopted, and in the implementation of the firmware, a part thereof may be realized as a new firmware function.

The data input unit 15 outputs the current signal i detected by the current sensor 6 to the window function unit 1211 of the signal preprocessing unit 121. This current signal i is represented in the frequency spectrum format. The window function unit 1211 of the signal preprocessing unit 121 executes window function processing in which a frequency spectrum other than a specific section is set to 0, and outputs the obtained frequency spectrum to the FFT function unit 1212.

The FFT function unit 1212 performs FFT processing on this frequency spectrum, and outputs the frequency spectrum after the FFT processing to the prediction model unit 122. The frequency spectrum analysis of the current signal i after the FFT processing is used by the abnormality determination unit 123.

The prediction model unit 122 predicts an abnormality by AI based on the prediction model data 22 stored in the memory 20 and the frequency spectrum after the FFT processing. For example, when anomaly detection is performed using deep learning, which is a type of AI, the network structure, weight, bias value, etc. of the neural network are generated. In deep learning, while sequentially inputting the current signal i to be detected into the neural network, the error between the predicted value and the expected value by the neural network is calculated as the loss value, and the loss value is set so that the loss value becomes small. The learning process that feeds back to the weight and bias may be repeated. The prediction model unit 122 outputs the predicted value of the current signal i by AI to the abnormality determination unit 123. The prediction model unit 122 calculates the predicted value by selecting only the frequency portion necessary for abnormality detection from the frequency spectrum after FFT processing output from the FFT function unit 1212 based on the signal from the CPU 11. I am trying to be able to do it.

The abnormality determination unit 123 determines whether or not there is an abnormality in the motor 3 based on the predicted value of the current signal i input from the prediction model unit 122 and the control signal output from the peripheral function unit 14 to the motor driver 2. ..

In the present control system 1 of this example, since the abnormality detection function is incorporated in the system for controlling the rotating machine, the CPU 11 can easily obtain various information such as the target rotation speed of the motor from the motor control unit 13 and the peripheral function unit 14. Can be obtained in. Further, the prediction model unit 122 can calculate the predicted value by selecting only the frequency portion necessary for abnormality detection from the frequency spectrum after FFT processing output from the FFT function unit 1212 based on the signal from the CPU 11. I am able to do it. As described above, when anomaly detection is performed using deep learning, which is a type of AI, the amount of calculation of the prediction model greatly depends on the number of frequency spectra used in the calculation, so the frequency required for anomaly detection. By selecting only a part, the number of frequency spectra can be reduced, which can reduce the amount of calculation required for the entire system. For example, the rotation frequency is acquired from the motor control unit 13 or the peripheral function unit 14, and the CPU 11 is programmed to instruct the prediction model unit that only the width of several Hz above and below the rotation frequency should be used for abnormality detection. This can be achieved by keeping it.

<Another configuration example of a control system with a built-in abnormality detection function>
Next, another example of the configuration of the control system 1 incorporating the abnormality detection function will be described.

FIG. 3 is a block diagram showing another example (target frequency spectrum) of the present control system using the AI model according to the first embodiment. In this example, the rotation speed signal ω is used in addition to the current signal i shown in the example of FIG. 2 as the input data to the control system 1 incorporating the abnormality detection function. The control system 1 of this example detects the rotation speed of the rotating machine based on the rotation speed signal ω, and automatically switches the frequency band (frequency range) of the frequency spectrum according to the rotation speed. It is a feature.

As shown in FIG. 3, the control system 1 of this example includes a data input unit 15, a signal preprocessing unit 121 as an abnormality detection unit 12, a prediction model unit 122, and an abnormality determination unit 123. The control signal (current signal) input to the motor 3 via the motor driver 2 is detected by the current sensor 6, and the detected current signal i is input to the data input unit 15 of the control system 1. Since this current signal i is used for feedback control of the motor 3, it is also referred to as a feedback phase current i. Further, the rotation speed (speed signal) ω of the rotor (not shown) of the motor 3 and the rotary machine rotationally driven by the motor 3 is detected by the position / speed sensor 5, and the detected rotation speed signal ω is also the rotation speed signal ω of the present control system 1. It is input to the data input unit 15. This rotation speed signal ω is used in the calculation of the target frequency spectrum by the target spectrum unit 1213 of the signal preprocessing unit 121.

