Fault feature extraction method and device based on Internet of Things, equipment and medium

Abstract

The invention discloses a fault feature extraction method and device based on the Internet of Things, electronic equipment and a storage medium. The method comprises the following steps: sampling vibration signals sent by a vibration sensor on industrial equipment to obtain a discrete signal sequence; dividing the discrete signal sequence into a plurality of time frame signals; performing component decomposition on the time frame signal to obtain a plurality of signal components; performing wavelet transformation on the signal component to obtain a wavelet coefficient; determining the frequency scale of the wavelet coefficient based on the similarity objective function; and performing calculation based on all the frequency scales to obtain a fault characteristic parameter corresponding to the time frame signal. According to the method, the bandwidth requirement of fault diagnosis in the Internet of Things environment can be remarkably reduced, so that the real-time performance of work such as fault diagnosis is remarkably improved while it is ensured that equipment faults are effectively recognized.

Description

Technical Field

[0001] The present application relates to the field of Internet of Things technology, and specifically to a fault feature extraction method, device, electronic device, readable storage medium and computer program product based on the Internet of Things. Background Art

[0002] Industrial equipment operation and maintenance systems often rely on traditional Industrial Internet of Things (IIoT) architectures. These typically collect large amounts of data through sensors and upload it indiscriminately via network transmission devices, enabling servers to drive, store, predict, and issue early warnings based on this indiscriminate data. However, with the increasing number of industrial devices, the need for simultaneous data upload is also increasing. This makes it difficult for traditional IIoT architectures to withstand the increasing pressure on network bandwidth, directly impacting the real-time nature of various tasks. Summary of the Invention

[0003] In view of the above problems, the present application provides a fault feature extraction method, device, electronic device, readable storage medium and computer program product based on the Internet of Things, which can solve the problem of limited network bandwidth in the traditional industrial Internet of Things architecture and the problem of work processing timeliness caused by it.

[0004] In a first aspect, the present application provides a method for extracting fault features based on the Internet of Things, comprising:

[0005] Sampling the vibration signal sent by the vibration sensor on the industrial equipment to obtain a discrete signal sequence;

[0006] Dividing the discrete signal sequence into a plurality of time frame signals;

[0007] Decomposing the time frame signal into components to obtain multiple signal components;

[0008] Performing wavelet transform on the signal components to obtain wavelet coefficients;

[0009] determining a frequency scale of the wavelet coefficients based on a similarity objective function;

[0010] Calculations are performed based on all frequency scales to obtain fault characteristic parameters corresponding to the time frame signal.

[0011] In the above technical solution, the method can extract fault features from vibration signals of industrial equipment in an IoT environment, thereby significantly reducing the data capacity and facilitating the uploading of large amounts of data in an IoT environment, thereby solving the problem of limited network bandwidth in traditional industrial IoT architectures and ensuring the timeliness of corresponding work processing.

[0012] In some implementations, decomposing the time frame signal into components to obtain multiple signal components includes:

[0013] Decomposing the time frame signal into components to obtain a plurality of signal components that meet the component decomposition conditions;

[0014] The component decomposition conditions include:

[0015] The number of extreme points and the number of zero crossings are equal or differ by one; and

[0016] At any point, the mean value of the envelope defined by the local maximum and local minimum is zero.

[0017] In the above technical solution, the method can ensure the validity and accuracy of the signal components through strict component decomposition conditions, thereby more accurately extracting the fault characteristics in the vibration signal of the industrial equipment.

[0018] In some implementations, performing component decomposition on the time frame signal to obtain a plurality of signal components that satisfy the component decomposition condition includes:

[0019] Calculating the local maximum of the time frame signal and interpolating the local maximum to obtain an upper envelope sequence;

[0020] Calculating a local minimum value of the time frame signal and interpolating the local minimum value to obtain a lower envelope sequence;

[0021] Calculating a mean sequence of the upper envelope sequence and the lower envelope sequence;

[0022] determining whether a difference between the time frame signal and the mean value sequence satisfies the component decomposition condition;

[0023] If the difference satisfies the component decomposition condition, the difference is determined as a signal component.

[0024] In the above technical solution, the method can accurately decompose the signal components that meet specific conditions through envelope analysis and condition judgment, thereby providing a high-quality data basis for subsequent fault feature extraction.

[0025] In some embodiments, the method further comprises:

[0026] If the difference does not meet the component decomposition condition, the difference is used as the time frame signal, and the steps of calculating the local maximum of the time frame signal and interpolating the local maximum to obtain the upper envelope sequence are re-executed.

[0027] In the above technical solution, the method can iteratively process the difference signal that does not meet the component decomposition conditions, thereby gradually approximating and extracting the signal components that meet the conditions, ensuring the accuracy of signal decomposition, and thus improving the reliability of fault feature extraction.

[0028] In some embodiments, the method further comprises:

[0029] calculating a difference between the time frame signal and the signal component to obtain a residual value;

[0030] Determining whether the residual value is a monotonic function or has only one extreme value;

[0031] If the residual value is not a monotonic function or has more than one extreme value, then taking the residual value as the time frame signal, re-performing the step of calculating the local maximum value of the time frame signal, and interpolating the local maximum value to obtain the upper envelope sequence;

[0032] If the residual value is a monotonic function or has only one extreme value, all the obtained signal components are determined to be a plurality of signal components that meet the component decomposition condition.

[0033] In the above technical solution, the method can iteratively extract signal components in the time frame signal by judging the functional characteristics of the residual value, thereby ensuring that all signal components complete component decomposition, thereby improving the accuracy and completeness of fault feature extraction.

[0034] In some implementations, performing a wavelet transform on the signal component to obtain wavelet coefficients includes:

[0035] matching a wavelet function based on the operating environment of the industrial equipment;

[0036] Performing wavelet transform on the signal component based on the wavelet function to obtain wavelet coefficients.

[0037] In the above technical solution, the method can flexibly select the wavelet function according to the operating environment of the industrial equipment, so as to more accurately perform wavelet transform on the signal components, so that the obtained wavelet coefficients can more effectively reflect the signal characteristics, thereby improving the pertinence and accuracy of fault feature extraction.

