Method and system for detecting faults in dynamic equipment

Abstract

A method and system for fault detection in dynamic equipment, belonging to the field of fault detection technology, includes the steps of collecting detection signals of a target equipment (S10), classifying the detection signals into fault types based on their autocorrelation characteristics (S20), selecting a corresponding fault diagnosis model based on the classification results (S30), and performing fault diagnosis based on the selected fault diagnosis model and the detection signals to obtain a diagnosis result (S40). Adaptively selecting a diagnosis model based on the fault type classification results ensures that the selected diagnosis model is a fault detection model that is adapted to the current signal feature rules, thereby realizing a differentiated solution for fault diagnosis in dynamic equipment and improving the accuracy of the fault diagnosis results.

Description

[Technical Field]

[0001] The present invention relates to the field of fault detection technology, and in particular to a fault detection method for dynamic equipment and a fault detection system for dynamic equipment. [Background technology]

[0002] Dynamic equipment is widely used in enterprise production and is the core power equipment of enterprise production, such as motors and gears. Since the stability of these equipment is directly related to the stability of the production process, detecting the operating status of dynamic equipment is essential during production. Traditional dynamic equipment fault detection techniques mainly involve manual patrol inspections to collect vibration status signals from the dynamic equipment, and then performing fault detection based on the collected vibration signals. This method often involves pre-training a model, then training the collected feature signals as input parameters for the model, and obtaining fault diagnosis results based on the training results. However, as is well known, dynamic equipment varies greatly in operating status, operating structure, and operating fluids depending on different tasks. Therefore, the differences in corresponding fault feature signals for different fault types are very significant. Therefore, using a fixed single model to extract features from various signals with large differences and then performing fault diagnosis based on the extracted features makes it difficult to ensure the accuracy of the diagnosis results. To address the low accuracy of traditional dynamic equipment fault detection methods, a new dynamic equipment fault detection method is needed. Summary of the Invention [Problem to be solved by the invention]

[0003] SUMMARY OF THE INVENTION Embodiments of the present invention aim to at least solve the problem of low accuracy of conventional dynamic equipment fault detection schemes by providing a method and system for dynamic equipment fault detection. [Means for solving the problem]

[0004] To achieve the above object, a first aspect of the present invention provides a fault detection method for dynamic equipment, the method including the steps of collecting detection signals of target equipment, classifying the detection signals into fault types based on autocorrelation characteristics, selecting a corresponding fault diagnosis model based on the classification result, and performing fault diagnosis based on the selected fault diagnosis model and the detection signals to obtain a diagnosis result.

[0005] Optionally, the detection signals are acoustic signals and / or vibration signals, and operation process parameters of the target equipment, wherein the operation process parameters include one or more of operating temperature information, pressure information, flow rate information, and fastening method information.

[0006] Optionally, said dynamic facility is a facility in which, when activated, a moving member is present and which, when activated, generates a vibration signal and / or an acoustic signal.

[0007] Optionally, said fault type comprises a mechanical fault and / or a hydrodynamic fault.

[0008] Optionally, the method further includes steps of identifying an equipment type of the target equipment, determining detection signal characteristics of the target equipment based on the equipment type, determining filtering rules based on the detection signal characteristics, and collecting detection signals of the target equipment based on the filtering rules.

[0009] Optionally, said equipment types include mechanically dynamic equipment, including reciprocating dynamic equipment and centrifugal dynamic equipment, and pneumatically dynamic equipment, including gaseous medium equipment and liquid medium equipment.

[0010] Optionally, the step of classifying the fault type of the detection signal based on the autocorrelation characteristics includes the steps of obtaining an irregularity of the detection signal based on the autocorrelation of the detection signal, and determining a fault type of the equipment from which the current detection signal occurs based on the irregularity, and selecting a corresponding fault type diagnosis model based on the determined fault type of the equipment.

[0011] Optionally, the step of obtaining the irregularity of the detection signal based on the autocorrelation of the detection signal includes the steps of taking the detection signal for a certain time, performing time shift processing on the detection signal of the segment according to a preset time shift, and obtaining a corresponding time-shifted signal; constructing an autocorrelation function based on the detection signal of the segment and the corresponding time-shifted signal; calculating the entropy of the autocorrelation function; and representing the irregularity of the detection signal of the segment by the entropy of the autocorrelation function of the detection signal of the segment, wherein the smaller the entropy of the autocorrelation function, the greater the irregularity.

[0012] Optionally, the formula for the autocorrelation function is:

number

number

[0013] Optionally, the step of determining the fault type of the equipment in which each detection signal occurs based on the irregularity includes the steps of: comparing the entropy of the autocorrelation function of the detection signal of one segment with a predetermined entropy threshold; if the entropy of the autocorrelation function is less than or equal to the predetermined entropy threshold, determining that the fault type of the equipment in which the detection signal of the segment occurs is a mechanical fault type; and if the entropy of the autocorrelation function is greater than the predetermined entropy threshold, determining that the fault type of the equipment in which the detection signal of the segment occurs is a fluid dynamics anomaly type.

[0014] Optionally, when the current fault type diagnosis model is a mechanical fault diagnosis model, the step of performing feature extraction on the detection signal and obtaining feature information includes a step of performing a fusion analysis of a singular spectrum and a higher-order spectrum on the detection signal, obtaining post-analysis data, and using it as feature information; when the current fault type diagnosis model is a fluid dynamics fault diagnosis model, the step of performing feature extraction on the detection signal and obtaining feature information includes a step of performing a short-time Fourier transform process on the detection signal, obtaining a corresponding spectral image, and using it as feature information.

[0015] Optionally, the step of training a selected fault diagnosis model based on the feature information and identifying faulty equipment based on the training result includes the step of: using the feature information and operation process parameters as input parameters to train based on the machine fault diagnosis model, and using the obtained training result as the analysis result of the machine fault type; or using the feature information and operation process parameters as input parameters to train the fluid mechanics fault diagnosis model, and using the obtained training result as the analysis result of the fluid mechanics anomaly type, and during the training process of the fluid mechanics anomaly identification model, the training process is interfered with based on an image data augmentation method.

[0016] Optionally, the method further includes the steps of: within each audio frequency band, obtaining sound field distribution map information for each frequency band, comparing the sound field distribution map information for each frequency band with preset standard sound field distribution map information to obtain a sound field distribution map deviation degree for each frequency band; obtaining a total deviation degree matrix based on the sound field distribution map deviation degree for each frequency band; training a preset sound field fault identification model using the total deviation degree matrix as an input parameter to obtain an equipment abnormality result in a current operating state; and comparing the equipment abnormality result obtained based on the sound field fault identification model with the equipment abnormality result obtained based on a fault diagnosis model, and verifying the equipment abnormality result obtained by the current fault diagnosis model.

[0017] A second aspect of the present invention provides a fault detection apparatus for dynamic equipment, including: a signal collection module used for collecting detection signals of target equipment; a signal classification module used for classifying the detection signals into fault types based on autocorrelation characteristics; a model selection module used for selecting a corresponding fault diagnosis model based on the classification result; and a fault diagnosis module used for performing fault diagnosis based on the selected fault diagnosis model and the detection signals and obtaining a diagnosis result.

[0018] Optionally, the detection signals are acoustic signals and / or vibration signals, and operation process parameters of the target equipment, wherein the operation process parameters include one or more of operating temperature information, pressure information, flow rate information, and fastening method information.