The data input unit 15 outputs these current signals i and rotation speed signals ω to the motor control unit 13. The motor control unit 13 performs feedback control of the motor 3 based on the detected data i and ω and the motor control data 23 stored in the memory 20.

In this example, the signal preprocessing unit 121 includes a target spectrum unit 1213, a frequency shift unit 1214, a filter unit 1215, a downsampling unit 1216, a window function unit 1211, and an FFT function unit 1212.

The target spectrum unit 1213 calculates the range of the target frequency spectrum based on the rotation speed signal ω input from the data input unit 15. This velocity signal ω (also referred to as feedback velocity) is based on the cause of the abnormality and the influence of the current on the frequency spectrum. For example, the frequency analysis of the vibration caused by the wear or crevice of a gear (not shown) is reflected in the target frequency spectrum (frequency lower limit value f _bl to frequency upper limit value f _bu ). The calculation formula of the target frequency spectrum caused by the wear and tear of the gear is as follows.
_fr = ω / 2π
f _e = (pole pair) x _fr

Here, ω is a speed signal (rad / sec), _{fr is a motor rotation frequency (Hz), and fe is} a current frequency (Hz). Further, the range R of the target frequency spectrum is _fe −n × fr ≦ _R ≦ f _e + _n × fr using an integer n (0 ≦ n ≦ the number of poles / 2). That is, the lower limit of the frequency is f _bl = f _e −n × fr, and the upper limit of the frequency is f _bu = f _e + _n × _fr . Further, the motor rotation speed is given by the motor rotation frequency fr, the current frequency _fe , and the integer _n .

As described above, the target spectrum unit 1213 of the present embodiment automatically switches the range R of the target frequency spectrum to be FFT processed, which will be described later, based on the rotation speed signal ω detected by the position / speed sensor 5. It is a thing. Then, the target spectrum unit 1213 outputs the calculated target frequency spectrum to the frequency shift unit 1214.

The frequency shift unit 1214 shifts the lower limit cutoff point to frequency 0 by multiplying the time-series current signal by e ^{−j × 2π × fk × (1 / fs)} . Here, fk is the lower limit cutoff frequency (f _bl ), and fs is the sampling frequency. The frequency shift unit 1214 outputs the shifted target frequency spectrum to the filter unit 1215.

The filter unit 1215 is composed of, for example, an anti-aliasing filter which is a low-pass filter for preventing an aliasing (folding) error that may occur at the time of sampling. The filter unit 1215 removes frequency components having a frequency upper limit value f _bu or more of the target frequency spectrum after shifting input from the frequency shift unit 1214. The filter unit 1215 outputs the target frequency spectrum excluding the upper limit side to the downsampling unit 1216.

The downsampling unit 1216 increases the frequency resolution as shown below.

Frequency resolution Δf = sampling frequency fs / number of FFT bins N
Normally, the frequency resolution Δf increases in multiples of 2, such as 128, 256, 512, ... As the number of FFT bins increases. By limiting the maximum value of N, the FFT resolution can be increased by downsampling.

As described with reference to FIG. 2, the window function unit 1211 and the FFT function unit 1212 calculate the FFT. The FFT signal preprocessed by the signal preprocessing unit 121 is output to the prediction model unit 122. The prediction model unit 122 predicts an abnormality by AI based on the frequency spectrum after FFT processing, and the abnormality determination unit 123 determines an abnormality of the motor 3 based on a prediction value or the like input from the prediction model unit 122. Determine the presence or absence. As in the example of FIG. 2, the prediction model unit 122 selects only the frequency portion necessary for abnormality detection from the frequency spectrum after FFT processing output from the FFT function unit 1212 based on the signal from the CPU 11. It is possible to calculate the predicted value, which can reduce the amount of calculation of the predicted model.

