Abnormality diagnosis device, abnormality diagnosis method, and abnormality diagnosis program
By converting the vibration or action sound of the electrical equipment into a digital signal and performing short-term fast Fourier transformation, comparing the feature quantity with normal data, the problem of difficult to distinguish different abnormal types and unknown abnormalities in the prior art is solved, and the problem of realizing the Accurate diagnosis of electrical equipment abnormalities.
Patent Information
- Application Number
- CN202380031547.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Priority Date
- 2022-04-07
- Filing Date
- 2023-03-03
- Publication Date
- 2024-11-15
AI Technical Summary
The prior art is difficult to distinguish different anomalies types and unknown anomalies based on spectral patterns alone, and sensory tests rely on the feeling of the operator, making it difficult to quantitatively distinguish vibration or action sound abnormalities of electrical equipment.
The microphone is used to convert the sound into an analog electrical signal, and the signal converter is converted into a digital signal. The signal processing device is used to calculate the characteristic quantity for short-term fast Fourier transformation, and compare it with the stored normal waveform data characteristic quantity, and pass the discriminant index. Value determines the quality of the waveform data.
Accurate judgment of known and unknown abnormalities is achieved, the accuracy and reliability of abnormal diagnosis of electrical equipment is improved, and the dependence on operators' perception judgment is reduced.
Smart Images

Figure CN118974532A_ABST
Abstract
Description
Technical Field
[0001] The present disclosure relates to an abnormality diagnosis device, an abnormality diagnosis method, and an abnormality diagnosis program for diagnosing abnormality of an electrical device. Background Art
[0002] In the assembly process of general electrical equipment, an operation test of the electrical equipment is performed during or after the assembly. In addition, during the operation test, a sensory test is performed to confirm whether the vibration or operation sound generated by the electrical equipment is normal. Such a sensory test is performed by the hearing or touch of the operator, and therefore has the characteristic of relying on the operator's sense. Therefore, in the sensory test, in order to quantify the information of the vibration or operation sound as the material for judging the quality, a diagnostic device is sometimes used that converts the information of the vibration or operation sound obtained by a microphone, a vibration sensor, etc. into waveform data and performs signal processing.
[0003] Patent Document 1 below discloses a technique for comparing a spectrum pattern obtained by performing Fast Fourier Transform (FFT) processing on waveform data of vibration generated by an electrical device with a preset spectrum pattern of a known abnormality, and diagnosing the presence or absence of an abnormality in the electrical device based on the comparison result.
[0004] Prior art literature
[0005] Patent Literature
[0006] Patent Document 1: Japanese Patent Application Laid-Open No. 62-93620 Summary of the invention
[0007] Problems to be solved by the invention
[0008] As described above, the technology of the above-mentioned Patent Document 1 uses only the spectrum pattern to determine the abnormality. Therefore, the technology of Patent Document 1 has the following problem: when the spectrum patterns are equal for different types of abnormalities, for example, for abnormalities in which the spectrum patterns are equal but the time intervals of occurrence are different, it is difficult to distinguish the type of abnormality. In addition, the technology of the above-mentioned Patent Document 1 has the following problem: although it can distinguish known abnormalities such as the spectrum pattern changing over time relative to the waveform data, it is difficult to distinguish unknown abnormalities that are different from the usual ones that the operator feels.
[0009] The present disclosure has been made in view of the above-mentioned situation, and an object of the present disclosure is to obtain an abnormality diagnosis device that can distinguish known and unknown abnormalities that are different from normal ones.
[0010] Means for solving problems
[0011] In order to solve the above-mentioned problems and achieve the purpose, the abnormality diagnosis device disclosed in the present invention comprises: a microphone, which converts the sound of the object to be judged into an analog electric signal; and a signal converter, which converts the analog electric signal into a digital signal. The signal processing device that takes in the digital signal for signal processing comprises a signal processing unit, a data storage unit and a judgment unit. The signal processing unit performs a short-time fast Fourier transform on the waveform data of the input signal to calculate the feature quantity. The data storage unit stores feature quantity data based on normal known waveform data. The judgment unit compares the first feature quantity data composed of a plurality of feature quantities calculated by the signal processing unit with the feature quantity data stored in the data storage unit, that is, the second feature quantity data, to judge the quality of the waveform data of the input signal. The feature quantity is a quantity that represents the degree of deviation of the spectral intensity variation caused by time in a specific frequency band of the waveform data.
[0012] Effects of the Invention
[0013] According to the abnormality diagnosis device of the present disclosure, it is possible to distinguish known and unknown abnormalities that are different from normal ones. BRIEF DESCRIPTION OF THE DRAWINGS
[0014] Figure 1 This is a block diagram showing a configuration example of the abnormality diagnosis device according to the first embodiment.
[0015] Figure 2 This is a diagram for explaining filter bank processing that can be applied to the signal processing device of Embodiment 1.
[0016] Figure 3 This is a flowchart for explaining the abnormality diagnosis method of the electric equipment performed by using the abnormality diagnosis device according to the first embodiment.
[0017] Figure 4 This is a block diagram showing a configuration example of an abnormality diagnosis device according to the second embodiment.
[0018] Figure 5 This is a block diagram for explaining the algorithm of the estimation process used in the second embodiment.
[0019] Figure 6 This is a block diagram showing a configuration example of an abnormality diagnosis device according to the third embodiment.
[0020] Figure 7 This is a block diagram showing a configuration example of an abnormality diagnosis device according to a fourth embodiment.
[0021] Figure 8 This is a block diagram showing a configuration example of an abnormality diagnosis device according to the sixth embodiment.
[0022] Fig. 9 This is a diagram for explaining a known noise removal process performed using the abnormality diagnosis device according to the sixth embodiment. DETAILED DESCRIPTION
[0023] Hereinafter, an abnormality diagnosis device, an abnormality diagnosis method, and an abnormality diagnosis program according to an embodiment of the present disclosure will be described in detail with reference to the drawings. In addition, the embodiments described below are examples, and the scope of the present disclosure is not limited to the following embodiments.
[0024] Implementation method 1.
[0025] Figure 1 1 is a block diagram showing a configuration example of an abnormality diagnosis device according to Embodiment 1. The abnormality diagnosis device 100 according to Embodiment 1 includes a microphone 2 , a signal converter 3 , and a signal processing device 10 . The signal processing device 10 also includes a signal processing unit 4 , a data storage unit 5 , and a determination unit 6 .
