Fault intelligent diagnosis method, device and equipment and storage medium

Abstract

This invention discloses an intelligent fault diagnosis method, comprising: acquiring vibration signals of the device under test and extracting vibration feature information of the vibration signals using a data processing algorithm; processing the vibration feature information using a pre-trained machine learning classification model to determine whether a rotor shaft system fault has occurred; if so, outputting the cause of the rotor shaft system fault; otherwise, processing the vibration signals using a pre-trained deep learning classification model to determine whether a bearing fault has occurred; if so, outputting the cause of the bearing fault; otherwise, the fault diagnosis result is no fault. This invention is applicable not only to bearing-related fault diagnosis but also to rotor shaft system-related fault diagnosis; and the pre-processing machine learning classification model and post-processing deep learning classification model result in low requirements for computer resources.

Description

Technical Field

[0001] This invention relates to the field of rotating machinery fault diagnosis technology, specifically to a fault intelligent diagnosis method, device, equipment, and storage medium. Background Technology

[0002] With the development of the times, the complexity of modern mechanical equipment has increased dramatically, and the difficulty of fault diagnosis based on expert experience has also continued to rise. With the rise of machine learning, intelligent fault diagnosis methods have emerged. Intelligent fault diagnosis can identify fault characteristics that cannot be identified by expert experience, and can effectively improve the success rate of diagnosing complex faults.

[0003] Among related technologies, Chinese invention patent document CN111950377A describes an intelligent diagnosis method for rotating machinery faults based on fuzzy soft morphology image recognition. This method can convert vibration signals into three-dimensional images for processing to achieve intelligent diagnosis, but it requires a large amount of fault sample data for training and has high requirements for computing resources.

[0004] Chinese invention patent document CN114818825A describes a method for diagnosing rotating machinery faults based on a multilayer distributed residual network with convolution kernels. This method can solve the problem of feature extraction under varying operating conditions, but it is mainly applicable to bearing-related faults and is not suitable for rotor shaft system faults.

[0005] Since most existing intelligent fault diagnosis algorithms are trained on public datasets and are mostly focused on bearing fault diagnosis, they are not well adapted to rotor shaft system faults. In particular, due to the limited fault data of some specialized equipment in specific industries, it is difficult to use deep learning algorithms for targeted model training, which leads to a significant drop in the success rate of intelligent fault diagnosis.

[0006] Chinese invention patent document CN109102005A discloses a few-shot deep learning method based on shallow model knowledge transfer. This method first preprocesses the data, then transforms the original signal to different transform domains based on prior knowledge and expert experience in the relevant field, calculating artificial features. Based on these artificial features, different shallow models are selected and trained on a small amount of labeled sample data. Different shallow models with different feature combinations are screened based on indicators such as classification accuracy / prediction error, forming a candidate model pool. Then, based on this candidate model pool, a model is selected to predict unlabeled samples, obtaining predicted labels, and multiple predicted labels are fused. These predicted labels are combined with the existing small number of labeled samples to construct a training set. For a specific task, a deep neural network structure is designed and trained on the above hybrid training set. The effectiveness of this method is verified using a rotating machinery fault diagnosis dataset.

[0007] However, this scheme mainly identifies and diagnoses rotor-related faults, and its fault diagnosis adaptability is not high. In addition, it uses multiple deep learning algorithms to calculate at the same time, which places high demands on the computer. Summary of the Invention

[0008] The technical problem to be solved by this invention is how to provide a solution that is applicable to both bearing fault diagnosis and rotor shaft system fault diagnosis.

[0009] The present invention solves the above-mentioned technical problems through the following technical means:

[0010] This invention proposes an intelligent fault diagnosis method, which includes the following steps:

[0011] The vibration signal of the device under test is collected, and the vibration feature information of the vibration signal is extracted using a data processing algorithm;

[0012] The vibration feature information is processed using a pre-trained machine learning classification model to determine whether a fault has occurred in the rotor shaft system.

[0013] If so, output the cause of the rotor shaft system fault;

[0014] If not, the vibration signal is processed using a pre-trained deep learning classification model to determine whether the bearing has failed.

[0015] If so, then output the cause of the bearing failure;

[0016] If not, the fault diagnosis result is no fault.

[0017] This invention receives real-time vibration signals from the device under test and filters out the vibration feature information needed for diagnosis. A trained machine learning classification model is then used to diagnose the vibration feature information. If a fault is found, it indicates a problem with the device's rotor shaft system, and fault-related information is output. If no fault is found, the process jumps to a deep learning classification model. The deep learning classification model uses the real-time vibration signals for diagnosis; if a fault is found, it indicates a problem with the bearing, and the cause of the fault is output; otherwise, no fault is output. This invention is applicable not only to bearing-related fault diagnosis but also to rotor shaft system-related fault diagnosis; furthermore, the pre-processing machine learning classification model and post-processing deep learning classification model result in low requirements for computer resources.