<Operation of control system with built-in abnormality detection function>
Next, an example of the operation of the control system 1 incorporating the abnormality detection function will be described. FIG. 4 is a flowchart showing an example of an abnormality detection process executed by the control system shown in FIG. This abnormality detection process is executed at a predetermined timing while the motor 3 is being driven.

In the abnormality detection process, first, the current signal i detected by the current sensor 6 and the speed signal ω detected by the position / speed sensor 5 are input to the data input unit 15 (step S11). The data input unit 15 outputs these signals to the target spectrum unit 1213.

The target spectrum unit 1213 calculates a target frequency spectrum (target spectrum) based on the velocity signal ω input from the data input unit 15 (step S12), and outputs the calculated target frequency spectrum to the frequency shift unit 1214.

The frequency shift unit 1214 shifts the lower limit cutoff point to frequency 0 by a predetermined arithmetic process (step S13), and outputs the shifted target frequency spectrum to the filter unit 1215.

The filter unit 1215 removes a frequency equal to or higher than the frequency upper limit value f _bu of the shifted target frequency spectrum, that is, a high frequency component (step S14), and outputs the removed target frequency spectrum to the downsampling unit 1216.

The downsampling unit 1216 performs downsampling processing on the target frequency spectrum after removing the high frequency component, increases the frequency resolution by a predetermined arithmetic processing (step S15), and outputs the processed target frequency spectrum to the window function unit 1211. ..

The window function unit 1211 executes a window function process (window function) in which a frequency spectrum other than a specific section is set to 0 (step S16), and outputs the obtained frequency spectrum to the FFT function unit 1212. The FFT function unit 1212 performs FFT processing on this frequency spectrum (step S17), and outputs the frequency spectrum after the FFT processing to the prediction model unit 122.

The prediction model unit 122 predicts an abnormality by AI based on the frequency spectrum after FFT processing (step S18), and outputs the prediction result to the abnormality determination unit 123. The abnormality determination unit 123 determines the presence or absence of an abnormality in the motor 3 based on the prediction result of the prediction model unit 122 (step S19), and the control system 1 ends this abnormality detection process.

<Characteristics and effects of Embodiment 1>
Next, the main features and effects of the control system 1 incorporating the abnormality detection function according to the first embodiment will be described.

The main feature of the control system 1 incorporating the abnormality detection function according to the first embodiment shown in FIGS. 1 to 3 is that the abnormality detection function is integrated with the conventional motor control system. Further, a further feature of the control system 1 having an abnormality detection function is that the signal preprocessing unit 121 calculates the range of the target frequency spectrum according to the cause of the abnormality and also makes the target frequency spectrum after removing the high frequency component. It is configured to perform downsampling processing and increase the frequency resolution by a predetermined arithmetic processing. Further, the prediction model unit 122 detects an abnormality from the frequency spectrum after FFT processing output from the FFT function unit 1212 based on the signal from the CPU 11 calculated using the signal easily acquired from the inside of the control system 1. It is possible to select only the frequency part required for the calculation and calculate the predicted value, thereby reducing the amount of calculation of the predicted model.

Since the control system 1 with a built-in abnormality detection function has the above configuration, necessary information is acquired from the control signal (control command) of the motor that is moving the rotating machine, and the rotation speed of the rotating machine ( According to the rotation speed signal ω) of the motor 3, the frequency spectrum range R to be confirmed / determined when determining the presence / absence of an abnormality based on one cause is automatically switched according to the cause. Then, by performing the FFT process only on the range R of the frequency spectrum to be confirmed / determined according to the rotation speed of the rotating machine, the FFT bin number N can be reduced, and the prediction model unit 122 is from the CPU 11. By further possessing a function of reducing the number of frequency spectra only for the frequency spectra required for abnormality detection based on the signal, it is possible to reduce the memory usage as a whole and reduce the processing load.

(Embodiment 2)
Next, the second embodiment will be described. In the following, the same reference numerals will be given to the parts having the same functions as those in the first embodiment, and the description thereof will be omitted in principle. In the first embodiment, one abnormality detecting unit 12 detects the presence or absence of an abnormality based on one cause. In this embodiment, a control system incorporating an abnormality detection function capable of detecting the presence or absence of an abnormality based on a plurality of causes (here, two) will be described.