[0026] The microphone 2 converts the sound including the operation sound of the identification object 1 into an analog electric signal. The signal converter 3 is an analog to digital (AD) converter, and converts the analog electric signal output from the microphone 2 into a digital signal. The signal processing device 10 takes in the digital signal output from the signal converter 3 and performs the signal processing described later. In addition, the abnormality diagnosis device 100 may also include a plurality of signal processing devices 10.
[0027] In the signal processing device 10, the signal processing unit 4 performs necessary arithmetic processing on the waveform data of the input signal to calculate the feature quantity. The feature quantity is a quantity indicating the degree of deviation of the variation of the spectrum intensity of a specific frequency band possessed by the waveform data due to time, in other words, it is a quantity indicating how much the variation of the spectrum intensity of the specific frequency band deviates with respect to the direction of passage of time.
[0028] The feature quantity data based on normal known waveform data is stored in the data storage unit 5. The determination unit 6 compares the feature quantity data composed of a plurality of feature quantities calculated by the signal processing unit 4 with the feature quantity data based on normal known waveform data stored in the data storage unit 5, and determines whether the waveform data of the input signal input to the signal processing device 10 is good or bad.
[0029] In addition, in this specification, a data group consisting of multiple feature quantities calculated using the signal processing unit 4 is sometimes referred to as "first feature quantity data", and feature quantity data based on normal known waveform data or a collection of feature quantity data stored in the data storage unit 5 is referred to as "second feature quantity data".
[0030] The microphone 2 can be a general-purpose microphone, but it is preferable that the operating sound generated by the identification object 1 does not include surrounding noise as much as possible. This can be achieved by using a sound pickup microphone with sound pickup directivity, a parabolic sound pickup microphone formed by combining a general microphone and a parabolic reflector, or an appropriate combination of these sound pickup microphones.
[0031] The signal converter 3 can use a general audio interface. Here, if an audio interface with a large sampling frequency is used, the range in the frequency direction can be set larger when performing FFT processing in the signal processing described later, so it is preferred. In addition, when the signal processing device 10 has a function equivalent to the signal converter 3, this function can also be used.
[0032] The processing of the signal processing device 10 can be performed using a general computer or a programmable logic controller (PLC). PLC is also called a sequencer. In addition, the processing of the signal processing device 10 can also use an electronic substrate composed of a signal processing device and a storage medium. Examples of signal processing devices are microcomputers, FPGAs (Field Programmable Gate Arrays), and examples of storage media are non-volatile or volatile semiconductor memories such as RAMs (Random Access Memory), flash memory, EPROMs (Erasable Programmable ROMs), and EEPROMs (registered trademarks) (Electrically EPROMs). The signal processing device 10 can perform a series of operations from the input of waveform data to the determination according to program-type processing. In addition, the signal processing device 10 can also have a screen that can indicate and display the state of the determination. In this case, the screen can be used to make the settings required for the determination process, and the state of the determination and its results can be displayed on the screen to notify the operator.
[0033] Next, the algorithm of the discrimination processing in the signal processing device 10 is described. The signal processing unit 4 performs a short-time fast Fourier transform (STFT) on the waveform data converted by the signal converter 3, dividing the waveform data into an arbitrary number in the frequency direction and the time direction. In this specification, the processing result of the STFT is represented by a matrix F having i frequencies and j times as shown in the following formula (1).
[0034] [Formula 1]
[0035]
[0036] In the above formula (1), i is the number of divisions of the waveform data in the frequency direction, and j is the number of divisions of the waveform data in the time direction.
[0037] Furthermore, the signal processing unit 4 calculates a feature quantity vector T represented by the following equation (2) using the standard deviation of the spectrum intensity in the time direction in the matrix F as a feature quantity, based on the matrix F which is the result of the STFT process.
[0038] [Formula 2]
[0039]
[0040] In the above formula (2), each element t1...t i In the equation (3), the feature quantity t at any frequency i is calculated by the following formula (3).
[0041] [Formula 3]
[0042]
[0043] In the above formula (3), “f - " is a substitute mark with a "-" sign on top of a character such as "f". - l " is the j elements f in the i-th row of the above formula (1) i,1 …f i,j The average value of .
[0044] As described above, feature quantity data based on normal known waveform data is stored in the data storage unit 5. The feature quantity data stored in the data storage unit 5 is also calculated by STFT. Here, the number of normal known waveform data is assumed to be k, and T' is used. k represents a collection of feature quantities based on a plurality of normal known waveform data. T'1 is a feature quantity vector obtained using waveform data 1 as the first waveform data, and T' k is a feature vector obtained using waveform data k as the kth waveform data. In addition, in this specification, data related to the average ν and standard deviation σ of the collection T' is used as the second feature data. The average ν and standard deviation σ can be expressed by the following equations (4) and (5). At this time, the number k of normal known waveform data is preferably at least 100.
[0045] [Formula 4]
[0046]
[0047] [Formula 5]
[0048]
[0049] In addition, in this specification, the average ν and the standard deviation σ shown in the above equations (4) and (5) are examples of the second feature value data, and statistics other than the average ν and the standard deviation σ can also be stored in the data storage unit 5 as the second feature value data.
[0050] As described above, the determination unit 6 determines whether the waveform data of the input signal is good or bad. In this determination process, the determination index value Z shown in the following formula (6) is used. i .
[0051] [Formula 6]
[0052]
[0053] Discrimination index value Z i The determination unit 6 determines the deviation between the first feature quantity data and the second feature quantity data. i The value of is compared with the predetermined threshold value, and at least one or part of the discrimination index value Z i When the threshold value is exceeded, it is determined that the waveform data to be determined is abnormal or out of the normal state.
[0054] In addition, in the above STFT processing, it is also possible to use Figure 2 The filter bank processing shown. Figure 2 This is a diagram for explaining filter bank processing that can be applied to the signal processing device of Embodiment 1.
[0055] In the filter bank processing, with respect to the matrix F as the processing result of the STFT, the number of divisions of the waveform data in the frequency direction is set to i' (i'<i), and the number of divisions of the waveform data in the time direction is set to j' (j'<j). Specifically, Figure 2 , an example where i'=i / 2 and j'=j / 2 is shown.