[0019] Optionally, said dynamic facility is a facility in which, when activated, a moving member is present and which, when activated, generates a vibration signal and / or an acoustic signal.

[0020] Optionally, the signal classification module is further used for identifying an equipment type of the target equipment, determining detection signal characteristics of the target equipment based on the equipment type, determining filtering rules based on the detection signal characteristics, and collecting detection signals of the target equipment based on the filtering rules.

[0021] Optionally, the device further includes a verification module for, within each audio frequency band, obtaining sound field distribution map information for each frequency band, comparing the sound field distribution map information for each frequency band with preset standard sound field distribution map information to obtain sound field distribution map deviation degrees for each frequency band, obtaining a total deviation degree matrix based on the sound field distribution map deviation degrees for each frequency band, training a preset sound field fault identification model using the total deviation degree matrix as an input parameter to obtain equipment abnormality results in the current working state, and comparing the equipment abnormality results obtained based on the sound field fault identification model with the equipment abnormality results obtained based on a fault diagnosis model to verify the equipment abnormality results obtained by the current fault diagnosis model.

[0022] A third aspect of the present invention provides a fault detection system for dynamic equipment, the system including the above-described fault detection device for dynamic equipment.

[0023] A fourth aspect of the present invention provides a computer-readable storage medium having instructions stored thereon, the instructions, when run on a computer, causing the computer to perform the above-described method for detecting faults in dynamic equipment.

[0024] According to the above technical solution, the solution of the present invention fully considers the signal feature differences of different fault types and distinguishes between fault types with significant differences based on the autocorrelation characteristics that are most easily distinguishable, thereby realizing a technical solution for distinguishing between fault types based on signal features and avoiding the problem of being unable to distinguish between complex equipment fault types that exists in traditional solutions for distinguishing between fault types based on fault types. Adaptively selecting a diagnostic model based on the fault type distinction result ensures that the selected diagnostic model is a fault detection model that is adapted to the current signal feature rules, thereby realizing a differentiated dynamic equipment fault diagnosis solution and improving the accuracy of the fault diagnosis results.

[0025] Other features and advantages of the embodiments of the present invention are described in detail in the following specific embodiments. [Brief explanation of the drawings]

[0026] The drawings are included to provide a further understanding of embodiments of the invention, constitute a part of the specification, and, together with the following specific embodiments, are intended to explain embodiments of the invention but are not intended to be limitations thereon.

[0027] [Figure 1] 2 is a flowchart of a fault detection method for dynamic equipment according to an embodiment of the present invention. [Figure 2] 1 is a system configuration diagram of a fault detection system for dynamic equipment according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0028]

[0023] Specific embodiments of the present invention will be described in detail below with reference to the drawings. It should be understood that the specific embodiments described herein are for the purpose of explaining and interpreting the present invention, and are not intended to limit the present invention.

[0029] Dynamic equipment is widely used in enterprise production and is the core power equipment of enterprise production, such as motors and gears. Since the stability of these equipment is directly related to the stability of the production process, detecting the operating status of dynamic equipment is essential during production. Traditional dynamic equipment fault detection techniques mainly involve manual patrol inspections to collect vibration status signals from the dynamic equipment, and then performing fault detection on the dynamic equipment based on the collected vibration signals. This method often involves pre-training a model, then training the collected feature signals as input parameters for the model, and obtaining fault diagnosis results based on the training results. However, as is well known, dynamic equipment varies greatly in operating status, operating structure, and operating fluids depending on different tasks. Therefore, the differences in corresponding fault feature signals for different fault types are very significant. Therefore, using a fixed single model to extract features from various signals with large differences and then performing fault diagnosis based on the extracted features makes it difficult to ensure the accuracy of the diagnosis results.

[0030] In response to the problem of low accuracy of conventional dynamic equipment fault detection methods, the solution of the present invention proposes a new dynamic equipment fault detection method, which fully considers the signal feature differences of different fault types and distinguishes between fault types with significant differences based on the autocorrelation characteristics that are most easily distinguishable, thereby realizing a technical solution for distinguishing fault types based on signal features and avoiding the problem of conventional solutions for distinguishing fault types based on fault types that are unable to distinguish between complex equipment fault types. Adaptive diagnosis model selection is performed based on the fault type distinction results, ensuring that the selected diagnosis model is a fault detection model that adapts to the current signal feature rules, thereby realizing a differentiated dynamic equipment fault diagnosis solution and improving the accuracy of fault diagnosis results.

[0031] 1 is a flowchart of a fault detection method for dynamic equipment according to an embodiment of the present invention. As shown in FIG. 1, an embodiment of the present invention provides a fault detection method for dynamic equipment, which includes the following steps:

[0032] Step S10: Collect the detection signals of the target equipment.

[0033] Specifically, the solution of the present invention aims to realize fault detection based on vibration signals or acoustic signals. Since both acoustic signals and vibration signals are waveform signals and there are differences between the two in both the normal operating state and the abnormal operating state of dynamic equipment, the present invention can be applied to fault detection using acoustic signals, and similarly, can be applied to fault detection using vibration signals.

[0034] Preferably, the detection signals are acoustic signals and / or vibration signals, and operation process parameters of the target equipment, and the operation process parameters include one or more of operating temperature information, pressure information, flow rate information, and fixing method information.

[0035] In an embodiment of the present invention, the solution of the present invention further collects the operation process parameters of the equipment in addition to the required feature signals. For example, if a liquid storage tank is present in a dynamic equipment, the liquid storage amount in the liquid storage tank will cause differences in the equipment signals. When performing fault diagnosis, it is necessary to avoid misidentifying the signal differences caused by such differences in the equipment process state as feature differences in the fault signal. Based on this, the solution of the present invention must use the equipment process parameters as a reference quantity when building a model and performing subsequent diagnostic training, thereby ensuring the accuracy of the identification results.

[0036] In addition, the solution of the present invention can theoretically be applied to any signal collection method, and whether it is a signal collection method using traditional manual patrol inspections or a signal collection method using fixed or mobile sensors, as long as it can accurately collect the signal characteristics of the target equipment, it should theoretically fall within the protection scope of the solution of the present invention.

[0037] Step S20: Classify the fault type of the detection signal based on the autocorrelation characteristics.

[0038] Preferably, the dynamic installation is an installation in which a moving member is present when activated, and which generates a vibration signal and / or an acoustic signal when activated.

[0039] Specifically, dynamic equipment refers to process equipment in petrochemical production equipment whose main operating components are moving. Generally, it can be classified into fluid transport machinery, heterogeneous separation machinery, stirring and mixing machinery, refrigeration machinery, crystallization and drying equipment, etc. Specifically, it includes centrifugal compressors, rotary compressors, piston compressors, flue turbines, turbines, blowers, centrifugal pumps, reciprocating pumps, gear pumps, centrifugal pumps, screw pumps, filters, centrifuges, pulverizers, rotary kilns, agitators, rotary dryers, and other fluid transport machinery. The solution of the present invention can be applied to any equipment that generates vibration signals and / or acoustic signals during operation, but is not limited to specific equipment. To better explain the technical solution of the present invention, mechanical faults and fluid dynamic faults will be described below, but the present invention is not limited to these two fault types.

[0040] Preferably, this includes mechanical and / or hydrodynamic failures.