<Configuration of control system with built-in abnormality detection function>
FIG. 5 is a configuration diagram showing an example of a control system 1'in which the abnormality detection function according to the second embodiment is incorporated. As shown in FIG. 5, the control system 1'is a microcomputer for controlling the entire system including the motor 3 and detecting an abnormality of a rotating machine, similarly to the control system 1 of the first embodiment. The MCU) 10 and a memory 20 for storing various applications and data necessary for motor control and abnormality detection are provided.

As will be described later with reference to FIG. 6, in the present embodiment, in order to detect an abnormality based on two causes, the memory 20 has the first preprocessing data 21a and the second preprocessing data 21a. 21b, the first prediction model data 22a, the second prediction model data 22b, and the motor control data 23 are stored.

The abnormality detection unit 12 includes two systems, but a specific configuration will be described with reference to the block diagram of FIG. FIG. 6 is a block diagram showing an example (target frequency spectrum) of the present control system using the AI model according to the second embodiment.

As shown in FIG. 6, in the control system 1'in which the abnormality detection function according to the present embodiment is incorporated, the abnormality detection unit 12 includes two signal preprocessing units 121a and 121b and two prediction model units 122a. It includes 122b and two abnormality determination units 123a and 123b.

The first signal preprocessing unit 121a includes a first target spectrum unit 1213a, a first frequency shift unit 1214a, a first filter unit 1215a, a first downsampling unit 1216a, and a first window function. A unit 1211a and a first FFT function unit 1212a are included. Further, the second signal preprocessing unit 121b includes a second target spectrum unit 1213b, a second frequency shift unit 1214b, a second filter unit 1215b, a second downsampling unit 1216b, and a second. It includes a window function unit 1211b and a second FFT function unit 1212b.

Similar to the first embodiment, the rotation speed (speed signal) ω and the current signal i of the rotating machine are input to the data input unit 15. The data input unit 15 outputs the rotation speed signal ω to the first and second signal preprocessing units 121a and 121b.

Here, the first and second signal preprocessing units 121a and 121b perform signal preprocessing for detecting two abnormalities based on two causes. Therefore, even if the same rotation speed signal ω detected by the position / speed sensor 5 is input to the first and second target spectrum units 1213a and 1213b via the data input unit 15, the first preprocessing data. Since the contents of the second preprocessing data 21b are different from those of the 21a, the two target spectrum units 1213a and 1213b will calculate the ranges of the two target frequency spectra corresponding to the two different causes, respectively. Two such different causes include, for example, wear of a gear (not shown) and damage to an inverter (not shown).

The processing of each part after the two target spectrum parts 1213a and 1213b is substantially the same as the processing of the corresponding parts of the control system 1 according to the first embodiment. Therefore, in order to avoid duplicate explanations, detailed explanations thereof will be omitted here.

<Operation of control system with built-in abnormality detection function>
Next, the operation of the control system 1'in which the abnormality detection function is incorporated will be described. FIG. 7 is a flowchart showing an example of the abnormality detection process executed by the control system 1'shown in FIG. This abnormality detection process is executed at a predetermined timing while the motor 3 is being driven.

In the abnormality detection process, first, the current signal i detected by the current sensor 6 and the speed signal ω detected by the position / speed sensor 5 are input to the data input unit 15 (step S101). The data input unit 15 outputs these signals to the first and second target spectrum units 1213a and 1213b.

The first target spectrum unit 1213a calculates a target frequency spectrum (target spectrum) based on the velocity signal ω input from the data input unit 15 (step S102), and shifts the calculated target frequency spectrum to the first frequency. Output to unit 1214a.

The first frequency shift unit 1214a shifts the lower limit cutoff point to frequency 0 (step S103) by a predetermined arithmetic process, and outputs the shifted target frequency spectrum to the first filter unit 1215a.