[0056] exist Figure 2 In the left figure, for the top four elements f of the matrix F 1,1 、f 1,2 、f 2,1 、f 2,2 Conduct 1,1 +f 1,2 +f 2,1 +f 2,2 The operation is the addition operation, let it be f' 1,1 . Similarly, for the top four elements f of the matrix F 1,j-1 、f 1,j 、f 2,j-1、f 2,j Conduct 1,j-1 +f 1,j +f 2,j-1 +f 2,j operation, and set it to f' 1,j’ , for the four lower left elements f of the matrix F i-1,1 、f i-1,2 、f i,1 、f i,2 Conduct i-1,1 +f i-1,2 +f i,1 +f i,2 operation, and set it to f' i’,1 In addition, for the four elements f at the bottom right of the matrix F i-1,j-1 、f i-1,j 、f i,j-1 、f i,j Conduct i-1,j-1 +f i-1,j +f i,j-1 +f i,j operation, and set it to f' i’,j’ For the other elements of matrix F, we also perform addition operations every 4 elements. In this way, we generate Figure 2 The re-merged new matrix F' is shown in the right figure.
[0057] also, Figure 2 This is an example, and any i' and j' that satisfy the relationship of i'<i, j'<j can be selected. In addition, the division of the waveform data may not be at equal intervals. Therefore, the number of divisions i' does not necessarily have to be a divisor of the number of divisions i, and the number of divisions j' does not necessarily have to be a divisor of the number of divisions j. In addition, there is no dependency between the number of divisions in the frequency direction and the number of divisions in the time direction, and they can be performed independently.
[0058] By appropriately setting the division numbers i' and j', the sound features of the discrimination object 1 can be highlighted in the re-integrated matrix F'. Therefore, by using the re-integrated matrix F' through the filter bank processing, the accuracy of discrimination of good and bad waveform data can be improved.
[0059] Figure 3 This is a flowchart for explaining a method for diagnosing abnormality of electric equipment using the abnormality diagnosis device according to the first embodiment. Figure 3 The processing flow can be executed using a computer, a PLC, or a processor having a storage unit. Hereinafter, these computers, PLCs, or processors are simply referred to as "computers."
[0060] The computer stores second feature quantity data, that is, feature quantity data representing the degree of deviation of the spectral intensity variation caused by time in a specific frequency band of normal known waveform data. In addition, the computer stores a program for realizing the functions of the microphone 2, signal converter 3, and signal processing device 10. Alternatively, the computer is configured to be able to load the program from the outside and execute it.
[0061] The computer converts the operating sound of the electrical device into an analog electrical signal (step S11). The computer converts the analog electrical signal converted in step S11 into a digital signal (step S12). The computer performs STFT on the waveform data converted in step S12 to calculate the feature value (step S13).
[0062] The computer generates a discrimination index value Z based on the first feature quantity data composed of the plurality of feature quantities calculated in step S13 and the second feature quantity data stored in the computer. i Compare with the threshold value (step S14). Supplement the processing of step S14. As described above, the feature vector T representing the first feature data is represented by the above formula (2), and each element of the feature vector T is represented by the above formula (3). In addition, the discrimination index value Z i An example of is represented by the above formula (6).
[0063] The computer determines whether there is a discrimination index value Z that exceeds the threshold i (Step S15). i (Step S15, No), the computer determines that the electrical equipment is normal (Step S16). In addition, if there is at least one discrimination index value Z exceeding the threshold i In the case (step S15, yes), the computer determines that the electrical device is abnormal or there is a possibility of abnormality (step S17).
[0064] By using Figure 3 The flowchart can determine whether the electrical equipment to be determined is a normal product or an abnormal product. In addition, preferably, the computer has a function to display the determination index value Z on the display screen. i The function of displaying the threshold value in a graph can visually and easily inform the operator of the good and bad discrimination results.
[0065] As also explained in the section "Problems to be Solved by the Invention", in the prior art, only the spectrum pattern is used to determine the abnormality, so there is a problem that it is difficult to distinguish the type of abnormality for abnormalities such as the spectrum patterns being equal but the time intervals of occurrence being different. In addition, there is a problem in the prior art that it is difficult to distinguish unknown abnormalities that are different from the usual ones that the operator feels. In contrast, in the abnormality diagnosis method of embodiment 1, the presence or absence of abnormality in the waveform data to be determined is determined based on the deviation between the feature quantity of the waveform data to be determined and the feature quantity of the normal known waveform data stored in the computer, so that even unknown abnormalities can be determined to be good or bad.
[0066] In addition, in the above description, it is assumed that the feature quantity of normal known waveform data is stored in the computer, but it is not limited to this. The feature quantity of normal known waveform data can be referenced by the computer, and the feature quantity can be stored in a storage unit or storage area that is not a component of the computer. In other words, the computer can be configured to reference the feature quantity of normal known waveform data.
[0067] As described above, the abnormality diagnosis device of embodiment 1 includes: a microphone that converts the sound of the object to be judged into an analog electric signal; and a signal converter that converts the analog electric signal into a digital signal. The signal processing device that takes in the digital signal and performs signal processing includes a signal processing unit, a data storage unit, and a judgment unit. The signal processing unit performs a short-time fast Fourier transform on the waveform data of the input signal to calculate the feature quantity. The feature quantity data based on normal known waveform data is stored in the data storage unit. The judgment unit compares the first feature quantity data composed of a plurality of feature quantities calculated by the signal processing unit with the feature quantity data stored in the data storage unit, that is, the second feature quantity data, to judge the quality of the waveform data of the input signal. The feature quantity is a quantity that indicates the degree of deviation of the spectral intensity variation caused by time in a specific frequency band of the waveform data. According to the abnormality diagnosis device configured in this way, based on the deviation between the feature quantity of the waveform data as the object to be judged and the feature quantity of the normal known waveform data stored in the data storage unit, it is judged whether the waveform data as the object to be judged has an abnormality. As a result, known and unknown abnormalities that are different from the usual, such as the spectral pattern changing over time, can be judged.
[0068] In addition, in the above-mentioned processing, the signal processing device may perform filter bank processing on the feature quantity generated by short-time fast Fourier transform, and generate the first feature quantity data based on the feature quantity processed by the filter bank. In this way, the time width or frequency width when obtaining the first feature quantity data can be arbitrarily changed. Thus, the accuracy of distinguishing the quality of waveform data can be improved.