[0041] Specifically, there are two main types of dynamic equipment failures. The first is mechanical failure, such as general bearing damage, shaft misalignment, shaft imbalance, and loose pump foundations or machinery. These mechanical failures result from damage or changes to the structural properties of dynamic equipment, causing changes in friction or vibration, resulting in primarily mechanical vibration and friction noise. The second is fluid power abnormalities. Mechanical and fluid power transmissions are common in production sites, and the fluid may be the product flow. If a fault, such as cavitation or turbulence, exists in the fluid flow path, the corresponding sound will also be generated. Frequent faults such as cavitation can cause abnormal pump body vibration, inlet / outlet pressure pulsations, and noise in centrifugal pumps, affecting their performance, service life, and safety.

[0042] The dynamic equipment is mainly rolling bearings. The acoustic signal characteristics of rolling bearings are severe signal modulation, weak fault signals, low signal-to-noise ratios, wide signal frequency bands, and strong signal periodicity. In the noise radiated by rotation, the peak frequency of the generated noise is often a multiple of the rotation speed of the centrifugal pump, and the signal exhibits periodicity. On the other hand, the acoustic signal radiated during abnormal fluid dynamics operation is random. There are significant differences in the characteristics of the fault noise between these two types, and the solution of the present invention is to identify the fault type based on these differences and apply corresponding acoustic signal analysis methods to different fault types.

[0043] Preferably, the mechanical fault type includes any of bearing damage, shaft misalignment, shaft unbalance, loose pump foundation and loose machinery, and the fluid dynamics anomaly type includes any of cavitation and turbulence.

[0044] Preferably, the method further includes the steps of identifying an equipment type of the target equipment, determining detection signal characteristics of the target equipment based on the equipment type, determining filtering rules based on the detection signal characteristics, and collecting detection signals of the target equipment based on the filtering rules.

[0045] In the solution of the present invention, differences in equipment types will cause differences in signals. For example, in the case of a single motor and a motor-combined equipment combined with a storage tank, there will be differences in the signal characteristics of the two signals. Based on this difference, the solution of the present invention will apply a corresponding filtering collection method to filter out most of the noise signals that do not belong to the signal characteristic frequency band of the target equipment.

[0046] Preferably, the equipment types include mechanically dynamic equipment, including reciprocating dynamic equipment and centrifugal dynamic equipment, and pneumatically dynamic equipment, including gaseous medium equipment and liquid medium equipment.

[0047] Preferably, the step of classifying the fault type of the detection signal based on the autocorrelation characteristic includes the steps of obtaining an irregularity of the detection signal based on the autocorrelation of the detection signal, and determining a fault type of the equipment from which the current detection signal occurs based on the irregularity, and selecting a corresponding fault type diagnosis model based on the determined fault type of the equipment.

[0048] Specifically, the step of acquiring the irregularity of the detection signal based on the autocorrelation of the detection signal includes the steps of taking the detection signal for a certain time, performing time shift processing on the detection signal of the segment according to a preset time shift, and acquiring a corresponding time-shifted signal; constructing an autocorrelation function based on the detection signal of the segment and the corresponding time-shifted signal; calculating the entropy of the autocorrelation function; and expressing the irregularity of the detection signal of the segment by the entropy of the autocorrelation function of the detection signal of the segment, wherein the smaller the entropy of the autocorrelation function, the greater the irregularity.

[0049] Specifically, the formula for the autocorrelation function is:

number

number

[0050] Example 1

[0051] The autocorrelation of an audio signal is the cross-correlation of a signal with itself at different time points. The corresponding autocorrelation function is a function of the similarity between two observations relative to their time difference. It is a mathematical tool used to find repetitive patterns (e.g., periodic signals buried in noise) or identify missing fundamental frequencies among harmonic frequencies of a signal. It is often used in signal processing to analyze functions or sequences, such as time-domain signals. The solution of the present invention utilizes the autocorrelation function to determine whether regularity exists in an audio signal. Specifically, an audio signal of a certain time period is taken, and a time-shift process is performed on the audio signal of the certain segment based on a preset time shift to obtain a corresponding time-shifted signal. An autocorrelation function is constructed based on the audio signal of the certain segment and the corresponding time-shifted signal, and the entropy of the autocorrelation function is calculated. The entropy of the autocorrelation function represents the irregularity of the corresponding audio signal. The larger the entropy of the autocorrelation function, the greater the irregularity of the corresponding audio signal. The essence of entropy is the "degree of internal disorder" of a system. In the field of signal processing, it can indicate the degree of irregularity of a signal. The smaller the entropy value, the greater the corresponding irregularity, and the less likely the signal is to be a regular signal. The entropy of the autocorrelation function can be calculated to represent the irregularity of a signal. The formula for the autocorrelation function is as follows:

number

number

[0052] Based on historical data statistics and the empirical judgment of personnel, an entropy value threshold can be used to label the equipment. For the labeling criteria, entropy values ​​below the threshold are often fluid dynamics anomalies, while entropy values ​​above the threshold are often mechanical faults. Based on this, the entropy of the autocorrelation function of the acoustic signal is compared with a preset entropy threshold. If the entropy of the autocorrelation function is below the preset entropy threshold, the equipment fault type from which the current acoustic signal occurs is a mechanical fault, including bearing damage, shaft misalignment, shaft imbalance, loose pump foundations, and loose machinery. If the entropy of the autocorrelation function is greater than the preset entropy threshold, the equipment fault type from which the current acoustic signal occurs is a fluid dynamics anomaly, including cavitation and turbulence.

[0053] Step S30: Select a corresponding fault diagnosis model based on the classification result.

[0054] Specifically, after the signal type distinction is completed, an adaptive fault diagnosis model can be selected based on the signal distinction result.

[0055] Example 2

[0056] Preferably, a machine fault identification model needs to be pre-trained to meet usage needs and directly invoke the corresponding fault identification model during subsequent fault identification. First, a large amount of historical operating data is sorted through and labeled as machine faults. Then, the corresponding machine fault types and historical audio collection signals are obtained for each data. Of course, the corresponding audio feature signals are recorded at the production site where the audio monitoring-based fault identification method was used. If no audio monitoring signals existed in the past, equipment with machine faults is labeled. Then, the labeled equipment is operated based on each machine fault type to collect audio feature information. The collected audio feature information is subjected to a fusion analysis process of singular spectrum and higher-order spectrum to extract the corresponding features. Singular spectrum analysis is a powerful method for studying nonlinear time series data that has emerged in recent years. A trajectory matrix is ​​constructed based on the observed time series, and then the trajectory matrix is ​​decomposed and reconstructed to extract signals representing different components of the original time series, such as long-term trend signals, periodic signals, and noise signals, which can be used to analyze the structure of the time series and further predict it. On the other hand, higher-order spectrum is a kind of representation form of higher-order statistics, which is a broad definition of statistical methods in which the order of statistical features is greater than 2. As a nonlinear signal processing tool, higher-order statistics can reflect the nonlinear relationship of higher-order correlation.

[0057] In the embodiment of the present invention, the effect of singular spectrum analysis mainly lies in two aspects: spectral line smoothness and the degree of decay. Compared with the method of feature extraction using only singular spectra, the feature extraction method using singular spectra fused with higher-order spectra has better spectral smoothness, faster decay, and more obvious decay phenomenon under the same embedding dimensionality (under large noise conditions), that is, it can better reflect the dimensionality features. Based on this, the feature extraction solution fused with higher-order spectra and singular spectra has better feature extraction effect and higher robustness.