[0038] In some embodiments, determining the frequency scale of the wavelet coefficients based on a similarity objective function includes:

[0039] Substituting the wavelet coefficients and the signal components into a similarity objective function, a frequency scale is obtained that maximizes the result of the similarity objective function; the frequency scale is the frequency scale of the wavelet coefficients.

[0040] In the above technical solution, the method can automatically determine the optimal frequency scale of the wavelet coefficients through the similarity objective function, thereby facilitating more accurate disclosure of fault characteristics in the signal.

[0041] In some embodiments, the calculation based on all frequency scales to obtain the fault characteristic parameter corresponding to the time frame signal includes:

[0042] Aggregate calculations are performed on multiple frequency scales corresponding to each time frame signal to obtain multiple fault characteristic parameters corresponding to the multiple time frame signals.

[0043] In the above technical solution, the method can effectively integrate information at different frequency scales by performing aggregation calculations on multiple frequency scales corresponding to each time frame signal, thereby obtaining more representative and stable fault characteristic parameters.

[0044] In some embodiments, the method further comprises:

[0045] Convert the multiple fault characteristic parameters into JSON format.

[0046] The JSON format data is uploaded to the cloud platform via the MQTT protocol, so that the cloud platform performs fault analysis on the industrial equipment based on the multiple fault characteristic parameters.

[0047] In the above technical solution, the method can convert multiple fault characteristic parameters into JSON format data, and efficiently upload it to the cloud platform through the MQTT protocol, realizing standardized data transmission and cloud-based fault analysis, thereby ensuring the fault diagnosis capability of industrial equipment in the Internet of Things environment.

[0048] In a second aspect, the present application provides a fault feature extraction device based on the Internet of Things, comprising:

[0049] A sampling unit, used to sample the vibration signal sent by the vibration sensor on the industrial equipment to obtain a discrete signal sequence;

[0050] a dividing unit, configured to divide the discrete signal sequence into a plurality of time frame signals;

[0051] A component decomposition unit, configured to decompose the time frame signal into components to obtain a plurality of signal components;

[0052] A wavelet transform unit, configured to perform wavelet transform on the signal components to obtain wavelet coefficients;

[0053] a frequency scale calculation unit, configured to determine the frequency scale of the wavelet coefficients based on a similarity objective function;

[0054] The fault characteristic calculation unit is used to perform calculations based on all frequency scales to obtain fault characteristic parameters corresponding to the time frame signal.

[0055] In the above technical solution, the device can extract fault features of industrial equipment vibration signals based on the Internet of Things, thereby significantly reducing the data capacity and facilitating the uploading of large amounts of data in the Internet of Things environment, thereby solving the problem of limited network bandwidth in the traditional industrial Internet of Things architecture and ensuring the timeliness of corresponding work processing.

[0056] In some embodiments, the component decomposition unit is specifically configured to perform component decomposition on the time frame signal to obtain a plurality of signal components that meet the component decomposition conditions;

[0057] The component decomposition conditions include:

[0058] The number of extreme points and the number of zero crossings are equal or differ by one; and

[0059] At any point, the mean value of the envelope defined by the local maximum and local minimum is zero.

[0060] In the above technical solution, the device can ensure the validity and accuracy of the signal components through strict component decomposition conditions, thereby more accurately extracting the fault characteristics in the vibration signal of the industrial equipment.

[0061] In some embodiments, the component decomposition unit includes:

[0062] a calculation subunit, configured to calculate local maximum values of the time frame signal and interpolate the local maximum values to obtain an upper envelope sequence;

[0063] The calculation subunit is further configured to calculate a local minimum value of the time frame signal and interpolate the local minimum value to obtain a lower envelope sequence;

[0064] The calculation subunit is further configured to calculate a mean value sequence of the upper envelope sequence and the lower envelope sequence;

[0065] a judging subunit, configured to judge whether the difference between the time frame signal and the mean value sequence satisfies the component decomposition condition;

[0066] The determination subunit is configured to determine the difference as a signal component when the difference satisfies the component decomposition condition.

[0067] In the above technical solution, the device can accurately decompose the signal components that meet specific conditions through envelope analysis and conditional judgment, thereby providing a high-quality data foundation for subsequent fault feature extraction based on the Internet of Things.

[0068] In some embodiments, the calculation subunit is further used to use the difference as a time frame signal when the difference does not meet the component decomposition condition, calculate the local maximum of the time frame signal, and interpolate the local maximum to obtain an upper envelope sequence.

[0069] In the above technical solution, the device can iteratively process the difference signal that does not meet the component decomposition conditions, thereby gradually approximating and extracting the signal components that meet the conditions, ensuring the accuracy of signal decomposition, and thus improving the reliability of fault feature extraction based on the Internet of Things.

[0070] In some embodiments, the component decomposition unit includes:

[0071] The calculation subunit is further configured to calculate a difference between the time frame signal and the signal component to obtain a residual value;

[0072] The judging subunit is further configured to judge whether the residual value is a monotonic function or has only one extreme value;

[0073] The calculation subunit is further configured to, when the residual value is not a monotonic function or has more than one extreme value, use the residual value as a time frame signal, calculate a local maximum value of the time frame signal, and interpolate the local maximum value to obtain an upper envelope sequence;

[0074] The determining subunit is further configured to determine all obtained signal components into a plurality of signal components that meet the component decomposition condition when the residual value is a monotonic function or has only one extreme value.

[0075] In the above technical solution, the device can iteratively extract signal components in the time frame signal by judging the functional characteristics of the residual value, thereby ensuring that all signal components complete component decomposition, thereby improving the accuracy and completeness of fault feature extraction based on the Internet of Things.

[0076] In some embodiments, the wavelet transform unit includes:

[0077] a matching subunit, configured to match a wavelet function based on an operating environment of the industrial equipment;

[0078] The transform subunit is configured to perform wavelet transform on the signal component based on the wavelet function to obtain wavelet coefficients.

[0079] In the above technical solution, the device can flexibly select the wavelet function according to the operating environment of the industrial equipment, so as to more accurately perform wavelet transform on the signal components, so that the obtained wavelet coefficients can more effectively reflect the signal characteristics, thereby improving the pertinence and accuracy of fault feature extraction based on the Internet of Things.