The first filter unit 1215a removes a frequency equal to or higher than the frequency upper limit value f _bu of the target frequency spectrum after shifting, that is, a high frequency component (step S104), and removes the removed target frequency spectrum from the first downsampling unit 1216a. Output to.

The first downsampling unit 1216a performs downsampling processing on the target frequency spectrum after removing the high frequency component, increases the frequency resolution by a predetermined arithmetic processing (step S105), and displays the processed target frequency spectrum in the first window. Output to the function unit 1211a.

The first window function unit 1211a executes a window function process (window function) in which a frequency spectrum other than a specific section is set to 0 (step S106), and outputs the obtained frequency spectrum to the first FFT function unit 1212a. .. The first FFT function unit 1212a performs FFT processing on this frequency spectrum (step S107), and outputs the frequency spectrum after the FFT processing to the first prediction model unit 122a.

The first prediction model unit 122a predicts an abnormality by AI based on the frequency spectrum after FFT processing (step S108), and outputs the prediction result to the first abnormality determination unit 123a. The first abnormality determination unit 123a determines whether or not there is an abnormality in the motor 3 based on the first cause based on the prediction result of the first prediction model unit 122a (step S109), and the control system 1'is , Ends this abnormality detection process. As in the example of FIG. 2, the first prediction model unit 122a has a frequency required for abnormality detection from the frequency spectrum after FFT processing output from the first FFT function unit 1212a based on the signal from the CPU 11. It is possible to select only a part and calculate the predicted value, thereby reducing the amount of calculation of the predicted model.

On the other hand, when the current signal i and the speed signal ω are input from the data input unit 15, the second signal preprocessing unit 121b, the second prediction model unit 122b, and the second abnormality determination unit 123b are in steps S102 to The process of steps S110 to S117, which is substantially the same as that of S109, is executed, and the control system 1'ends this abnormality detection process. As in the example of FIG. 2, the second prediction model unit 122b has a frequency required for abnormality detection from the frequency spectrum after FFT processing output from the second FFT function unit 1212b based on the signal from the CPU 11. It is possible to select only a part and calculate the predicted value, thereby reducing the amount of calculation of the predicted model.

As shown in the flowchart of FIG. 7, the abnormality detection processes of these two systems based on the two causes may be executed in parallel, and although not shown, the abnormality detection processes of the two systems are the abnormalities of one system. After the detection / determination is completed, it may be sequentially executed so as to execute the abnormality detection / determination of another system.

<Characteristics and effects of Embodiment 2>
Next, the main features and effects of the control system 1'in which the abnormality detection function according to the second embodiment is incorporated will be described.

As shown in FIG. 6, the feature of the control system 1'in which the abnormality detection function according to the present embodiment is incorporated is that each part of two systems is provided in the abnormality detection unit 12, and two parts are based on two different causes. The purpose is to detect and determine anomalies in parallel or in sequence.

Since the control system 1'has the above configuration, it is possible to detect anomalies with a high resolution spectrum corresponding to two target frequency spectra in order to detect anomalies based on two different causes. can. Further, since the abnormality detection processing of two systems is performed in order to detect the abnormality based on two different causes, the reliability of the control system 1'with respect to the motor control system can be improved.

Further, with such a configuration, it is possible to detect an abnormality of a rotating machine in real time with a higher frequency resolution while reducing the cost required for memory resources and calculation resources.

(Embodiment 3)
Next, the third embodiment will be described. In the following, the same reference numerals will be given to the parts having the same functions as those of the first embodiment or the second embodiment, and the description thereof will be omitted in principle. In the second embodiment, the abnormality detection unit 12 is composed of two systems, and the control system 1'detects an abnormality based on two different causes. In this embodiment, the present control system to which a third system using an environmental signal is added will be described.

<Configuration of control system with built-in abnormality detection function>
FIG. 8 is a block diagram showing an example (target frequency spectrum) of the control system 1 ”incorporated with the abnormality detection function using the AI model according to the third embodiment. As shown in FIG. 8, the present embodiment is shown. The control system 1 "in which the abnormality detection function according to the embodiment is incorporated includes three signal preprocessing units 121a, 121b, 121c, three prediction model units 122a, 122b, 122c, and three abnormality determination units 123a, 123b. It includes 123c, a meta prediction model unit 124, and a meta abnormality determination unit 125. The configuration of the two systems of abnormality detection units in which the detection data is input from the first data input unit 15a is the same as the configuration of the abnormality detection unit 12 (see FIGS. 5 and 6) of the second embodiment. Therefore, the detailed description thereof will be omitted here.