[0058] After processing by fusion analysis of the singular spectrum and the higher-order spectrum, the intensity and phase features in the frequency domain are extracted from the high-order spectrum data after noise removal. Then, the intensity and phase features are used as input parameters and the corresponding fault type is used as a guide to output a machine fault type identification model. The larger the training sample, the more consistent the model obtained by the corresponding training will be. After the model training is completed, a portion of the unused training sample data is selected for model testing to determine whether the model training results meet the requirements. If they do meet the requirements, the model obtained by the training can be used as the machine fault identification model to be used subsequently.

[0059] Example 3

[0060] Furthermore, for fluid dynamics anomalies, the acoustic signals have high randomness, making the processing and analysis of fluid noise relatively complex, and it is difficult to achieve accurate identification through manual feature extraction. To better extract the features of fluid dynamics anomalies, the solution of the present invention proposes a method for processing fluid noise based on a spectral image visual model. A spectral diagram can be simply understood as a frequency distribution diagram, and a complex signal is primarily a signal containing different frequency components. Simply put, the Fourier transformation is used to organize the disorder, and then a specific representation is based on the spectral diagram. This is because the collected acoustic signal is a complex signal, and a complex signal actually contains many signal frequencies rather than a single one.

[0061] Preferably, the solution of the present invention uses short-time Fourier transform to process the acoustic signal. Short-time Fourier transform is a versatile tool used in audio signal processing, defining a very useful time and frequency distribution class and specifying the complex amplitude along the time and frequency variations of any signal. In practice, the calculation process of short-time Fourier transform involves dividing a long time signal into shorter segments of the same length and calculating the Fourier transform, i.e., the Fourier spectrum, for each shorter segment. Since constructing a spectral image model based on short-time Fourier transform is a conventional method, its implementation process will not be repeated here. After processing the acoustic signal into a spectral signal, the spectral signal is used as the input parameter of a corresponding fault identification model. The fault identification model is constructed based on two-dimensional convolution, the construction method of which is similar to the construction method of a machine fault type identification model. In both cases, a large amount of historical data and conventional labeling data are collected to train the model. Of course, the corresponding training sample data also needs to be processed into a corresponding spectral image pattern.

[0062] Preferably, during the training process of the fault identification model, interference is directly added to the frequency domain information by using the image data extension method, and processing such as frequency shifting, warping, and shielding is performed on the spectral image at a certain time step, forcing the identification model to have a fine-grained representation, improving the model's processing ability for interference information, and improving the model's generalization performance.

[0063] Step S40: A fault diagnosis is performed based on the selected fault diagnosis model and the detection signal, and a diagnosis result is obtained.

[0064] Specifically, the entropy of the autocorrelation function of the detection signal of one segment is compared with a predetermined entropy threshold, and if the entropy of the autocorrelation function is equal to or less than the predetermined entropy threshold, the fault type of the equipment generating the detection signal of that segment is a mechanical fault type, and if the entropy of the autocorrelation function is greater than the predetermined entropy threshold, the fault type of the equipment generating the detection signal of that segment is a fluid dynamics anomaly type. When the current fault type diagnostic model is a mechanical fault diagnostic model, the step of extracting features from the detection signal and obtaining feature information includes a step of performing a fusion analysis of a singular spectrum and a higher-order spectrum on the detection signal, obtaining post-analysis data, and setting the analyzed data as feature information. When the current fault type diagnostic model is a fluid dynamics fault diagnostic model, the step of extracting features from the detection signal and obtaining feature information includes a step of performing a short-time Fourier transform on the detection signal, obtaining a corresponding spectral image, and setting the analyzed data as feature information.

[0065] Specifically, the step of training the selected fault diagnosis model based on the feature information and identifying faulty equipment based on the training result includes the step of using the feature information and operation process parameters as input parameters to train based on the machine fault diagnosis model and use the acquired training result as the analysis result of the machine fault type, or the step of using the feature information and operation process parameters as input parameters to train the fluid mechanics fault diagnosis model and use the acquired training result as the analysis result of the fluid mechanics anomaly type, and during the training process of the fluid mechanics anomaly identification model, the training process is interfered with based on the image data augmentation method.

[0066] Example 4

[0067] Within each audio frequency band, sound field distribution map information for each frequency band is obtained, the sound field distribution map information for each frequency band is compared with preset standard sound field distribution map information, the sound field distribution map deviation degree for each frequency band is obtained, a total deviation degree matrix is ​​obtained based on the sound field distribution map deviation degree for each frequency band, the total deviation degree matrix is ​​used as an input parameter to train a preset sound field fault identification model, obtain equipment abnormality results in the current operating state, compare the equipment abnormality results obtained based on the sound field fault identification model with the equipment abnormality results obtained based on the fault diagnosis model, and verify the equipment abnormality results obtained by the current fault diagnosis model.

[0068] Specifically, whether the identification result is a mechanical fault type or a fluid dynamics abnormality type, the information on the fault type caused by the voice can be ultimately estimated based on the corresponding voice feature information, and the information on the fault type can be directly pushed to the user terminal to help the user identify and investigate the corresponding fault.

[0069] Preferably, in order to further ensure the accuracy of the fault identification results, the solution of the present invention further proposes a fault diagnosis result verification method, which includes: obtaining sound field distribution map information for each audio frequency band, comparing the sound field distribution map information for each frequency band with preset standard sound field distribution map information, obtaining the sound field distribution map deviation degree for each frequency band, obtaining a total deviation degree matrix based on the sound field distribution map deviation degrees for each frequency band, training a preset sound field fault identification model using the total deviation degree matrix as an input parameter, obtaining equipment abnormality results in the current operating state, comparing the equipment abnormality results obtained based on the sound field fault identification model with the equipment abnormality results obtained based on the fault diagnosis model, and verifying the equipment abnormality results obtained by the current fault diagnosis model.

[0070] Specifically, because sound propagation is overlapping, noise in a complex on-site environment is often the result of the overlap of multiple sound sources, and changes in the spectral characteristics of the sound emitted by the sound source, changes in the spatial distribution of the sound source, and increases or decreases in the number of sound sources will all cause changes in the sound waveform collected at a fixed detection point. Based on this, the verification method proposed in this invention uses acoustic imaging technology to collect noise audio data and sound field video data of the object being measured, and through frequency band discrimination and judgment, comprehensively determine the sound spectral characteristics and distribution state within the entire frequency range, thereby achieving more comprehensive, accurate, and intuitive discrimination and visualization of various types of abnormal operations.

[0071] Specifically, the fault identification results of the comparison form must be obtained first, and then compared based on the two fault identification results. If the difference between the two meets the requirement, it indicates that the fault identification result meets the requirement. If the difference between the two exceeds the requirement, it indicates that there is a recognition error in a certain form, and fault identification must be attempted again to ensure recognition accuracy.