[0069] In addition, the abnormality diagnosis method of embodiment 1 is as follows: use a computer to diagnose the abnormality of the electrical device, and the computer is configured to be able to refer to a feature quantity, which represents the degree of deviation of the spectrum intensity change caused by time in a specific frequency band of normal known waveform data. The abnormality diagnosis method includes the first to fourth steps shown below. The first step is a step of converting the operation sound of the electrical device into an analog electrical signal. The second step is a step of converting the analog electrical signal into a digital signal. The third step is a step of performing a short-time fast Fourier transform on the waveform data converted by the second step to calculate the feature quantity. The fourth step is a step of comparing the first feature quantity data composed of a plurality of feature quantities calculated by the third step with the second feature quantity data stored in the computer to determine the quality of the waveform data. By making the computer execute these first to fourth steps, it is possible to distinguish known and unknown abnormalities that are different from the usual, such as the spectrum pattern changing over time.
[0070] In addition, the abnormality diagnosis program of embodiment 1 is a program that causes a computer to diagnose abnormalities of electrical equipment, and the computer is configured to be able to refer to a feature quantity, which represents the degree of deviation of the spectral intensity variation caused by time in a specific frequency band of normal known waveform data. The abnormality diagnosis program includes the first to fourth steps shown below. The first step is a step of converting the operating sound of the electrical equipment into an analog electrical signal. The second step is a step of converting the analog electrical signal into a digital signal. The third step is a step of performing a short-time fast Fourier transform on the waveform data converted by the second step to calculate the feature quantity. The fourth step is a step of comparing the first feature quantity data composed of a plurality of feature quantities calculated by the third step with the second feature quantity data stored in the computer to determine the quality of the waveform data. By causing the computer to execute the program including these first to fourth steps, it is possible to determine known and unknown abnormalities that are different from the usual, such as the spectral pattern changing over time.
[0071] Implementation method 2.
[0072] exist Figure 1 In the configuration of the abnormality diagnosis device of the first embodiment shown, there is a possibility that noise from the surroundings may be mixed in, resulting in erroneous judgment of noise and abnormal sound. In the second embodiment, a configuration that improves this is proposed.
[0073] Figure 4 : is a block diagram showing a configuration example of an abnormality diagnosis device according to Embodiment 2. Figure 4 In the abnormality diagnosis device 100A of the second embodiment, Figure 1 The abnormality diagnosis device 100 of the first embodiment shown in the figure has a noise collecting microphone 11 added thereto for collecting ambient noise. Figure 1The same or equivalent components are denoted by the same reference numerals, and repeated descriptions are omitted. Figure 4 In the embodiment, the number of the noise collecting microphones 11 is three, but the present invention is not limited to this example. The number of the noise collecting microphones 11 may be one, two, or four or more.
[0074] Next, the operation of the abnormality diagnosis device of the second embodiment is described. As described above, in the structure of the first embodiment, noise from the surroundings may be mixed in and the noise and abnormal sound may be misjudged. Therefore, in the second embodiment, the waveform data collected by the noise collecting microphone 11 is used to estimate the time interval in which the surrounding noise is mixed. Then, the method of the first embodiment is applied to the waveform data after excluding the estimated time interval. Thus, it is possible to provide a discrimination method that reduces the influence of the surrounding noise.
[0075] Next, refer to Figure 5 The algorithm of the estimation process used in the second embodiment will be described. Figure 5 FIG. 1 is a block diagram for explaining an algorithm for estimation processing used in Embodiment 2. Figure 5 In FIG. 2 , adders 21 and 23 and a filter 22 are shown as control elements for realizing the method of the second embodiment.
[0076] First, it is assumed that the matrix F shown in the above equation (1) and the feature vector T shown in the above equation (2) are obtained based on the waveform data of the sound signal collected by the microphone 2 and the noise collection microphone 11. At this time, at time t, s(t) represents the feature vector obtained from the microphone 2, and n1(t), n2(t), and n3(t) represent the feature vectors obtained from the three noise collection microphones 11, respectively.
[0077] exist Figure 5 In the above, the filter 22 can reproduce the attenuation of the values of each element of the matrix F. Here, w represents the constant vector representing the characteristics of the filter 22, and d represents ~ (t) represents the estimated value of the noise contained in the eigenvector s(t) observed at microphone 2. ~ " refers to a substitute notation in which a symbol "~" is added to the upper part of a character such as "d". In this case, the estimated value of noise d~(t) can be expressed by the following equation (7).
[0078] [Formula 7]
[0079]
[0080] In the above equation (7), the symbol "○" is an operator indicating the multiplication process of each element of the vector. In addition, the constant vector w representing the characteristics of the filter 22 is a vector whose elements are constants obtained by appropriately calculating the attenuation rate of each element. The attenuation rate of each element can be calculated using any method.
[0081] According to the above formula (7), the characteristic amount of noise is calculated by multiplying the value obtained by adding the sound pressure levels of the three noise collecting microphones 11 by a constant. Therefore, the period in which the characteristic amount of noise calculated by the above formula (7) exceeds a predetermined threshold value can be estimated as a period in which noise is mixed. Therefore, if the estimated period, i.e., the period in which noise is mixed is removed and the judgment is performed based on the above algorithm, the quality of the waveform data in which the influence of the mixed noise is reduced can be judged.
[0082] Next, the characteristics of the filter 22 are considered. First, it is considered that the characteristics of the filter 22 are good and equivalent to the actual noise d(t) observed by the microphone 2 and its estimated value d ~ In order to take this error into account, an observation signal matrix N represented by the following equation (8) is defined.
[0083] [Formula 8]
[0084]
[0085] In the above equation (8), the symbol "T" represents transposition. In addition, L represents the number of frames in the observation signal matrix N. In addition, the number of frames L corresponds to the number of short-time frames in STFT.
[0086] In addition, a noise matrix D represented by the following equation (9) which represents actual noise observed by the microphone 2 is defined.
[0087] [Formula 9]
[0088]
[0089] Here, the operation of taking out the i-th column of the observation signal matrix N shown in the above equation (8) to generate the longitudinal quantity is denoted as N i The operation of taking out the i-th column of the noise matrix D shown in the above formula (9) and generating the longitudinal quantity is marked as D i In addition, the operation of taking out the i-th value of the constant vector w is marked as w i At this time, in the matrix F representing the STFT processing result, when the square error when focusing on the i-th frequency is represented by e, the square error e can be represented as in the following equation (10).