[0080] In some embodiments, the frequency scale calculation unit is specifically configured to substitute the wavelet coefficients and the signal components into a similarity objective function to obtain a frequency scale that maximizes the result of the similarity objective function; the frequency scale is the frequency scale of the wavelet coefficients.

[0081] In the above technical solution, the device can automatically determine the optimal frequency scale of the wavelet coefficients through the similarity objective function, which is conducive to more accurately revealing the fault characteristics in the signal.

[0082] In some implementations, the fault feature calculation unit is specifically configured to perform aggregate calculation on multiple frequency scales corresponding to each time frame signal to obtain multiple fault feature parameters corresponding one-to-one to the multiple time frame signals.

[0083] In the above technical solution, the device can effectively integrate information at different frequency scales by performing aggregate calculations on multiple frequency scales corresponding to each time frame signal, thereby obtaining more representative and stable fault characteristic parameters.

[0084] In some embodiments, the IoT-based fault feature extraction device further includes:

[0085] A conversion unit, configured to convert the format of the plurality of fault characteristic parameters to obtain data in JSON format;

[0086] The uploading unit is used to upload the JSON format data to the cloud platform through the MQTT protocol, so that the cloud platform can perform fault analysis on the industrial equipment based on the multiple fault characteristic parameters.

[0087] In the above technical solution, the device can convert multiple fault characteristic parameters into JSON format data and efficiently upload it to the cloud platform through the MQTT protocol, realizing standardized data transmission and cloud-based fault analysis, thereby ensuring the fault diagnosis capability of industrial equipment in the Internet of Things environment.

[0088] In a third aspect, the present application provides an electronic device, comprising a memory and a processor, wherein the memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to execute the Internet of Things-based fault feature extraction method described in any one of the first aspects.

[0089] In a fourth aspect, the present application provides a readable storage medium, wherein the readable storage medium stores a computer program. When the computer program is executed by a processor, the fault feature extraction method based on the Internet of Things described in any one of the first aspects is executed.

[0090] In a fifth aspect, the present application provides a computer program product, which includes a computer program. When the computer program is executed by a processor, it executes the fault feature extraction method based on the Internet of Things described in any one of the first aspects.

[0091] The beneficial effects of the present application are as follows: the method and device can significantly reduce the bandwidth requirements for fault diagnosis in an Internet of Things environment through operations such as fine signal component decomposition, adaptive wavelet transform, optimized calculation of frequency scales, and standardized transmission and cloud analysis of fault characteristic parameters, thereby ensuring that a large number of industrial equipment faults are effectively identified while significantly improving the real-time performance of fault diagnosis and other tasks. BRIEF DESCRIPTION OF THE DRAWINGS

[0092] To more clearly illustrate the technical solutions of the embodiments of the present application, the following briefly describes the drawings required for use in the embodiments of the present application. The following drawings illustrate only certain embodiments of the present application and should not be construed as limiting the scope thereof. The same reference numerals are used throughout the drawings to represent the same content.

[0093] Figure 1 This is a flowchart of a fault feature extraction method based on the Internet of Things in some embodiments of the present application;

[0094] Figure 2 This is a schematic diagram of a system architecture for applying the fault extraction method in some embodiments of the present application;

[0095] Figure 3 This is a flowchart of a fault feature extraction method based on the Internet of Things in some embodiments of the present application;

[0096] Figure 4 This is a schematic structural diagram of a fault feature extraction device based on the Internet of Things in some embodiments of the present application;

[0097] Figure 5 This is a schematic diagram of the structure of an electronic device in some embodiments of the present application. DETAILED DESCRIPTION

[0098] The following embodiments of the technical solution of the present application will be described in detail with reference to the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present application and are therefore only examples and are not intended to limit the scope of protection of the present application.

[0099] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which this application belongs; the terms used herein are only for the purpose of describing specific embodiments and are not intended to limit this application; the terms "including" and "having" and any variations thereof in the specification and claims of this application and the above-mentioned figure descriptions are intended to cover non-exclusive inclusions.

[0100] In the description of the embodiments of the present application, “multiple” means two or more (including two), unless otherwise clearly and specifically defined.

[0101] References herein to "embodiments" mean that a particular feature, structure, or characteristic described in connection with the embodiments may be included in at least one embodiment of the present application. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor does it constitute an independent or alternative embodiment that is mutually exclusive of other embodiments. It is understood, both explicitly and implicitly, by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0102] Industrial equipment operation and maintenance systems often rely on traditional Industrial Internet of Things (IIoT) architectures, where sensors installed on industrial equipment typically indiscriminately upload collected data to servers. However, as the number of industrial devices increases, the amount of data uploaded also increases dramatically, making it difficult for traditional architectures to cope with bandwidth pressures.

[0103] In response to the above technical problems, an embodiment of the present application provides a fault feature extraction method based on the Internet of Things. The method proposes a fault feature extraction method based on the Internet of Things consisting of steps such as fine signal component decomposition, adaptive wavelet transform, optimized calculation of frequency scale, and standardized processing of fault feature parameters.

[0104] In the above technical solution, this method can significantly reduce the bandwidth requirement for fault diagnosis in the Internet of Things environment, thereby significantly improving the real-time performance of fault diagnosis and other tasks while ensuring that a large number of industrial equipment faults are effectively identified.

[0105] like Figure 1 As shown, some embodiments of the present application provide a method for extracting fault features based on the Internet of Things, and the method for extracting fault features based on the Internet of Things includes:

[0106] S101. Sample the vibration signal sent by the vibration sensor on the industrial equipment to obtain a discrete signal sequence.

[0107] S102: Divide the discrete signal sequence into multiple time frame signals.

[0108] S103: Decompose the time frame signal into components to obtain multiple signal components.

[0109] S104: Perform wavelet transform on the signal components to obtain wavelet coefficients.

[0110] S105 . Determine the frequency scale of the wavelet coefficients based on the similarity objective function.

[0111] S106 . Perform calculations based on all frequency scales to obtain fault characteristic parameters corresponding to the time frame signal.

[0112] In some embodiments, the IoT-based fault feature extraction method is performed by an edge device, which is used to preliminarily process data acquired by vibration sensors on industrial equipment and then upload it to the cloud for fault diagnosis.

[0113] In some embodiments, industrial equipment refers to equipment at an industrial site, specifically an industrial robot.