[0072] When obtaining the comparison results, the corresponding acoustic field fault identification model must first be constructed. When the on-site equipment is in normal operating mode, the acoustic imaging equipment is used to collect and record video and audio signals, and a sample library of normal operating mode is established. The steps are as follows:

[0073] First, select the total frequency range [Φmin, Φmax] for sound analysis, where Φmax is less than half the microphone sampling frequency. Then, select N groups of characteristic frequency ranges based on the noise frequency characteristics of the target abnormal operating state. Divide the total frequency range [Φmin, Φmax] into N groups, each with a frequency range of [Φ1, Φ2], [Φ2, Φ3]...[ΦN-1, ΦN], where Φ1 = Φmin and ΦN = Φmax. Next, define the sound frequency range [Φ1, Φ2] for the first group, and calculate a sound field distribution map. Record the coordinate γ1[x1, y1] of the maximum sound source point. Then, adjust the sound frequency ranges for the second, third, and Nth groups. Repeat the above steps to obtain the sound source point coordinates γ2[x2, y2], γ3[x3, y3]...γN[xN, yN] for each frequency band. The sound energy signal s(t) of a microphone in the microphone array is collected and stored, and the acoustic signal is Fourier transformed to obtain the spectral density s(w) within the frequency range [Φmin, Φmax]. The total spectral density is filtered according to the previously set N groups of frequency ranges, and the N groups of characteristic spectral densities are calculated.

number

number

[0074] After obtaining the sample library of normal operating modes, in the case of abnormal operating modes, or when using a simulation experimental device to conduct a simulation experiment of abnormal operating states, the acoustic imaging equipment is used to collect and record video and audio signals, and the abnormal operating state identification model is trained. The steps are as follows:

[0075] The first group of spectral density sequences s1(w) in the speech analysis frequency range [Φ1, Φ2] is selected and input to the trained unsupervised discriminant model.

number

number

number

number

number

number

number

[0076] The total deviation degree matrix is ​​taken as input, the type number of the abnormal operating state is taken as output, and the abnormal operating state identification model is established through training.

[0077] After obtaining the sound field fault identification model, the total deviation degree matrix is ​​used as an input parameter to train the preset sound field fault identification model, and obtain the equipment abnormality result in the current operating state. The equipment abnormality result obtained based on the sound field fault identification model is compared with the equipment abnormality result obtained based on the fault diagnosis model, and the equipment abnormality result obtained by the current fault diagnosis model is verified. After the consistency test between the two is passed, the final fault diagnosis result is determined.

[0078] Example 5

[0079] Various fault types were artificially labeled, including bearing wear, misalignment, imbalance, machine looseness, shaft misalignment, loose foundation, cavitation, and turbulence. For the same rotating equipment, only one fault type was labeled at a time, and 200 groups of data were collected for each fault type, divided by time and background. Finally, each fault type had 200 groups of data, for a total of 1,600 groups of data, each of which represented an audio or vibration signal. As will be further explained, the 1,600 groups of data included data collected from multiple pre-defined points, each at a different distance and orientation from the rotating equipment. These data served as comparative samples to form multiple comparative examples.

[0080] Comparative Example 1

[0081] The fault type of each group of data is hidden, and all data is randomly combined to obtain a sequence of data groups with unknown fault types and confusing sequence numbers. Fault detection is performed based on a fault diagnosis model based on the vibration signals obtained through simulation, and a diagnosed fault type is obtained for each group of data. Then, the corresponding data displays the hidden fault type, and the two are compared. The data groups with the same fault type are 1420 groups.

[0082] Comparative Example 2

[0083] The fault type of each group of data is hidden, and all data are randomly combined to obtain a sequence of data groups with unknown fault types and confusing sequence numbers. Fault detection is performed based on a fault diagnosis model obtained through simulation that distinguishes fault types based on acoustic signal features, and a diagnosed fault type is obtained for each group of data. Then, the corresponding data displays the hidden fault type, and the two are compared. There are 1532 data groups where the two are the same.

[0084] Comparative Example 3

[0085] The fault type of each group of data is hidden, and all data is randomly combined to obtain a sequence of data groups with unknown fault types and confusing sequence numbers. Fault detection is performed based on a fault diagnosis model obtained through simulation that distinguishes fault types based on acoustic signal features, and the diagnosed fault type of each group of data is obtained. The detection results are corrected based on the deviation degree of the sound field distribution map in each frequency band. The corrected detection results are compared with the original results, and the data groups where the two are the same are 1551 groups.

[0086] 2 is a block diagram of a fault detection device for dynamic equipment according to an embodiment of the present invention. As shown in FIG. 2, an embodiment of the present invention provides a fault detection device for dynamic equipment, which includes a signal collection module, a signal classification module, a model selection module, and a fault diagnosis module.

[0087] The signal collecting module is used to collect the detection signal of the target equipment.

[0088] Specifically, the solution of the present invention aims to realize fault detection based on vibration signals or acoustic signals. Since both acoustic signals and vibration signals are waveform signals and there are differences between the two in both the normal operating state and the abnormal operating state of dynamic equipment, the present invention can be applied to fault detection using acoustic signals, and similarly, can be applied to fault detection using vibration signals.

[0089] Preferably, the detection signals are acoustic signals and / or vibration signals, and operation process parameters of the target equipment, and the operation process parameters include one or more of operating temperature information, pressure information, flow rate information, and fixing method information.

[0090] In an embodiment of the present invention, the solution of the present invention further collects the operation process parameters of the equipment in addition to the required feature signals. For example, if a dynamic equipment has a liquid storage tank, the liquid storage amount in the liquid storage tank will cause differences in the equipment signals. When performing fault diagnosis, it is necessary to avoid misidentifying the signal differences caused by such differences in the equipment process state as feature differences in the fault signals. Based on this, the solution of the present invention must use the equipment process parameters as a reference quantity when building a model and performing subsequent diagnosis training, thereby ensuring the accuracy of the identification results.

[0091] In addition, the solution of the present invention can theoretically be applied to any signal collection method, and whether it is a signal collection method using traditional manual patrol inspections or a signal collection method using fixed or mobile sensors, as long as it can accurately collect the signal characteristics of the target equipment, it should theoretically fall within the protection scope of the solution of the present invention.

[0092] The signal classification module is used to perform fault type classification on the detection signal based on autocorrelation characteristics.

[0093] Specifically, this includes mechanical and hydrodynamic failures.

[0094] Specifically, there are two main types of dynamic equipment failures. The first is mechanical failure, such as general bearing damage, shaft misalignment, shaft imbalance, and loose pump foundations or machinery. These mechanical failures result from damage or changes to the structural properties of dynamic equipment, causing changes in friction or vibration, resulting in primarily mechanical vibration and friction noise. The second is fluid power abnormalities. Mechanical and fluid power transmissions are common in production sites, and fluids may be the product flow. If there are faults in the fluid flow path, such as cavitation or turbulence, the corresponding sound will also be generated. Frequent faults such as cavitation can cause abnormal pump body vibration, inlet and outlet pressure pulsations, and noise in centrifugal pumps, affecting their performance, service life, and safety.

[0095] The dynamic equipment is mainly rolling bearings. The acoustic signal characteristics of rolling bearings are poor signal modulation, weak fault signals, low signal-to-noise ratios, wide signal frequency bands, and strong signal periodicity. In the noise radiated by rotation, the peak frequency of the generated noise is often a multiple of the rotation speed of the centrifugal pump, and the signal exhibits periodicity. On the other hand, the acoustic signal radiated during abnormal fluid dynamics operation is random. There are significant differences in the characteristics of the fault noise between these two types. The solution of the present invention is to identify the fault type based on these differences and apply corresponding acoustic signal analysis methods to different fault types.

[0096] Preferably, the mechanical fault type includes any of bearing damage, shaft misalignment, shaft unbalance, loose pump foundation and loose machinery, and the fluid dynamics anomaly type includes any of cavitation and turbulence.

[0097] Preferably, the method further includes the steps of identifying an equipment type of the target equipment, determining detection signal characteristics of the target equipment based on the equipment type, determining filtering rules based on the detection signal characteristics, and collecting detection signals of the target equipment based on the filtering rules.