[0090] [Formula 10]
[0091] e=(D i -w i N i ) T (D i -w i N i )…(10)
[0092] Find the w that minimizes the above equation (10) i The w i It is using w i The (de / w i ) becomes zero w i , that is, it satisfies de / w i =0 i , expressed by the following formula (11).
[0093] [Formula 11]
[0094]
[0095] If the above equation (11) is calculated for all i, the constant vector w with the optimal parameters can be obtained. In addition, in the calculation process described above, information on the noise d(t) collected by the microphone 2 and information on the characteristic vectors n1(t), n2(t), and n3(t) collected by the noise collection microphone 11 are required, but an alternative method can be used. An example of an alternative method is a method of using data recorded in a state where only noise exists and there is no movement sound of the measured object. In the case of this method, data obtained by recording only noise in an actual noise environment can be used, or data obtained by recording a signal equivalent to noise using a speaker in a laboratory can be used. As an example of the latter, a test signal that has energy in all frequency bands and changes with time can be used. As the test signal, the TSP signal used in the TSP method (Time-Stretched-Pulse method) can be exemplified.
[0096] As described above, the abnormality diagnosis device of embodiment 2 is further provided with a noise collecting microphone for collecting surrounding noise. The signal processing device estimates the time interval in which the surrounding noise is mixed using the waveform data collected using the noise collecting microphone, performs a short-time fast Fourier transform on the waveform data after excluding the estimated time interval, and calculates the feature quantity. Thus, it is possible to avoid misjudging the noise contained in the sound of the object to be judged as abnormal.
[0097] In addition, in the abnormality diagnosis method of the second embodiment, in the abnormality diagnosis method described in the first embodiment, an estimation step is included between the second step and the third step, in which the time interval mixed with the surrounding noise is estimated using the waveform data collected using the noise collection microphone for collecting the surrounding noise. Then, in the third step, the waveform data after excluding the time interval estimated in the estimation step is subjected to a short-time fast Fourier transform to calculate the feature quantity. In this way, a discrimination method that reduces the influence of the surrounding noise can be provided. As a result, it is possible to avoid misjudging the noise contained in the sound of the discrimination object as abnormal.
[0098] Implementation method 3.
[0099] Figure 6 : is a block diagram showing a configuration example of an abnormality diagnosis device according to Embodiment 3. Figure 6 In the abnormality diagnosis device 100B of the third embodiment, Figure 1 The structure of the abnormality diagnosis device 100 of the first embodiment shown in the figure is supplemented with a soundproof wall 12 for reducing surrounding noise. The soundproof wall 12 is arranged so as to cover the surroundings of the identification target object 1. Figure 1 The same or equivalent components are denoted by the same reference numerals, and duplicate descriptions are omitted.
[0100] exist Figure 6 In the structure of , a general sound absorbing material, a metal plate, or a combination of these can be used as the sound insulation wall 12. The thickness of the sound insulation wall 12 is preferably changed according to the size of the surrounding noise.
[0101] In the third embodiment, the method for determining whether the waveform data is abnormal is basically the same as that in the first embodiment. However, in the case of the structure of the first embodiment, noise from the surroundings may be mixed in. In addition, in the case of the structure of the second embodiment, in the case of an environment where the noise from the surroundings cannot be completely removed or the noise is always mixed in, it is possible to misjudge the noise and abnormal sound. In contrast, in the third embodiment, the noise from the surroundings can be attenuated by the soundproof wall 12. Although the method of the third embodiment is a simple method, it can provide a determination method that reduces the influence of the noise from the surroundings.
[0102] As described above, the abnormality diagnosis device of the third embodiment includes a soundproof wall arranged so as to cover the surroundings of the identification object. The soundproof wall attenuates the noise from the surroundings, thereby providing a determination method that reduces the influence of the noise from the surroundings. As a result, it is possible to avoid misjudging the noise contained in the sound of the identification object as abnormality.
[0103] Implementation method 4.
[0104] Figure 7 : is a block diagram showing a configuration example of an abnormality diagnosis device according to a fourth embodiment. Figure 7 In the abnormality diagnosis device 100C of the fourth embodiment, Figure 1 The abnormality diagnosis device 100 of the first embodiment shown in the figure has a movable part 13 for moving the microphone 2 and a movable control part 14 for controlling the movement of the movable part 13. Figure 1 The same or equivalent components are denoted by the same reference numerals, and duplicate descriptions are omitted.
[0105] Here, the movable part 13 can use a multi-joint robot, a linear motion robot, an air cylinder, a hydraulic cylinder, etc. In addition, the movable control part 14 can use a computer, a sequencer, etc. The movable control part 14 can also be equipped with a position detection sensor, a camera, etc. for determining the movement of the movable part 13.
[0106] In Embodiment 4, the method of determining whether waveform data is abnormal is basically the same as that in Embodiment 1. However, in Embodiment 1, for example, when the determination object 1 moves, if the microphone 2 is fixed, the distance between the determination object 1 and the microphone 2 changes, making it difficult to stably measure the sound generated from the determination object 1. The case where the determination object 1 moves is, for example, a case where the determination object 1 is placed on a belt conveyor and flows on a production line.
[0107] In contrast, in the fourth embodiment, the movable control unit 14 can cause the movable unit 13 to move in accordance with the movement of the identification object 1. This can reduce the change in the distance between the identification object 1 and the microphone 2, or make the change in the distance between the identification object 1 and the microphone 2 zero, so that the sound generated from the identification object 1 can be stably measured.
[0108] In addition, in the case where the structure of the discrimination object 1 is complicated, it is assumed that a sufficient volume cannot be obtained due to the increase in the distance between the discrimination object 1 and the microphone 2. Failure to obtain a sufficient volume means that waveform data with sufficient amplitude cannot be obtained. In the case where waveform data with sufficient amplitude cannot be obtained, it is possible that the sound required for discrimination is not included in the waveform data, resulting in errors in discrimination. When the operator makes the discrimination, the operator's ear is brought close to the action sound generating part in the discrimination object 1 for confirmation, but in the case of the structure of the first embodiment, it is difficult to bring the microphone 2 close to the action sound generating part in the discrimination object 1. In contrast, in the fourth embodiment, the movable part 13 can be moved by the movable control part 14 so that the microphone 2 is sufficiently close to the action sound generating part in the discrimination object 1. As a result, the microphone 2 can be brought close to a position where sufficient amplitude can be obtained for measurement. As a result, waveform data with sufficient amplitude can be obtained, and the quality of the waveform data can be judged with high accuracy.