The rotation speed signal ω detected by the position / speed sensor 5 and the environmental signal of the motor 3 detected by the environmental sensor (not shown) (for example, vibration, temperature, humidity, etc.) are input to the second data input unit 15b. .. The second data input unit 15b outputs these rotation speed signals ω and the environmental signal to the third signal preprocessing unit 121c.

The third signal preprocessing unit 121c, like the first and second signal preprocessing units 121a and 121b, has a third target spectrum unit 1213c, a third frequency shift unit 1214c, and a third filter unit. It includes a 1215c, a third downsampling unit 1216c, a third window function unit 1211c, and a third FFT function unit 1212c.

In this example, the third target spectrum unit 1213c calculates the range of the target frequency spectrum based on the rotation speed signal ω input from the second data input unit 15b. Further, the third frequency shift unit 1214c shifts the lower limit cutoff point to frequency 0 by multiplying the time-series current signal by e ^{−j × 2π × fk × (1 / fs)} . For each function, the above-described first embodiment may be referred to.

The third filter unit 1215c removes frequency components having a frequency upper limit value f _bu or more of the target frequency spectrum after shifting. The third downsampling unit 1216c increases the frequency resolution (FFT resolution).

The first to third prediction model units 122a, 122b, and 122c predict an abnormality by AI based on the frequency spectrum after FFT processing input from the corresponding FFT function units 1212a, 1212b, and 1212c, respectively. Then, the first to third abnormality determination units 123a, 123b, 123c perform the first to third signal preprocessing based on the prediction results of the corresponding first to third prediction model units 122a, 122b, 122c. It is determined whether or not there is an abnormality in the motor 3 based on the cause corresponding to the portions 121a, 121b, and 121c.

The determination results of the first to third abnormality determination units 123a, 123b, 123c are input to the meta prediction model unit 124. The meta prediction model unit 124 predicts an abnormality of the entire motor control system based on these determination results, and outputs the prediction result to the meta abnormality determination unit 125. The meta abnormality determination unit 125 determines whether or not there is an abnormality in the entire motor control system.

<Operation of control system with built-in abnormality detection function>
Next, an example of the operation of the control system 1 "in which the abnormality detection function is incorporated will be described. FIG. 9 is a flowchart showing an example of the abnormality detection process executed by the control system 1" shown in FIG. The process of steps S201 to S217 in FIG. 9 is substantially the same as the process of steps S101 to S117 of FIG. 7 in the second embodiment. Therefore, in order to avoid duplicate explanations, detailed explanations thereof will be omitted here.

When the presence or absence of an abnormality in the motor 3 based on the first cause is detected in step S209, the processing flow shifts to step S227. Further, when the presence or absence of the abnormality of the motor 3 based on the second cause is detected in step S217, the processing flow similarly shifts to step S227.

Then, the abnormality detection process relating to the third cause is executed in parallel with the processes of steps S201 to S217. First, the speed signal ω detected by the position / speed sensor 5 and the environmental signal of the motor 3 (for example, vibration, temperature, humidity, etc.) detected by the environmental sensor (not shown) are input to the second data input unit 15b. (Step S218).

The third target spectrum unit 1213c calculates the target frequency spectrum (target spectrum) based on the velocity signal ω input from the second data input unit 15b (step S219), and the third frequency shift unit 1214c The lower limit cutoff point is shifted to frequency 0 by a predetermined arithmetic process (step S220), and the third filter unit 1215c has a frequency equal to or higher than the frequency upper limit value f _bu of the shifted target frequency spectrum, that is, a high frequency component. Is removed (step S221).