[0098] In the solution of the present invention, differences in equipment types will cause differences in signals. For example, in the case of a single motor and a motor-combined equipment combined with a storage tank, there will be differences in the signal characteristics of the two signals. Based on this difference, the solution of the present invention will apply a corresponding filtering collection method to filter out most of the noise signals that do not belong to the signal characteristic frequency band of the target equipment.

[0099] Preferably, the equipment types include mechanically dynamic equipment, including reciprocating dynamic equipment and centrifugal dynamic equipment, and pneumatically dynamic equipment, including gaseous medium equipment and liquid medium equipment.

[0100] Preferably, the step of classifying the fault type of the detection signal based on the autocorrelation characteristic includes the steps of obtaining an irregularity of the detection signal based on the autocorrelation of the detection signal, and determining a fault type of the equipment from which the current detection signal occurs based on the irregularity, and selecting a corresponding fault type diagnosis model based on the determined fault type of the equipment.

[0101] Specifically, the step of acquiring the irregularity of the detection signal based on the autocorrelation of the detection signal includes the steps of taking the detection signal for a certain time, performing time shift processing on the detection signal of the segment according to a preset time shift, and acquiring a corresponding time-shifted signal; constructing an autocorrelation function based on the detection signal of the segment and the corresponding time-shifted signal; calculating the entropy of the autocorrelation function; and expressing the irregularity of the detection signal of the segment by the entropy of the autocorrelation function of the detection signal of the segment, wherein the smaller the entropy of the autocorrelation function, the greater the irregularity.

[0102] Specifically, the formula for the autocorrelation function is:

number

number

[0103] The autocorrelation of an audio signal is the cross-correlation of a signal with itself at different time points. The corresponding autocorrelation function is a function of the similarity between two observations relative to their time difference. It is a mathematical tool used to find repetitive patterns (e.g., periodic signals buried in noise) or identify missing fundamental frequencies among harmonic frequencies of a signal. It is often used in signal processing to analyze functions or sequences, such as time-domain signals. The solution of the present invention utilizes the autocorrelation function to determine whether regularity exists in an audio signal. Specifically, an audio signal of a certain time period is taken, and a time-shift process is performed on the audio signal of the certain segment based on a preset time shift to obtain a corresponding time-shifted signal. An autocorrelation function is constructed based on the audio signal of the certain segment and the corresponding time-shifted signal, and the entropy of the autocorrelation function is calculated. The entropy of the autocorrelation function represents the irregularity of the corresponding audio signal. The larger the entropy of the autocorrelation function, the greater the irregularity of the corresponding audio signal. The essence of entropy is the "degree of internal disorder" of a system. In the field of signal processing, it can indicate the degree of irregularity of a signal. The smaller the entropy value, the greater the corresponding irregularity, and the less likely the signal is to be a regular signal. The entropy of the autocorrelation function can be calculated to represent the irregularity of a signal. The formula for the autocorrelation function is as follows:

number

number

[0104] Based on historical data statistics and the empirical judgment of personnel, an entropy value threshold can be used to label the equipment. For the labeling criteria, entropy values ​​below the threshold are often fluid dynamics anomalies, while entropy values ​​above the threshold are often mechanical faults. Based on this, the entropy of the autocorrelation function of the acoustic signal is compared with a preset entropy threshold. If the entropy of the autocorrelation function is below the preset entropy threshold, the equipment fault type from which the current acoustic signal occurs is a mechanical fault, including bearing damage, shaft misalignment, shaft imbalance, loose pump foundations, and loose machinery. If the entropy of the autocorrelation function is greater than the preset entropy threshold, the equipment fault type from which the current acoustic signal occurs is a fluid dynamics anomaly, including cavitation and turbulence.

[0105] The model selection module is used to select a corresponding fault diagnosis model based on the classification result.

[0106] Specifically, after the signal type distinction is completed, an adaptive fault diagnosis model can be selected based on the signal distinction result.

[0107] Preferably, a machine fault identification model needs to be pre-trained to meet usage needs and directly invoke the corresponding fault identification model during subsequent fault identification. First, a large amount of historical operating data is sorted through and labeled as machine faults. Then, the corresponding machine fault types and historical audio signals are obtained for each data. Of course, the corresponding audio feature signals are recorded at the production site where the audio monitoring-based fault identification method was used. If no audio monitoring signals existed in the past, equipment with machine faults is labeled. Then, the labeled equipment is operated based on each machine fault type to collect audio feature information. The collected audio feature information is subjected to a fusion analysis process of singular spectrum and higher-order spectrum to extract the corresponding features. Singular spectrum analysis is a powerful method for studying nonlinear time series data that has emerged in recent years. A trajectory matrix is ​​constructed based on the observed time series, and then the trajectory matrix is ​​decomposed and reconstructed to extract signals representing different components of the original time series, such as long-term trend signals, periodic signals, and noise signals, which can be used to analyze the structure of the time series and further predict it. On the other hand, higher-order spectrum is a kind of representation form of higher-order statistics, which is a broad definition of statistical methods in which the order of statistical features is greater than 2. As a nonlinear signal processing tool, higher-order statistics can reflect the nonlinear relationship of higher-order correlation.

[0108] In the embodiment of the present invention, the effect of singular spectrum analysis mainly lies in two aspects: spectral line smoothness and the degree of decay. Compared with the method of feature extraction using only singular spectra, the feature extraction method using singular spectra fused with higher-order spectra has better spectral smoothness, faster decay, and more obvious decay phenomenon under the same embedding dimensionality (under large noise conditions), that is, it can better reflect the dimensionality features. Based on this, the feature extraction solution fused with higher-order spectra and singular spectra has better feature extraction effect and higher robustness.

[0109] After processing by fusion analysis of the singular spectrum and the higher-order spectrum, the intensity and phase features in the frequency domain are extracted from the high-order spectrum data after noise removal. Then, the intensity and phase features are used as input parameters and the corresponding fault type is used as a guide to output a machine fault type identification model. The larger the training sample, the more consistent the model obtained by the corresponding training will be. After the model training is completed, a portion of the unused training sample data is selected for model testing to determine whether the model training results meet the requirements. If they do meet the requirements, the model obtained by the training can be used as the machine fault identification model to be used subsequently.

[0110] Furthermore, for fluid dynamics anomalies, the acoustic signals have high randomness, making the processing and analysis of fluid noise relatively complex, and it is difficult to achieve accurate identification through manual feature extraction. To better extract the features of fluid dynamics anomalies, the solution of the present invention proposes a method for processing fluid noise based on a spectral image visual model. A spectral diagram can be simply understood as a frequency distribution diagram, and a complex signal is primarily a signal containing different frequency components. Simply put, the Fourier transformation is used to organize the disorder, and then a specific representation is based on the spectral diagram. This is because the collected acoustic signal is a complex signal, and a complex signal actually contains many signal frequencies rather than a single one.

[0111] Preferably, the solution of the present invention uses short-time Fourier transform to process the acoustic signal. Short-time Fourier transform is a versatile tool used in audio signal processing, defining a very useful time and frequency distribution class and specifying the complex amplitude along the time and frequency variations of any signal. In practice, the calculation process of short-time Fourier transform involves dividing a long time signal into shorter segments of the same length and calculating the Fourier transform, i.e., the Fourier spectrum, for each shorter segment. Since constructing a spectral image model based on short-time Fourier transform is a conventional method, its implementation process will not be repeated here. After processing the acoustic signal into a spectral signal, the spectral signal is used as the input parameter of a corresponding fault identification model. The fault identification model is constructed based on two-dimensional convolution, the construction method of which is similar to the construction method of a machine fault type identification model. In both cases, a large amount of historical data and conventional labeling data are collected to train the model. Of course, the corresponding training sample data also needs to be processed into a corresponding spectral image pattern.