[0109] As described above, the abnormality diagnosis device of the fourth embodiment includes a movable part for moving the microphone and a movable control part for controlling the movement of the movable part. According to this structure, the microphone can be brought close to the action sound generating part in the identification object. Thus, waveform data with sufficient amplitude can be obtained, and the quality of the waveform data can be determined with high accuracy.
[0110] Implementation method 5.
[0111] exist Figure 1 In the structure of the abnormality diagnosis device of the embodiment 1 shown, it is possible that noise from the surroundings is mixed in and noise and abnormal sounds are misjudged. In addition, in the case of the structure of the embodiment 2, in the case of an environment where the noise from the surroundings cannot be completely removed or noise is always mixed in, noise and abnormal sounds may be misjudged. In addition, in the method of the embodiment 2, when the action sound generated by the identification object 1 is loud, there is a possibility that the action sound is also input to the noise collection microphone 11, and the time period mixed with noise cannot be estimated. Therefore, in the embodiment 5, a method for improving them is proposed. In addition, the method of the embodiment 5 can be used Figure 1 The structure of the abnormality diagnosis device 100 of the first embodiment shown in the figure is implemented.
[0112] Next, the algorithm of the estimation process used in Embodiment 5 is described. First, the signal processing unit 4 performs STFT on the waveform data converted by the signal converter 3, dividing the waveform data into an arbitrary number of STFTs in the frequency direction and the time direction. This process is the same as that in Embodiment 1. When the matrix F as the processing result of the STFT is shown again, it is represented by the following formula (12).
[0113] [Formula 12]
[0114]
[0115] In the above formula (12), i is the number of divisions of the waveform data in the frequency direction, and j is the number of divisions of the waveform data in the time direction.
[0116] The signal processing unit 4 calculates a matrix R represented by the following equation (13) for the matrix F which is the result of the STFT process.
[0117] [Formula 13]
[0118] R=F·ε…(13)
[0119] In the above formula (13), ε is a matrix including elements of the vector of the average ν shown in the above formula (4), and is represented by the following formula (14).
[0120] [Formula 14]
[0121]
[0122] As shown in the above equation (14), the matrix ε is a diagonal matrix, and its diagonal elements are represented by the following equation (15).
[0123] [Formula 15]
[0124]
[0125] As shown in the above formula (15), the diagonal elements e of the matrix ε are i,i By the average ν i The reciprocal of .
[0126] The matrix R is calculated as information about the frequency of the noise of the known normal product. In addition, as shown in the above equations (13) to (15), the matrix R is represented by the product of the matrix F and the matrix ε having the inverse of the average ν, and is calculated as an estimated value of the unstable noise. In addition, when a sufficient number of average values ν are not obtained, the unit matrix can be used instead of the matrix ε.
[0127] The signal processing unit 4 obtains the sum per unit time for each element of the matrix R representing the estimated value of the non-stationary noise. If a vector whose elements are the sums per unit time is represented by R', the sum vector R' can be represented as in the following equation (16).
[0128] [Formula 16]
[0129]
[0130] The signal processing unit 4 obtains the average value of the total sum vector R' and identifies the elements of the total sum vector R' that exceed the average value of the whole. Then, the signal processing unit 4 estimates the interval to which the elements exceeding the average value of the whole belong as a time interval mixed with unstable noise, that is, a time interval mixed with noise other than the operation sound. Then, the methods of the first and second embodiments are applied to the waveform data after excluding the estimated time interval. Thus, a discrimination method that reduces the influence of surrounding noise can be provided.
[0131] In the above description, the average value of the total sum vector R' is used as the reference value for comparison, that is, the threshold value, but the present invention is not limited thereto. The value obtained by multiplying the average value of the total sum vector R' by an arbitrary coefficient may be used as the threshold value.
[0132] As described above, according to the abnormality diagnosis device of the fifth embodiment, the signal processing device uses the waveform data of the operation sound of the identification object to estimate the time interval in which noise other than the operation sound is mixed, and performs short-time fast Fourier transform on the waveform data after excluding the estimated time interval to calculate the feature value. This can avoid misjudging the noise contained in the sound of the identification object as abnormality.
[0133] In addition, in the abnormality diagnosis method of the fifth embodiment, in the abnormality diagnosis method described in the first embodiment, an estimation step is included between the second step and the third step, in which the waveform data of the operation sound of the electrical device is used to estimate the time interval in which noise other than the operation sound is mixed. Then, in the third step, the waveform data after excluding the time interval estimated in the estimation step is subjected to a short-time fast Fourier transform to calculate the feature quantity. In this way, a discrimination method that reduces the influence of surrounding noise can be provided. Thus, it is possible to avoid misjudging the noise contained in the sound of the discrimination object as abnormal.
[0134] Implementation method 6.
[0135] exist Figure 1In the structure of the abnormality diagnosis device 100 of the first embodiment shown, known noises from the surroundings, such as the ringtone of the notification time and the melody produced by the AGV (Automated Guided Vehicle) running in the factory, may be mixed in, thereby misjudging noise and abnormal sounds. In addition, in the case of the structure of the second embodiment, in the case of an environment where the noise from the surroundings cannot be completely removed or the noise is always mixed in, it is possible to misjudge noise and abnormal sounds. In addition, in the method of the second embodiment, when the action sound generated by the identification object 1 is loud, there is a possibility that the action sound is also input to the noise collection microphone 11, and the time period mixed with the noise cannot be estimated. Therefore, in the sixth embodiment, a method for improving them is proposed. In addition, with respect to Figure 1 The structure of the abnormality diagnosis device 100 of the first embodiment shown can be used Figure 8 The method of the sixth embodiment is implemented by the abnormality diagnosis device 100D including the signal processing device 10D as shown. Figure 8 1 is a block diagram showing a configuration example of an abnormality diagnosis device 100D according to a sixth embodiment. Figure 1 The signal processing device 10D included in the abnormality diagnosis device 100D further includes a preprocessing unit 30. The data storage unit 5 of the signal processing device 10D stores feature data based on normal known waveform data and known noise data.