[0114] For example, see Figure 2 , Figure 2 A schematic diagram of the system architecture for applying this fault extraction method is shown. Multiple vibration sensors are installed on the moving parts of an industrial robot. At least one of these sensors collects the moving part's vertical vibration signal, and at least one collects the moving part's horizontal vibration signal. After the vibration sensors acquire the corresponding signals, the edge device extracts the signal's features using digital signal processing technology, and finally transmits the features to a cloud server for storage.

[0115] For example, assuming the original vibration signal is x(o), then sampling the original signal at a specific frequency can produce a discrete signal sequence x(t) (t = 1, 2, 3...). The method then divides the discrete signal sequence x(t) into m time frame signals and decomposes each time frame signal into n signal components. Each component is then transformed to find the corresponding frequency scale a, thereby maximizing the objective function and ultimately determining all fault characteristic parameters corresponding to the time frame signal.

[0116] In these embodiments, the method can perform IoT-based fault feature extraction on vibration signals of industrial equipment, thereby significantly reducing the data capacity and facilitating the uploading of large amounts of data in an IoT environment, thereby solving the problem of limited network bandwidth in traditional industrial IoT architectures and ensuring the timeliness of corresponding work processing.

[0117] In order to obtain signal components more accurately, in some embodiments, the time frame signal is decomposed into components to obtain multiple signal components, including:

[0118] Decomposing the time frame signal into components to obtain multiple signal components that meet the component decomposition conditions;

[0119] The component decomposition conditions include:

[0120] The number of extreme points and the number of zero crossings are equal or differ by one; and

[0121] At any point, the mean value of the envelope defined by the local maximum and local minimum is zero.

[0122] In some embodiments, the method requires that the number of extreme points and zero crossing points must be equal or differ by at most 1, and at any point, the average value of the envelope defined by the local maximum and local minimum is zero.

[0123] In these embodiments, the method can ensure the validity and accuracy of signal components through strict component decomposition conditions, thereby more accurately extracting fault features from the vibration signal of industrial equipment.

[0124] In order to more accurately obtain signal components, in some embodiments, the time frame signal is decomposed into components to obtain multiple signal components that meet the component decomposition conditions, including:

[0125] Calculate the local maximum of the time frame signal and interpolate the local maximum to obtain the upper envelope sequence;

[0126] Calculate the local minimum of the time frame signal and interpolate the local minimum to obtain the lower envelope sequence;

[0127] Calculate the mean sequence of the upper envelope sequence and the lower envelope sequence;

[0128] Determine whether the difference between the time frame signal and the mean sequence meets the component decomposition conditions;

[0129] If the difference satisfies the component decomposition condition, the difference is determined as the signal component.

[0130] In some embodiments, the sequence length calculated by the above-mentioned local maximum algorithm and local minimum algorithm is usually smaller than the length of the time frame signal, so further interpolation method is required to perform fitting to obtain the upper envelope sequence and the lower envelope sequence.

[0131] In some embodiments, the component decomposition condition refers to the definition of the component in the method.

[0132] Exemplarily, the method can decompose the sequence x(t) into signal components by the following steps:

[0133] Find the local maxima in x(t) and connect them using interpolation to obtain the upper envelope sequence e max(t);

[0134] Find the local minima in x(t) and connect them using interpolation to obtain the lower envelope sequence e min (t);

[0135] Calculate the mean sequence of two envelopes

[0136] Subtract m1(t) from x(t) to obtain h1(t) = x(t) - m1(t);

[0137] At this time, it is determined whether h1(t) is a component that satisfies the component decomposition conditions;

[0138] If h1(t) satisfies the component decomposition condition, then execute: c1(t)=h1(t); c1(t) is the signal component.

[0139] In these embodiments, the method can accurately decompose signal components that meet specific conditions through envelope analysis and conditional judgment, thereby providing a high-quality data foundation for subsequent IoT-based fault feature extraction.

[0140] In order to find the signal components that meet the component decomposition conditions, in some embodiments, if the difference does not meet the component decomposition conditions, the difference is used as the time frame signal, the local maximum of the time frame signal is calculated again, and the local maximum is interpolated to obtain the upper envelope sequence.

[0141] Exemplarily, if h1(t) does not meet the component decomposition condition, x(t)=h1(t) is executed, and the mean sequence calculation process is repeated until c1(t) that meets the component decomposition condition is found; c1(t) is the signal component.

[0142] In these embodiments, the method can iteratively process the difference signal that does not meet the component decomposition conditions, thereby gradually approximating and extracting the signal components that meet the conditions, ensuring the accuracy of signal decomposition, and thereby improving the reliability of fault feature extraction based on the Internet of Things.

[0143] In order to obtain all signal components, in some embodiments, the method further includes:

[0144] Calculate the difference between the time frame signal and the signal components to obtain the residual value;

[0145] Determine whether the residual value is a monotonic function or has only one extreme value;

[0146] If the residual value is not a monotonic function or has more than one extreme value, then the residual value is used as the time frame signal, and the steps of calculating the local maximum of the time frame signal are re-executed, and the local maximum is interpolated to obtain the upper envelope sequence;

[0147] If the residual value is a monotonic function or has only one extreme value, all the obtained signal components are determined as multiple signal components that meet the component decomposition conditions.

[0148] For example, the method can calculate the residual value r1(t)=x(t)-c1(t), and perform x(t)=r1(t), and repeat the signal component decomposition process until n components are found, and the residual value r n (t) is a monotonic function or has only one extreme value. That is,

[0149] r n (t) = r n-1 (t)-c n (t);

[0150] And you can get:

[0151]

[0152] In these embodiments, the method can iteratively extract signal components in the time frame signal by determining the functional characteristics of the residual value, thereby ensuring that all signal components complete component decomposition, thereby improving the accuracy and completeness of fault feature extraction based on the Internet of Things.

[0153] In order to obtain more suitable wavelet coefficients, in some embodiments, wavelet transform is performed on the signal components to obtain wavelet coefficients, including:

[0154] Matching wavelet functions based on the working environment of industrial equipment;

[0155] The signal components are transformed by wavelet function to obtain wavelet coefficients.