[0112] Preferably, during the training process of the fault identification model, interference is directly added to the frequency domain information by using the image data extension method, and processing such as frequency shifting, warping, and shielding is performed on the spectral image at a certain time step, forcing the identification model to have a fine-grained representation, improving the model's processing ability for interference information, and improving the model's generalization performance.

[0113] The fault diagnosis module is used to perform fault diagnosis based on the selected fault diagnosis model and the detection signal, and obtain a diagnosis result.

[0114] Specifically, the entropy of the autocorrelation function of the detection signal of one segment is compared with a predetermined entropy threshold, and if the entropy of the autocorrelation function is equal to or less than the predetermined entropy threshold, the fault type of the equipment generating the detection signal of that segment is a mechanical fault type, and if the entropy of the autocorrelation function is greater than the predetermined entropy threshold, the fault type of the equipment generating the detection signal of that segment is a fluid dynamics anomaly type. When the current fault type diagnostic model is a mechanical fault diagnostic model, the step of extracting features from the detection signal and obtaining feature information includes a step of performing a fusion analysis of a singular spectrum and a higher-order spectrum on the detection signal, obtaining post-analysis data, and setting the analyzed data as feature information. When the current fault type diagnostic model is a fluid dynamics fault diagnostic model, the step of extracting features from the detection signal and obtaining feature information includes a step of performing a short-time Fourier transform on the detection signal, obtaining a corresponding spectral image, and setting the analyzed data as feature information.

[0115] Specifically, the step of training the selected fault diagnosis model based on the feature information and identifying faulty equipment based on the training result includes the step of using the feature information and operation process parameters as input parameters to train based on the machine fault diagnosis model and use the acquired training result as the analysis result of the machine fault type, or the step of using the feature information and operation process parameters as input parameters to train the fluid mechanics fault diagnosis model and use the acquired training result as the analysis result of the fluid mechanics anomaly type, and during the training process of the fluid mechanics anomaly identification model, the training process is interfered with based on the image data augmentation method.

[0116] Within each audio frequency band, sound field distribution map information for each frequency band is obtained, the sound field distribution map information for each frequency band is compared with preset standard sound field distribution map information, the sound field distribution map deviation degree for each frequency band is obtained, a total deviation degree matrix is ​​obtained based on the sound field distribution map deviation degree for each frequency band, the total deviation degree matrix is ​​used as an input parameter to train a preset sound field fault identification model, obtain equipment abnormality results in the current operating state, compare the equipment abnormality results obtained based on the sound field fault identification model with the equipment abnormality results obtained based on the fault diagnosis model, and verify the equipment abnormality results obtained by the current fault diagnosis model.

[0117] Specifically, whether the identification result is a mechanical fault type or a fluid dynamics abnormality type, the information on the fault type caused by the voice can be ultimately estimated based on the corresponding voice feature information, and the information on the fault type can be directly pushed to the user terminal to help the user identify and investigate the corresponding fault.

[0118] Preferably, in order to further ensure the accuracy of the fault identification results, the solution of the present invention further proposes a fault diagnosis result verification method, which includes: obtaining sound field distribution map information for each audio frequency band, comparing the sound field distribution map information for each frequency band with preset standard sound field distribution map information, obtaining the sound field distribution map deviation degree for each frequency band, obtaining a total deviation degree matrix based on the sound field distribution map deviation degrees for each frequency band, training a preset sound field fault identification model using the total deviation degree matrix as an input parameter, obtaining equipment abnormality results in the current operating state, comparing the equipment abnormality results obtained based on the sound field fault identification model with the equipment abnormality results obtained based on the fault diagnosis model, and verifying the equipment abnormality results obtained by the current fault diagnosis model.

[0119] Specifically, because sound propagation is overlapping, noise in a complex on-site environment is often the result of the overlap of multiple sound sources, and changes in the spectral characteristics of the sound emitted by the sound source, changes in the spatial distribution of the sound source, and increases or decreases in the number of sound sources will all cause changes in the sound waveform collected at a fixed detection point. Based on this, the verification method proposed in this invention uses acoustic imaging technology to collect noise audio data and sound field video data of the object being measured, and through frequency band discrimination and judgment, comprehensively determine the sound spectral characteristics and distribution state within the entire frequency range, thereby achieving more comprehensive, accurate, and intuitive discrimination and visualization of various types of abnormal operations.

[0120] Specifically, the fault identification results of the comparison form must be obtained first, and then compared based on the two fault identification results. If the difference between the two meets the requirement, it indicates that the fault identification result meets the requirement. If the difference between the two exceeds the requirement, it indicates that there is a recognition error in a certain form, and fault identification must be attempted again to ensure recognition accuracy.

[0121] The solution of the present invention further proposes a fault detection system for dynamic equipment, the system including the above-mentioned fault detection device for dynamic equipment.

[0122] An embodiment of the present invention further provides a computer-readable storage medium having instructions stored thereon, the instructions, when run on a computer, causing the computer to perform the above-described method for detecting faults in dynamic equipment.

[0123] As can be understood by those skilled in the art, all or part of the steps in the methods of the above embodiments can be achieved by instructing relevant hardware through a program, which is stored in a storage medium and includes a plurality of instructions for causing a one-chip microcomputer, chip, or processor to perform all or part of the steps of the methods of the above embodiments. The storage medium includes various media capable of storing program code, such as a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0124] Although the above describes in detail the optional embodiments of the present invention with reference to the drawings, the embodiments of the present invention are not limited to the details in the above embodiments. Various simple modifications can be made to the technical solutions of the embodiments of the present invention within the technical concept of the embodiments of the present invention, and all of these simple modifications fall within the scope of protection of the embodiments of the present invention. It should be noted that the specific technical features described in the above specific embodiments can be combined in any appropriate manner if not contradictory. In order to avoid unnecessary repetition, the embodiments of the present invention will not separately describe various possible combinations.

[0125] Furthermore, various different embodiments of the present invention can be combined in any manner, and as long as the combination does not deviate from the technical idea of ​​the embodiments of the present invention, the combination should also be considered to be the content disclosed in the embodiments of the present invention.

Claims

1. 1. A method for fault detection in dynamic equipment, the method comprising: collecting detection signals of the target equipment; classifying the detection signal into a fault type based on an autocorrelation characteristic; selecting a corresponding fault diagnosis model based on the classification result; performing a fault diagnosis based on the selected fault diagnosis model and the detection signal, and obtaining a diagnosis result.

2. the detection signals are acoustic signals and / or vibration signals and operating process parameters of the target equipment; 2. The method of claim 1, wherein the operating process parameters include one or more of operating temperature information, pressure information, flow rate information, and fixation method information.

3. 2. The method of claim 1, wherein the dynamic installation is an installation in which a moving member is present when activated, and the moving member generates a vibration signal and / or an acoustic signal when activated.