[0136] Next, the algorithm of the estimation process used in the sixth embodiment is described. First, the preprocessing unit 30 performs STFT on the waveform data and the known noise data converted by the signal converter 3, respectively, to divide the waveform data and the noise data into an arbitrary number of STFTs in the frequency direction and the time direction. In addition, the preprocessing unit 30 performs pattern matching between the matrices as the processing results of the STFT. A general algorithm can be used for pattern matching.
[0137] Here, the matrix representing the STFT result of the waveform data converted by the signal converter 3 is represented by I, and is called the "first signal matrix". In addition, the matrix representing the STFT result of the known noise data is represented by T, and is called the "first noise matrix". When the size of the first noise matrix T in the time direction and the frequency direction is set to M and N respectively, R represented by the following formula (17) can be used in pattern matching. NCC (Normalized Cross-Correlation: normalized cross-correlation).
[0138] [Formula 17]
[0139]
[0140] When the normalized cross-correlation R obtained by pattern matching NCC In a certain index (i, j), the approximation threshold Th is exceeded. Near When the preprocessing unit 30 converts the normalized cross-correlation R in the first signal matrix I converted by the signal converter 3, NCC The index at which the value of takes the maximum value is set as the start time, and the data from the start time to the size of the first noise matrix T in the time direction is judged as a known noise mixing interval.
[0141] Next, use Fig. 9 Describe the known noise removal process. Fig. 9 This is a diagram for explaining a known noise removal process performed using the abnormality diagnosis device 100D according to the sixth embodiment.
[0142] The preprocessing unit 30 determines whether the waveform data converted by the signal converter 3 contains known noise data. Fig. 9 In the example of , the portion indicated by the thick solid line frame is determined to be a known noise mixing interval.
[0143] If it is determined that the interval is mixed with known noise, the preprocessing unit 30 sets the value of the first noise matrix T to exceed the replacement target threshold value Th. pow The value of the element is replaced by the noise removal variable P based on pattern matching rep , generate the second noise matrix T' by replacing some elements of the STFT result of the known noise data. Fig. 9 In the example, the replacement target threshold Th pow = 100, and in the first noise matrix T, the values of "150", "200", and "450" exceeding the value of 100 are replaced by the noise removal variable P rep The second noise matrix T' is generated by using the value of. Here, as the noise removal variable Prep based on pattern matching, the average value of the data of the part other than the known noise mixing interval in the first signal matrix I can be used to calculate the external value. Fig. 9 , it is shown that the average value of each frequency of the portion other than the data of the known noise mixing interval in the first signal matrix I is set as the noise removal variable P rep Example.
[0144] Finally, the preprocessing unit 30 replaces the data of the part of the mixing interval of the known noise in the first signal matrix I with the second noise matrix T', thereby generating a second signal matrix I' from which the influence caused by the mixing of the known noise is removed. The subsequent processing uses the second signal matrix I' to perform any of the processing in the above-mentioned embodiments 1 to 5.
[0145] In the sixth embodiment, the structure in which the pre-processing unit 30 is added to the front stage of the signal processing unit 4 is applied to Figure 1 The structure of the abnormality diagnosis device 100 of the first embodiment shown in the figure can also be applied to Figure 4 The structure of the second embodiment shown, Figure 6 The structure of the third embodiment shown and Figure 7 The structure of embodiment 4 is shown.
[0146] As described above, according to the abnormality diagnosis device of the sixth embodiment, the signal processing device includes a preprocessing unit, the preprocessing unit uses known noise data, estimates the time interval in which the known noise data is mixed based on the result of the short-time fast Fourier transform, and the signal processing unit performs processing to calculate the feature amount based on the data from which the noise data is removed. This can avoid misjudging the known noise contained in the sound of the identification object as abnormality.
[0147] In the abnormality diagnosis device of the sixth embodiment, the preprocessing unit performs short-time fast Fourier transform on the waveform data and the known noise data converted by the signal converter, respectively, to divide the waveform data and the noise data into any number in the frequency direction and the time direction, and performs pattern matching between the matrices as the processing results of the short-time fast Fourier transform. In addition, the signal processing unit calculates the first feature quantity data using the data generated based on the processing result of the pattern matching of the preprocessing unit. In this way, it is possible to provide an abnormality diagnosis device that reduces the influence of the surrounding noise.
[0148] In addition, in the abnormality diagnosis method of embodiment 6, in the abnormality diagnosis method described in embodiments 1 to 5, an estimation step is included between the second step and the third step, in which a time interval in which known noise data is mixed is estimated based on the result of short-time fast Fourier transform. Then, in step 3, the waveform data after replacing the data of the time interval estimated in the estimation step with other data is subjected to short-time fast Fourier transform to calculate the feature quantity. In this way, a discrimination method that reduces the influence of surrounding noise can be provided. Thus, it is possible to avoid misjudging the noise contained in the sound of the discrimination object as abnormal.
[0149] The configurations described in the above embodiments are merely examples, and may be combined with other known techniques, the embodiments may be combined with each other, and a portion of the configuration may be omitted or modified within a scope not departing from the gist.
[0150] Description of symbols
[0151] 1: Identification object; 2: Microphone; 3: Signal converter; 4: Signal processing unit; 5: Data storage unit; 6: Identification unit; 10, 10D: Signal processing device; 11, Noise collection microphone; 12: Sound insulation wall; 13: Movable unit; 14: Movable control unit; 21, 23: Adder; 22: Filter; 30: Preprocessing unit; 100, 100A, 100B, 100C, 100D: Abnormality diagnosis device.
Claims
1. An abnormality diagnosis device, characterized in that: The abnormality diagnosis device comprises: A microphone that converts the sound of the object to be identified into an analog electrical signal; a signal converter that converts the analog electrical signal into a digital signal; and a signal processing device that receives the digital signal and performs signal processing, The signal processing device comprises: A signal processing unit, which performs short-time fast Fourier transform on the waveform data of the input signal to calculate the feature quantity; A data storage unit storing feature quantity data based on normal known waveform data; as well as a determination unit that compares first feature quantity data composed of a plurality of feature quantities calculated by the signal processing unit with second feature quantity data, which is feature quantity data stored in the data storage unit, to determine whether the waveform data of the input signal is good or bad; The feature amount is an amount indicating the degree of variation in the intensity of a spectrum of a specific frequency band of waveform data over time.
2. The abnormality diagnosis device according to claim 1, characterized in that: The signal processing device performs filter bank processing on the feature quantity generated by the short-time fast Fourier transform, and generates the first feature quantity data based on the feature quantity subjected to the filter bank processing.