[0156] Exemplarily, the signal transformation formula used for wavelet transform is as follows:

[0157]

[0158] Where c(t) is the signal component, ψ * is the complex conjugate of the wavelet function, a is the scale parameter, b is the translation parameter, and z(a, b) is the wavelet coefficient.

[0159] The above formula describes the process of performing an inner product operation on a signal with a wavelet function at different scales a and translation parameters b. By adjusting the translation parameter b, this method can change the position of the wavelet function on the time axis, thereby matching it with the signal at different time locations. z(a, b) describes the local energy distribution of the signal at different frequency scales a and different time locations. By analyzing z(a, b), signal features can be extracted.

[0160] In some embodiments, since the results of wavelet transform and analysis effects are directly affected by the selected wavelet function, the method may select different wavelet functions in different application environments, such as Morlet wavelet, Haar wavelet, etc.

[0161] For example, when selecting a suitable wavelet function, the following strategies can be considered:

[0162] (1) Theoretical matching: Select the wavelet function related to the specific research problem based on its characteristics and requirements. For example, the Morlet wavelet performs well in processing periodic signals, while the Daubechies wavelet is suitable for analyzing signals with local singularities.

[0163] (2) Frequency characteristics consideration: When selecting a wavelet function, the frequency characteristics of the signal must be considered. If the signal contains specific frequency components, a wavelet function that can capture these components should be selected.

[0164] (3) Time domain resolution assessment: The resolution of the wavelet function in the time domain is also an important factor in the selection. Some wavelet functions have high time domain resolution and can capture subtle changes in the signal; while other wavelet functions have lower time domain resolution and are more suitable for capturing the overall structure of the signal.

[0165] (4) Frequency domain resolution analysis: Similarly, the resolution of the wavelet function in the frequency domain also needs to be considered. Some wavelet functions have a narrow frequency range and are suitable for analyzing local frequency characteristics, while others have a wide frequency range and are suitable for analyzing the overall frequency distribution.

[0166] (5) Computational efficiency considerations: Computational efficiency is also a factor that cannot be ignored when selecting a wavelet function. Some wavelet functions have concise mathematical forms and efficient computational methods, making them suitable for large-scale data analysis and real-time processing scenarios.

[0167] In summary, the selection of wavelet function requires comprehensive consideration of factors such as signal characteristics, frequency distribution, time domain and frequency domain resolution, and computational efficiency. At the same time, the best choice can be made by combining theoretical analysis and experimental verification.

[0168] For example, the formula of Morlet wavelet is:

[0169]

[0170] Among them, w0 is the frequency parameter, which is used to control the frequency of the wavelet function;

[0171] i is the imaginary unit;

[0172] π is the ratio of a circle to its circumference.

[0173] The method is described using Morlet wavelet as an example, but is not limited to Morlet wavelet.

[0174] In these embodiments, the method can flexibly select wavelet functions according to the operating environment of the industrial equipment, thereby more accurately performing wavelet transform on the signal components, so that the obtained wavelet coefficients can more effectively reflect the signal characteristics, thereby improving the pertinence and accuracy of fault feature extraction based on the Internet of Things.

[0175] In order to accurately obtain the frequency scale, in some embodiments, the frequency scale of the wavelet coefficients is determined based on the similarity objective function, including:

[0176] Substitute the wavelet coefficients and signal components into the similarity objective function to obtain the frequency scale that maximizes the similarity objective function result; the frequency scale is the frequency scale of the wavelet coefficients.

[0177] Exemplarily, there are n signal components in each selected time frame, and the similarity between the wavelet coefficient w and the corresponding signal component c is calculated by the following objective function d.

[0178]

[0179] Where c is the component signal, e is its wavelet coefficient, and n is the number of data.

[0180] The first objective function mentioned above adopts the form of weighted covariance, which is mainly used to measure the weighted sum of the deviation products of two series after removing the mean. It is mainly suitable for highlighting specific points (through weight w j ), or scenarios that are sensitive to overall trends and deviations in the data.

[0181] The second objective function mentioned above uses the Pearson correlation coefficient, which is primarily used to measure the linear correlation between two sequences. It is primarily suitable for scenarios that require a concise and standardized similarity measure or focus on the overall linear relationship between sequences.

[0182] Based on the above, the method can find the frequency scale a that maximizes the objective function d. The objective function indicates the similarity of the signals, not only in amplitude but also in geometry. The higher the exponent of the objective function, the greater the similarity.

[0183] In these embodiments, the method can automatically determine the optimal frequency scale of the wavelet coefficients through a similarity objective function, thereby facilitating more accurately revealing fault features in the signal.

[0184] In order to effectively integrate information at different frequency scales, in some embodiments, calculations are performed based on all frequency scales to obtain fault characteristic parameters corresponding to the time frame signal, including:

[0185] Aggregate calculations are performed on multiple frequency scales corresponding to each time frame signal to obtain multiple fault characteristic parameters corresponding to the multiple time frame signals.

[0186] For example, various algorithms (such as arithmetic mean algorithm, weighted mean algorithm, majority algorithm, etc.) can be used to calculate the scale of the original signal using the scales of the various components. This method is illustrated using a simple arithmetic mean algorithm.

[0187] Among them, this method performs a simple arithmetic average on the frequency scale a of each component and mixes the initial features obtained from the signal components through the following formula:

[0188]

[0189] Among them, s represents the mixed scale corresponding to the time frame signal, that is, the fault characteristic parameter;

[0190] a i is the frequency scale of the ith signal component.

[0191] In this way, this method can obtain m fault characteristic parameters of x(t).

[0192] In these embodiments, the method can effectively integrate information at different frequency scales by performing aggregate calculations on multiple frequency scales corresponding to each time frame signal, thereby obtaining more representative and stable fault characteristic parameters.

[0193] In order to achieve standardized and efficient transmission, in some embodiments, the method further includes:

[0194] Convert multiple fault characteristic parameters into JSON format data.

[0195] Upload JSON format data to the cloud platform via the MQTT protocol, so that the cloud platform can perform fault analysis on industrial equipment based on multiple fault characteristic parameters.

[0196] In some embodiments, JSON (JavaScript Object Notation) is a lightweight data exchange format that is easy for humans to read and write, as well as for machines to parse and generate. It is based on a subset of ECMAScript (the JS specification developed by the European Computer Society) and uses a completely language-independent text format to store and represent data. Simply put, JSON is a data format, not a programming language. While it has a structure similar to programming languages, it is untyped and exceptionally concise.