4. The failure type is:

4. The method of claim 3, including mechanical and / or hydrodynamic failures.

5. The method comprises: identifying an equipment type of the target equipment; determining a detection signal characteristic of a target facility based on the facility type; The method of claim 1 , further comprising: determining a filtering rule based on the detection signal characteristics; and collecting detection signals of target equipment based on the filtering rule.

6. The equipment type is: mechanical dynamic equipment, including reciprocating dynamic equipment and centrifugal dynamic equipment; and a pneumatic dynamic facility including a gas medium facility and a liquid medium facility.

7. The step of classifying the fault type of the detection signal based on the autocorrelation characteristic includes: obtaining an irregularity of the detection signal based on the autocorrelation of the detection signal, and determining a fault type of the equipment causing the current detection signal based on the irregularity; and selecting a corresponding fault type diagnostic model based on the determined fault type of the equipment.

8. The step of obtaining the irregularity of the detection signal based on the autocorrelation of the detection signal includes: Taking a detection signal for a certain period of time, and performing time shift processing on the detection signal of the segment according to a preset time shift to obtain a corresponding time shifted signal; constructing an autocorrelation function based on the detected signals of the segments and corresponding time-shifted signals; calculating the entropy of the autocorrelation function; 8. The method of claim 7, further comprising a step of expressing the degree of irregularity of the detected signal of said segment by the entropy of the autocorrelation function of the detected signal of said segment, wherein the smaller the entropy of said autocorrelation function, the greater the degree of irregularity.

9. The formula for the autocorrelation function is: [Equation 1] where t is the signal acquisition time, τ is the time shift, x(t) is the detected signal at time t within the current time segment; x(t-τ) is the time-shifted signal of the detected signal at time t within the current time segment; The formula for calculating the entropy of the autocorrelation function is as follows: [Equation 2] where c is a preset constant, 9. The method of claim 8, wherein H is the entropy of the correlation function.

10. The step of determining the fault type of the equipment causing each detection signal based on the irregularity degree includes: comparing the entropy of the autocorrelation function of the detected signal of one segment with a preset entropy threshold; If the entropy of the autocorrelation function is equal to or less than a predetermined entropy threshold, the fault type of the equipment in which the detection signal of the segment occurs is determined to be a machine fault type; The method according to claim 8, further comprising the step of: if the entropy of the autocorrelation function is greater than a preset entropy threshold, determining that the fault type of the equipment in which the detection signal of the segment occurs is a hydrodynamic anomaly type.

11. When the current fault type diagnosis model is a machine fault diagnosis model, the step of extracting features from the detection signal and obtaining feature information includes: The method according to claim 10, further comprising a step of performing a fusion analysis of a specific spectrum and a higher-order spectrum on the detection signal, obtaining post-analysis data, and using the data as feature information.

12. When the current fault type diagnosis model is a fluid dynamics fault diagnosis model, the step of extracting features from the detection signal and obtaining feature information includes: The method according to claim 10, further comprising the step of performing a short-time Fourier transform on the detection signal to obtain a corresponding spectral image as feature information.

13. The step of training a selected fault diagnosis model based on the feature information and identifying faulty equipment based on the training result includes: Using the feature information and the operating process parameters as input parameters, training is performed based on the machine fault diagnosis model, and the obtained training result is used as an analysis result of the machine fault type; or and performing training of the fluid dynamics fault diagnosis model using the characteristic information and the operating process parameters as input parameters, and using the acquired training result as an analysis result of the fluid dynamics abnormality type; The method according to claim 10 , wherein in the training process of the fluid dynamics anomaly identification model, the training process is interfered with based on an image data augmentation method.

14. The method comprises: further comprising training a machine fault diagnosis model, the step comprising: Sorting through the historical operating data and sorting out operating data labeled as a machine failure; Obtaining a machine fault type and history detection signal corresponding to each data; performing a fusion analysis process of the singular spectrum and the higher-order spectrum on the collected history detection signal, and extracting corresponding features to obtain intensity and phase features in the frequency domain of the higher-order spectrum data; The method of claim 13, further comprising: using the intensity and phase features in the frequency domain of the high-order spectral data as input parameters, outputting the corresponding fault types as guides, training a machine fault type identification model, and obtaining a machine fault diagnosis model.

15. The method comprises: training a fluid dynamics fault diagnosis model, the step comprising: filtering historical operating data and filtering operating data labeled as a fluid dynamics fault; obtaining a hydrodynamic fault type and history detection signal corresponding to each data point; performing a short-time Fourier transform process on the collected historical detection signal, and performing corresponding feature extraction to obtain a spectral image; 14. The method of claim 13, further comprising: taking the spectral image as an input parameter and outputting the corresponding fault type as a guide to train a fluid mechanics fault type identification model and obtain a fluid mechanics fault diagnosis model.

16. The method comprises: In each audio frequency band, obtain sound field distribution map information of each frequency band, compare the sound field distribution map information of each frequency band with preset standard sound field distribution map information, and obtain the sound field distribution map deviation degree of each frequency band; obtaining a total deviation degree matrix based on the sound field distribution map deviation degree of each frequency band; Using the total deviation degree matrix as an input parameter, training a pre-defined sound field fault identification model to obtain equipment abnormality results in the current operating state; The method according to claim 1, further comprising a step of comparing the equipment abnormality results obtained based on the sound field fault identification model with the equipment abnormality results obtained based on the fault diagnosis model, and verifying the equipment abnormality results obtained using the current fault diagnosis model.

17. 1. A fault detection apparatus for dynamic equipment, the apparatus comprising: a signal collection module for collecting the detection signal of the target device; a signal classification module used for classifying fault types for the detection signal based on autocorrelation characteristics; a model selection module used for selecting a corresponding fault diagnosis model according to the classification result; a fault diagnosis module used for performing fault diagnosis based on the selected fault diagnosis model and the detection signal and obtaining a diagnosis result.

18. the detection signals are acoustic signals and / or vibration signals and operating process parameters of the target equipment; 18. The apparatus of claim 17, wherein the operational process parameters include one or more of the following: operational temperature information, pressure information, flow rate information, and fixation method information.

19. 18. The device according to claim 17, wherein the dynamic facility is a facility in which a moving member is present when activated, and which, when activated, generates a vibration signal and / or an acoustic signal.

20. The signal classification module further comprises: Identifying an equipment type of the target equipment; determining a detection signal characteristic of the target equipment based on the equipment type; 20. The apparatus of claim 17, further comprising: determining a filtering rule based on the detection signal characteristics; and collecting detection signals of target equipment based on the filtering rule.

21. The device comprises: In each audio frequency band, obtain sound field distribution map information of each frequency band, compare the sound field distribution map information of each frequency band with preset standard sound field distribution map information, and obtain the sound field distribution map deviation degree of each frequency band; Obtaining a total deviation degree matrix based on the sound field distribution map deviation degree of each frequency band; Using the total deviation degree matrix as an input parameter, a preset sound field fault identification model is trained to obtain an equipment abnormality result in the current operating state; The apparatus according to claim 17, further comprising a verification module used for comparing the equipment abnormality result obtained based on the sound field fault identification model with the equipment abnormality result obtained based on the fault diagnosis model, and verifying the equipment abnormality result obtained by the current fault diagnosis model.

22. A dynamic equipment fault detection system, characterized in that the system includes a dynamic equipment fault detection device according to any one of claims 17 to 21.

23. A computer-readable storage medium having instructions stored thereon, the instructions causing the computer to perform the method for detecting faults in dynamic equipment according to any one of claims 1 to 16 when run on the computer.

Images