3. The abnormality diagnosis device according to claim 1 or 2, characterized in that: The signal processing unit performs a short-time fast Fourier transform on the waveform data converted by the signal converter, calculates a matrix F having i frequency and j time dimensions represented by the following formula (1), and calculates a feature quantity vector T represented by the following formula (2) using the standard deviation of the spectrum intensity in the time direction in the matrix F as the feature quantity as the first feature quantity data, The data storage unit stores the data based on T' k The data related to the average ν and the standard deviation σ calculated by the following equations (3) and (4) are used as the second feature quantity data. k It is a collection of feature quantities based on multiple normal and known waveform data. The determination unit determines whether the waveform data of the input signal is abnormal based on the deviation between the first feature quantity data and the second feature quantity data. [Formula 1] [Formula 2] [Formula 3] [Formula 4] 4. The abnormality diagnosis device according to claim 3, characterized in that: The abnormality diagnosis device includes a noise collecting microphone for collecting surrounding noise. The signal processing device estimates a time period in which surrounding noise is mixed using the waveform data collected using the noise collecting microphone, and performs a short-time fast Fourier transform on the waveform data excluding the estimated time period to calculate a feature value.
5. The abnormality diagnosis device according to claim 3 or 4, characterized in that: The signal processing device estimates a time interval in which noise other than the operation sound is mixed using the waveform data of the operation sound of the identification target object, and performs short-time fast Fourier transform on the waveform data excluding the estimated time interval to calculate a feature value.
6. The abnormality diagnosis device according to any one of claims 3 to 5, characterized in that: The signal processing device includes a preprocessing unit, which uses known noise data to estimate the time interval mixed with the known noise data based on the result of short-time fast Fourier transform. The signal processing unit performs short-time fast Fourier transform on the waveform data after replacing the data of the time interval mixed with the noise data with other data, and calculates the feature value.
7. The abnormality diagnosis device according to claim 6, characterized in that: The preprocessing unit performs short-time fast Fourier transform on the waveform data and the known noise data converted by the signal converter, respectively dividing the waveform data and the noise data into any number in the frequency direction and the time direction, and performs pattern matching between matrices as processing results of the short-time fast Fourier transform. The signal processing unit calculates the first feature value data using data generated based on a processing result of the pattern matching by the pre-processing unit.
8. The abnormality diagnosis device according to claim 7, characterized in that: When a matrix representing a result of a short-time fast Fourier transform of the waveform data is set as a first signal matrix, and a matrix representing a result of a short-time fast Fourier transform of the known noise data is set as a first noise matrix, The preprocessing unit determines that there is a known noise mixing interval when the waveform data includes the known noise data, generates a second noise matrix by replacing the values of the elements of the first noise matrix whose values exceed the first threshold value with noise removal variables based on pattern matching, and generates a second signal matrix by replacing the data of the part of the known noise mixing interval in the first signal matrix with the data of the second noise matrix, The signal processing unit calculates the first feature quantity data using the second signal matrix.
9. The abnormality diagnosis device according to claim 8, characterized in that: The preprocessing unit uses normalized cross-correlation in pattern matching between the first signal matrix and the first noise matrix, and when the normalized cross-correlation exceeds a second threshold value at a certain index, in the first signal matrix, the index at which the normalized cross-correlation value takes a maximum value is set as a start time, and the data from the start time to the size of the first noise matrix in the time direction is judged as a known noise mixing interval.
10. The abnormality diagnosis device according to any one of claims 3 to 9, characterized in that: The abnormality diagnostic device includes a sound insulation wall disposed so as to cover the periphery of the identification target object.
11. The abnormality diagnosis device according to any one of claims 3 to 10, characterized in that: The abnormality diagnosis device comprises: a movable portion for moving the microphone; and A movable control unit controls the movement of the movable unit.
12. An abnormality diagnosis method for diagnosing abnormalities in electrical equipment using a computer, wherein the computer is configured to refer to a feature quantity representing a degree of deviation of a spectral intensity variation caused by time in a specific frequency band of normal known waveform data, characterized in that: The abnormality diagnosis method comprises: Step 1, converting the operating sound of the electrical device into an analog electrical signal; Step 2, converting the analog electrical signal into a digital signal; The third step is to perform a short-time fast Fourier transform on the waveform data converted in the second step to calculate the feature value; as well as In the fourth step, the first feature quantity data composed of the plurality of feature quantities calculated in the third step is compared with the second feature quantity data stored in the computer to determine whether the waveform data is good or bad.
13. The abnormality diagnosis method according to claim 12, characterized in that: An estimation step is included between the second step and the third step, in which a time interval in which the surrounding noise is mixed is estimated using waveform data collected using a noise collecting microphone for collecting surrounding noise, In the third step, short-time fast Fourier transform is performed on the waveform data excluding the time interval estimated in the estimation step.
14. The abnormality diagnosis method according to claim 12, characterized in that: An estimation step is included between the second step and the third step, in which a time interval in which noise other than the operating sound is mixed is estimated using waveform data of the operating sound of the electrical device, In the third step, short-time fast Fourier transform is performed on the waveform data excluding the time interval estimated in the estimation step.
15. The abnormality diagnosis method according to claim 12, characterized in that: An estimation step is included between the second step and the third step, in which a time interval in which known noise data is mixed is estimated based on a result of a short-time fast Fourier transform. In the third step, short-time fast Fourier transform is performed on the waveform data obtained by replacing the data of the time interval estimated in the estimation step with other data.
16. An abnormality diagnosis program that causes a computer to diagnose abnormalities in electrical equipment, the computer being configured to refer to a feature quantity that indicates a degree of deviation in a time-dependent spectrum intensity variation in a specific frequency band of normal known waveform data, characterized in that: The abnormality diagnosis program causes the computer to execute a process including the following steps: Step 1, converting the operating sound of the electrical device into an analog electrical signal; Step 2, converting the analog electrical signal into a digital signal; The third step is to perform a short-time fast Fourier transform on the waveform data converted in the second step to calculate the feature value; as well as In the fourth step, the first feature quantity data composed of the plurality of feature quantities calculated in the third step is compared with the second feature quantity data stored in the computer to determine whether the waveform data is good or bad.
Citation Information
Cited By
Mecanum wheel automatic guided vehicle fault detection method and system
CN122088133A