[0197] JSON is built on two structures: objects (consisting of key-value pairs) and arrays (consisting of ordered lists of values).

[0198] Characteristics of JSON:

[0199] (1) Easy to read: JSON uses a syntax similar to JavaScript objects, which makes it very easy for humans to read;

[0200] (2) Easy to write: JSON has a very concise structure and is easy to write and modify;

[0201] (3) Easy to parse: Most programming languages support JSON and provide methods to directly parse JSON data;

[0202] (4) Lightweight: Compared with XML, JSON is lighter and has no complex tags. It only uses simple brackets and commas to separate data.

[0203] The basic syntax of JSON is:

[0204] Object: enclosed by curly braces {}, key-value pairs are separated by commas, and keys and values are separated by colons; keys must be strings (enclosed in double quotes);

[0205] Array: enclosed by square brackets [], and the values in the array are separated by commas;

[0206] Value: Can be a string (enclosed in double quotes), a number, an object, an array, a Boolean value (true or false), or null.

[0207] For example, in order to design a JSON that describes the structure of devices, components, and sensors, the method uses nested objects and arrays to represent the hierarchical relationship.

[0208] The following is a sample JSON that describes an industrial device (industrial robot). The device contains multiple parts, each part has multiple sensors, and each sensor has a scale parameter (i.e., fault signature parameter):

[0209]

[0210]

[0211] In this JSON structure:

[0212] deviceName and deviceID are used to identify the device;

[0213] parts is an array containing all the parts in the device; each part has a partName and partID for identification;

[0214] sensors is an array containing all sensors in the component; each sensor has a sensorName and sensorID for identification;

[0215] scale represents the scale parameter of the sensor.

[0216] In these embodiments, the method can be easily extended through the structure to include more devices, components or sensors, and each level can add additional properties to describe more information.

[0217] In some embodiments, the method uses MQTT to transmit data between the cloud platform and the edge. The MQTT (Message Queuing Telemetry Transport) protocol is widely used in the Industrial Internet of Things (IIoT). It is a lightweight publish / subscribe messaging protocol designed for low-bandwidth, unreliable, or unstable network environments, and is particularly suitable for industrial scenarios requiring remote monitoring, data collection, and device control. Sensors, robots, and various devices on production lines can all upload data to the cloud or control center in real time using the MQTT protocol.

[0218] The QoS (Quality of Service) level in the MQTT protocol defines the reliability of message delivery in the MQTT protocol. The MQTT protocol provides three QoS levels to meet the needs of different application scenarios:

[0219] (1) QoS 0 (at most once):

[0220] In this mode, after a message is sent, no confirmation is required from the receiver.

[0221] The sender will not retry message delivery. If the message is lost due to network abnormalities or other reasons, the message will not be received.

[0222] This mode is suitable for non-critical messages, such as high-frequency transmission of environmental sensor data, where data loss is acceptable.

[0223] (2) QoS1 (at least once):

[0224] In this mode, a sent message reaches the receiver at least once.

[0225] The sender will wait for the receiver's confirmation (PUBACK). If the sender does not receive PUBACK within a certain period of time, the message will be resent until PUBACK is received.

[0226] The receiver handles the duplication of messages by itself (e.g., detecting duplication through message ID).

[0227] (3) QoS2 (ensure only once):

[0228] In this mode, the message sent is guaranteed to reach the recipient and arrives only once.

[0229] A four-way handshake (PUBLISH->PUBREC->PUBREL->PUBCOMP) is performed between the sender and the receiver to ensure the unique delivery of the message.

[0230] If the confirmation message of a step in the process is lost or times out, the corresponding party will retry the corresponding step until the process is completed.

[0231] This mode is suitable for scenarios where the integrity and uniqueness of message delivery are very high.

[0232] This method uses MQTT QoS2 to transmit scale information, thereby improving transmission reliability.

[0233] In some embodiments, after the fault characteristic parameters arrive at the cloud, they will enter the queue, and then a rule engine will be used to process the data in the queue, and finally a time series database will be used to store the data.

[0234] This approach uses a time series database instead of a traditional relational database. In the IoT scenario, the comparison between time series databases and relational databases is as follows:

[0235] (1) Design concept and storage structure

[0236] A. Relational database

[0237] Design concept: Based on the relational model, using row storage.

[0238] Storage structure: Each record contains multiple fields, supporting complex data relationships and diverse data types.

[0239] B. Time Series Database

[0240] Design concept: specially designed for processing time series data.

[0241] Storage structure: Using columnar storage, each record usually contains a timestamp, measurement value, and label.

[0242] (2) Performance optimization and storage efficiency

[0243] A. Relational database

[0244] Index structures such as B+ trees optimize query performance but increase write and storage overhead.

[0245] The vertical expansion capability is strong, and improving the performance of a single machine can temporarily alleviate the pressure caused by data growth, but it will increase hardware and maintenance costs.

[0246] B. Time Series Database

[0247] Improve storage efficiency and reduce hardware resource consumption through time slicing and data compression technology.

[0248] Many time series databases support horizontal scaling, which can increase the system's throughput and storage capacity by adding nodes.

[0249] High concurrent write performance optimizes real-time writing of large amounts of data.

[0250] (3) Query and data processing capabilities

[0251] A. Relational database

[0252] The read workload is mainly row-oriented queries, which optimizes the fast retrieval of single records.

[0253] The concurrent write rate is relatively low, which is suitable for transactional operations and ensures data consistency and integrity.

[0254] It is highly versatile, supports standard SQL query language, and is flexible and comprehensive.

[0255] B. Time Series Database

[0256] It is specially optimized for queries on time series data and supports fast aggregation and statistical analysis.

[0257] It provides a query language specifically for time series data queries, supporting complex time series analysis.

[0258] In these embodiments, this method uses a time series database to process time series data, such as sensor data and device status data, enabling efficient storage and query of time series data. The fault signature parameters in this method are extracted from time frame signals, which inherently possess temporal properties; at the same time, the fault signature parameters need to be efficiently stored and queried. Therefore, this method is more suitable for using a time series database to store scale parameters.