Next, the third downsampling unit 1216c performs a downsampling process on the target frequency spectrum after removing the high frequency component, increases the frequency resolution by a predetermined arithmetic process (step S222), and the third window function unit 1211c sets the third window function unit 1211c. A window function process (window function) in which a frequency spectrum other than a specific section is set to 0 is executed (step S223), and the third FFT function unit 1212c performs an FFT process on this frequency spectrum (step S224).

Next, the third prediction model unit 122c predicts the abnormality by AI based on the frequency spectrum after the FFT process (step S225), and the third abnormality determination unit 123c is the third prediction model unit 122c. Based on the prediction result, it is determined whether or not there is an abnormality in the motor 3 based on the third cause (step S226). As in the example of FIG. 2, the third prediction model unit 122c has a frequency required for abnormality detection from the frequency spectrum after FFT processing output from the third FFT function unit 1212c based on the signal from the CPU 11. It is possible to select only a part and calculate the predicted value, thereby reducing the amount of calculation of the predicted model.

Then, the meta prediction model unit 125 predicts the abnormality of the entire motor control system based on the determination results of the first to third abnormality determination units 123a, 123b, 123c (step S227), and the meta abnormality determination unit 125. Determines an abnormality in the entire motor control system (step S228), and the control system 1 "ends this abnormality detection process.

<Characteristics and effects of Embodiment 3>
Next, the main features and effects of the control system 1 "incorporating the abnormality detection function according to the third embodiment will be described.

The feature of the control system 1 "in which the abnormality detection function according to the present embodiment is incorporated is based on the environmental signal (environmental parameter signal) in addition to the configuration of the control system 1'of the second embodiment shown in FIG. The detection and determination of abnormalities based on the third cause are executed in parallel or in sequence. A further feature is the determination of the presence or absence of three or more types based on the first to third causes. Based on this, it is to detect and judge the abnormality of the whole system.

Since the control system 1 "has the above configuration, the effectiveness of the detection method for a specific cause of abnormality depends on the signal used for detection, but abnormality detection using both the control signal and the environmental signal. Thereby, the reliability of the normal / abnormal determination of the control system 1 "of the present embodiment can be improved. Further, in order to detect anomalies based on three different causes, it is possible to detect anomalies with a high resolution spectrum corresponding to three target frequency spectra. Further, since the abnormality detection processing of three systems is performed in order to detect the abnormality based on the three different causes, the reliability of the control system 1 "with respect to the motor control system can be further improved.

Further, with such a configuration, as in the case of the second embodiment, it is possible to detect an abnormality of the rotating machine in real time with a higher frequency resolution while reducing the cost required for the memory resource and the calculation resource.

Although the invention made by the present inventor has been specifically described above based on the embodiments thereof, the present invention is not limited to the above-described embodiments 1 to 3, and varies within the range not deviating from the gist thereof. Needless to say, it can be changed.

For example, in the first to third embodiments, each part of the abnormality detection unit 12 is provided in the CPU 11, and the case where the abnormality of the motor 3 is detected and determined as software processing has been described. However, the present invention has such a configuration. Not limited to. To the extent that the cost is acceptable, for example, each of these parts may be configured with dedicated hardware.

Further, in the first to third embodiments, the case where the abnormality detection processing of 1 to 3 systems is performed respectively has been described, but the present invention is not limited to such a configuration. For example, the control system incorporating the abnormality detection function may be configured to perform abnormality detection processing of four or more systems according to the target system or device for abnormality detection.

1, 1', 1 "Control system with built-in abnormality detection function 4 Environment sensor 5 Position / speed sensor 6 Current sensor 12 Abnormality detection unit 121, 121a, 121b, 121c Signal preprocessing unit 1213, 1213a, 1213b, 1213c Target Spectrum section 1216, 1216a, 1216b, 1216c Downsampling section 122, 122a, 122b, 122c Prediction model section 123, 123a, 123b, 123c Abnormality determination section

Claims (9)

A control system with a built-in abnormality detection function that detects machine abnormalities.
A signal preprocessing unit that performs FFT processing on the target frequency spectrum of the control signal according to the cause of the abnormality of the machine using the control signal to the machine.
Based on the frequency spectrum after FFT processing by the signal preprocessing unit, or by using the signal inside the control system, the part necessary for predicting the abnormality by AI is extracted from the frequency spectrum after FFT processing. Prediction model section that predicts abnormalities by AI,
An abnormality determination unit that determines the presence or absence of an abnormality in the machine based on the predicted value of the prediction model unit.
Equipped with
The signal preprocessing unit
A target spectrum unit that calculates a range of the target frequency spectrum according to the cause of the abnormality based on a control signal to the machine or a detection signal of a sensor that detects the environment of the machine.
A filter unit that removes frequency components above the upper limit of the target frequency spectrum calculated by the target spectrum unit, and a filter unit.
The FFT function unit that performs the FFT processing on the target frequency spectrum from which the high frequency band has been removed by the filter unit, and the FFT function unit.
including,
A control system with a built-in anomaly detection function. In the control system incorporating the abnormality detection function according to claim 1,
The signal preprocessing unit further includes a downsampling unit that increases the FFT resolution by downsampling with respect to the target frequency spectrum from which the frequency components above the upper limit value have been removed by the filter unit.
A control system with a built-in anomaly detection function. In the control system incorporating the abnormality detection function according to claim 1,
The anomaly detection system is integrated with a control system for controlling the machine.
A control system with a built-in anomaly detection function. In the control system incorporating the abnormality detection function according to claim 1,
A plurality of the signal preprocessing units according to the causes of the plurality of abnormalities of the machine, and
The plurality of prediction model units corresponding to the plurality of signal preprocessing units, and
A plurality of the abnormality detection units corresponding to the plurality of prediction model units, and
Equipped with
The plurality of target spectrum units included in the plurality of signal preprocessing units each calculate a range of a plurality of different target frequency spectra according to the causes of the plurality of abnormalities.
The plurality of prediction model units predict anomalies by AI based on the frequency spectra after FFT processing by the corresponding signal preprocessing units, respectively.
The plurality of abnormality detection units detect the presence or absence of a plurality of abnormalities in the machine based on the predicted values of the corresponding prediction model units.
A control system with a built-in anomaly detection function. In the control system incorporating the abnormality detection function according to claim 1,
A second signal preprocessing unit that performs FFT processing on the target frequency spectrum of the control signal according to the cause of the abnormality of the machine using the control signal to the machine and the environmental signal of the machine.
A second prediction model unit that predicts anomalies by AI based on the frequency spectrum after FFT processing by the second signal preprocessing unit, and
Based on the predicted value of the second prediction model unit, the second abnormality detection unit that detects the presence or absence of an abnormality in the machine, and the second abnormality detection unit.
Further prepare,
A control system with a built-in anomaly detection function. In the control system incorporating the abnormality detection function according to claim 1,
The machine includes a rotating machine that rotates by itself and a power machine that operates a machine to be rotated.
A control system with a built-in anomaly detection function. It is a machine abnormality detection method that detects machine abnormalities.
Signal preprocessing that applies FFT processing to the target frequency spectrum of the control signal according to the cause of the abnormality of the machine using the control signal to the machine.
Predictive model processing that predicts anomalies by AI based on the frequency spectrum after FFT processing, and
Anomaly detection processing that detects the presence or absence of an abnormality in the machine based on the predicted value,
Including
In the signal preprocessing,
Based on the control signal to the machine or the detection signal of the sensor that detects the environment of the machine, the range of the target frequency spectrum according to the cause of the abnormality is calculated.
The frequency components above the upper limit of the calculated target frequency spectrum are removed.
The FFT process is applied to the target frequency spectrum from which the high frequency band has been removed.
Machine abnormality detection method. In the method for detecting an abnormality in a machine according to claim 7,
In the signal preprocessing, further
The FFT resolution is increased by downsampling with respect to the target frequency spectrum from which the frequency components above the upper limit are removed.
Machine abnormality detection method. In the method for detecting an abnormality in a machine according to claim 7,
The machine includes a rotating machine that rotates by itself and a power machine that operates a machine to be rotated.
Machine abnormality detection method.

Images