[0259] In these embodiments, the method can convert multiple fault characteristic parameters into JSON format data, and efficiently upload them to the cloud platform through the MQTT protocol, thereby realizing standardized data transmission and cloud-based fault analysis, thereby ensuring the fault diagnosis capability of industrial equipment in the Internet of Things environment.

[0260] In order to make the purpose, technical solutions and advantages of this application clearer, the technical solutions in this application will be described clearly and completely below. In some embodiments, Figure 3 As shown, the fault feature extraction method based on the Internet of Things includes:

[0261] S201. Sample the vibration signal sent by the vibration sensor on the industrial equipment to obtain a discrete signal sequence.

[0262] S202: Divide the discrete signal sequence into multiple time frame signals.

[0263] S203: Calculate the local maximum of the time frame signal and interpolate the local maximum to obtain an upper envelope sequence.

[0264] S204: Calculate the local minimum value of the time frame signal and interpolate the local minimum value to obtain a lower envelope sequence.

[0265] S205: Calculate the mean sequence of the upper envelope sequence and the lower envelope sequence.

[0266] S206 , determining whether the difference between the time frame signal and the mean sequence meets the component decomposition condition, if so, executing step S207 ; if not, executing step S203 .

[0267] S207: Determine the difference as a signal component.

[0268] S208. Calculate the difference between the time frame signal and the signal component to obtain a residual value.

[0269] S209 , determining whether the residual value is a monotonic function or has only one extreme value, if so, executing step S210 ; if not, executing step S203 .

[0270] S210: Determine all obtained signal components into multiple signal components that meet component decomposition conditions.

[0271] S211. Wavelet function matching based on the working environment of industrial equipment.

[0272] S212. Perform wavelet transform on the signal components based on the wavelet function to obtain wavelet coefficients.

[0273] S213. Substitute the wavelet coefficients and the signal components into the similarity objective function to obtain a frequency scale that maximizes the result of the similarity objective function; the frequency scale is the frequency scale of the wavelet coefficients.

[0274] S214 , performing aggregation calculation on multiple frequency scales corresponding to each time frame signal to obtain multiple fault characteristic parameters corresponding to the multiple time frame signals.

[0275] S215. Convert the format of multiple fault characteristic parameters to obtain JSON format data.

[0276] S216. Upload JSON format data to the cloud platform through the MQTT protocol, so that the cloud platform can perform fault analysis on the industrial equipment based on multiple fault characteristic parameters.

[0277] Figure 4 The schematic diagram of the structure of a fault feature extraction device based on the Internet of Things is shown. It should be understood that the device is Figure 3 The method executed in the embodiment corresponds to the embodiment, and the steps involved in the aforementioned method can be executed. The specific functions and effects of the device can be found in the description above. To avoid repetition, detailed description is appropriately omitted here.

[0278] The fault feature extraction device based on the Internet of Things includes:

[0279] The sampling unit 310 is used to sample the vibration signal sent by the vibration sensor on the industrial equipment to obtain a discrete signal sequence;

[0280] A division unit 320, configured to divide the discrete signal sequence into a plurality of time frame signals;

[0281] A component decomposition unit 330 is used to decompose the time frame signal into components to obtain multiple signal components;

[0282] A wavelet transform unit 340 is used to perform wavelet transform on the signal components to obtain wavelet coefficients;

[0283] A frequency scale calculation unit 350 is used to determine the frequency scale of the wavelet coefficients based on the similarity objective function;

[0284] The fault feature calculation unit 360 is configured to perform calculations based on all frequency scales to obtain fault feature parameters corresponding to the time frame signal.

[0285] In some embodiments, the component decomposition unit 330 is specifically configured to perform component decomposition on the time frame signal to obtain a plurality of signal components that meet the component decomposition conditions;

[0286] The component decomposition conditions include:

[0287] The number of extreme points and the number of zero crossings are equal or differ by one; and

[0288] At any point, the mean value of the envelope defined by the local maximum and local minimum is zero.

[0289] In some embodiments, the component decomposition unit 330 includes:

[0290] The calculation subunit 331 is used to calculate the local maximum of the time frame signal and interpolate the local maximum to obtain an upper envelope sequence;

[0291] The calculation subunit 331 is further used to calculate the local minimum value of the time frame signal and interpolate the local minimum value to obtain the lower envelope sequence;

[0292] The calculation subunit 331 is further used to calculate the mean sequence of the upper envelope sequence and the lower envelope sequence;

[0293] A judging subunit 332 is used to judge whether the difference between the time frame signal and the mean value sequence satisfies the component decomposition condition;

[0294] The determining subunit 333 is configured to determine the difference as a signal component when the difference satisfies a component decomposition condition.

[0295] In some embodiments, the calculation subunit 331 is further used to use the difference as a time frame signal when the difference does not meet the component decomposition condition, calculate the local maximum of the time frame signal, and interpolate the local maximum to obtain an upper envelope sequence.

[0296] In some embodiments, the component decomposition unit 330 includes:

[0297] The calculation subunit 331 is further configured to calculate the difference between the time frame signal and the signal component to obtain a residual value;

[0298] The judging subunit 332 is further used to judge whether the residual value is a monotonic function or has only one extreme value;

[0299] The calculation subunit 331 is further configured to, when the residual value is not a monotonic function or has more than one extreme value, use the residual value as a time frame signal, calculate a local maximum value of the time frame signal, and interpolate the local maximum value to obtain an upper envelope sequence;

[0300] The determination subunit 333 is further configured to determine all obtained signal components into a plurality of signal components that meet a component decomposition condition when the residual value is a monotonic function or has only one extreme value.

[0301] In some embodiments, the wavelet transform unit 340 includes:

[0302] A matching subunit 341 is used to match the wavelet function based on the operating environment of the industrial equipment;

[0303] The transform subunit 342 is configured to perform wavelet transform on the signal components based on the wavelet function to obtain wavelet coefficients.

[0304] In some embodiments, the frequency scale calculation unit 350 is specifically configured to substitute the wavelet coefficients and the signal components into the similarity objective function to obtain a frequency scale that maximizes the similarity objective function result; the frequency scale is the frequency scale of the wavelet coefficients.

[0305] In some embodiments, the fault feature calculation unit 360 is specifically configured to perform aggregate calculation on multiple frequency scales corresponding to each time frame signal to obtain multiple fault feature parameters corresponding to the multiple time frame signals.

[0306] In some embodiments, the fault feature extraction device based on the Internet of Things further includes:

[0307] The conversion unit 370 is used to convert the format of the multiple fault characteristic parameters to obtain JSON format data;

[0308] The uploading unit 380 is used to upload JSON format data to the cloud platform through the MQTT protocol, so that the cloud platform can perform fault analysis on the industrial equipment based on multiple fault characteristic parameters.

[0309] like Figure 5 As shown, the present application provides an electronic device 300, which includes a processor 301 and a memory 302. The processor 301 and the memory 302 are interconnected and communicate with each other through a communication bus 303 and / or other forms of connection mechanisms (not marked). The memory 302 stores a computer program executable by the processor 301. When the computing device is running, the processor 301 executes the computer program to perform the method in any of the aforementioned optional implementations.

[0310] The present application provides a computer-readable storage medium, wherein a computer program is stored in the computer-readable storage medium. When the computer program is executed by a processor, the method in any of the aforementioned optional implementations is executed.

[0311] Among them, the computer-readable storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk or optical disk.

[0312] The present application provides a computer program product, which includes computer programmability. When the computer program is executed by a processor, the method in any of the aforementioned optional implementations is executed.

[0313] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some or all of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the scope of the technical solutions of the embodiments of the present application, and they should all be included in the scope of the claims and specification of the present application. In particular, as long as there is no conflict, the various technical features mentioned in the various embodiments can be combined in any way. The present application is not limited to the specific embodiments disclosed herein, but includes all technical solutions that fall within the scope of the claims.

Claims

1. A fault feature extraction method based on the Internet of Things, characterized in that: include: Sampling the vibration signal sent by the vibration sensor on the industrial equipment to obtain a discrete signal sequence; Dividing the discrete signal sequence into a plurality of time frame signals; Decomposing the time frame signal into components to obtain multiple signal components; Performing wavelet transform on the signal components to obtain wavelet coefficients; determining a frequency scale of the wavelet coefficients based on a similarity objective function; Calculations are performed based on all frequency scales to obtain fault characteristic parameters corresponding to the time frame signal.

2. The method for extracting fault features based on the Internet of Things according to claim 1, characterized in that: The component decomposition of the time frame signal to obtain multiple signal components includes: Decomposing the time frame signal into components to obtain a plurality of signal components that meet the component decomposition conditions; The component decomposition conditions include: The number of extreme points and the number of zero crossings are equal or differ by one; and At any point, the mean value of the envelope defined by the local maximum and local minimum is zero.

3. The method for extracting fault features based on the Internet of Things according to claim 2, characterized in that: The component decomposition of the time frame signal to obtain a plurality of signal components that meet the component decomposition conditions includes: Calculating the local maximum of the time frame signal and interpolating the local maximum to obtain an upper envelope sequence; Calculating a local minimum value of the time frame signal and interpolating the local minimum value to obtain a lower envelope sequence; Calculating a mean sequence of the upper envelope sequence and the lower envelope sequence; determining whether a difference between the time frame signal and the mean value sequence satisfies the component decomposition condition; If the difference satisfies the component decomposition condition, the difference is determined as a signal component.

4. The method for extracting fault features based on the Internet of Things according to claim 3, characterized in that: The method further comprises: If the difference does not meet the component decomposition condition, the difference is used as the time frame signal, and the steps of calculating the local maximum of the time frame signal and interpolating the local maximum to obtain the upper envelope sequence are re-executed.

5. The method for extracting fault features based on the Internet of Things according to claim 3, characterized in that: The method further comprises: calculating a difference between the time frame signal and the signal component to obtain a residual value; Determining whether the residual value is a monotonic function or has only one extreme value; If the residual value is not a monotonic function or has more than one extreme value, then taking the residual value as the time frame signal, re-performing the step of calculating the local maximum value of the time frame signal, and interpolating the local maximum value to obtain the upper envelope sequence; If the residual value is a monotonic function or has only one extreme value, all the obtained signal components are determined to be a plurality of signal components that meet the component decomposition condition.

6. The method for extracting fault features based on the Internet of Things according to claim 1, characterized in that: The step of performing wavelet transform on the signal components to obtain wavelet coefficients includes: matching a wavelet function based on the operating environment of the industrial equipment; Performing wavelet transform on the signal component based on the wavelet function to obtain wavelet coefficients.

7. The method for extracting fault features based on the Internet of Things according to claim 1, characterized in that: The determining of the frequency scale of the wavelet coefficients based on the similarity objective function includes: Substituting the wavelet coefficients and the signal components into a similarity objective function to obtain a frequency scale that maximizes the result of the similarity objective function; The frequency scale is the frequency scale of the wavelet coefficients.

8. The method for extracting fault features based on the Internet of Things according to claim 1, characterized in that: The calculation based on all frequency scales to obtain the fault characteristic parameters corresponding to the time frame signal includes: Aggregate calculations are performed on multiple frequency scales corresponding to each time frame signal to obtain multiple fault characteristic parameters corresponding to the multiple time frame signals.

9. The method for extracting fault features based on the Internet of Things according to claim 1, characterized in that: The method further comprises: Convert the multiple fault characteristic parameters into JSON format. The JSON format data is uploaded to the cloud platform via the MQTT protocol, so that the cloud platform performs fault analysis on the industrial equipment based on the multiple fault characteristic parameters.

10. A fault feature extraction device based on the Internet of Things, characterized in that: include: A sampling unit, used to sample the vibration signal sent by the vibration sensor on the industrial equipment to obtain a discrete signal sequence; a dividing unit, configured to divide the discrete signal sequence into a plurality of time frame signals; A component decomposition unit, configured to decompose the time frame signal into components to obtain a plurality of signal components; A wavelet transform unit, configured to perform wavelet transform on the signal components to obtain wavelet coefficients; a frequency scale calculation unit, configured to determine the frequency scale of the wavelet coefficients based on a similarity objective function; The fault characteristic calculation unit is used to perform calculations based on all frequency scales to obtain fault characteristic parameters corresponding to the time frame signal.

Images