[0018] Furthermore, the sampling frequency of the vibration signal is greater than or equal to the natural frequency of the bearing and rotor, and greater than or equal to twice the maximum rotational frequency of the faulty equipment.

[0019] Furthermore, the process of collecting vibration signals from the faulty equipment and extracting vibration characteristic information from the vibration signals using data processing algorithms includes:

[0020] The vibration signal is processed using a data processing algorithm to obtain the time domain and frequency domain information of the vibration signal;

[0021] The vibration feature information is extracted from the time and frequency domain information of the vibration signal. The vibration feature information includes the vibration passband value and the vibration amplitude of each harmonic.

[0022] Furthermore, the machine learning classification model employs a decision tree model, a support vector machine model, a Naive Bayes model, or a random forest model.

[0023] Furthermore, the deep learning classification model employs a feedforward neural network model, a recurrent neural network model, a convolutional neural network model, or a long short-term memory neural network model.

[0024] Furthermore, the pre-training process of the machine learning classification model includes:

[0025] Collect the fault vibration signal of the faulty equipment, establish a first fault dataset based on the fault vibration signal, and establish fault labels according to the actual fault type on site;

[0026] Extract the fault vibration feature information of the fault vibration signal, and establish a second fault dataset based on the fault vibration feature information;

[0027] The cause of the fault is determined by expert experience, and the fault type with a success rate greater than a set threshold is used as the fault label of the second fault dataset.

[0028] The machine learning classification model is trained using the second fault dataset with fault labels to obtain the pre-trained machine learning classification model.

[0029] Furthermore, the pre-training process of the deep learning classification model includes:

[0030] The deep learning classification model is trained using a public dataset and a first fault dataset with fault labels to obtain the pre-trained deep learning classification model.

[0031] Furthermore, the present invention also proposes an intelligent fault diagnosis device, the device comprising:

[0032] The acquisition module is used to acquire vibration signals from the device under test and to extract vibration characteristic information from the vibration signals using data processing algorithms.

[0033] The first judgment module is used to process the vibration feature information using a pre-trained machine learning classification model to determine whether a fault has occurred in the rotor shaft system.

[0034] The fault result output module is used to output the cause of the rotor shaft system fault when the output result of the first judgment module is yes;

[0035] The second judgment module is used to process the vibration signal using a pre-trained deep learning classification model when the output result of the first judgment module is negative, and to determine whether the bearing has failed.

[0036] The fault result output module is used to output the bearing fault cause when the output result of the second judgment module is yes; and to output the fault diagnosis result as no fault when the output result of the second judgment module is no.

[0037] Furthermore, the present invention also proposes an intelligent fault diagnosis device, the device including a memory and a processor; wherein the processor reads executable program code stored in the memory to run a program corresponding to the executable program code, so as to implement the method described above.

[0038] Furthermore, the present invention also proposes a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method described above.

[0039] The advantages of this invention are:

[0040] (1) By receiving real-time vibration signals from the device under test and filtering out the vibration feature information required for diagnosis from the vibration signals, a trained machine learning classification model is used to diagnose the vibration feature information. If a fault is found, it indicates a problem with the rotor shaft system of the device, and fault-related information is output. If no fault is found, the process jumps to the deep learning classification model. The deep learning classification model uses real-time vibration signals for diagnosis. If a fault is found, it indicates a problem with the bearing, and the cause of the fault is output. If no fault is found, no fault is output. This invention is applicable not only to bearing-related fault diagnosis but also to rotor shaft system-related fault diagnosis. Furthermore, the pre-processing machine learning classification model and the post-processing deep learning classification model have low requirements for computer resources.

[0041] (2) It combines the advantages of deep learning classification models based on public datasets for better diagnosis of bearing-related faults and machine learning classification models based on expert experience for better adaptability when there are fewer fault samples. Machine learning based on expert experience has a better diagnosis effect on common fault types, and the number of parameters can be reduced by processing with expert experience, which greatly reduces the number of fault samples required to train the machine learning model and greatly reduces the time required for intelligent fault diagnosis.

[0042] (3) In the process of training the deep learning classification model, the model is trained using the public dataset and the first fault dataset with fault labels. The public dataset is used to make up for the small sample size of the actual running data.

[0043] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0044] Figure 1 This is a flowchart illustrating an intelligent fault diagnosis method proposed in an embodiment of the present invention;

[0045] Figure 2 This is a flowchart illustrating the intelligent fault diagnosis method proposed in an embodiment of the present invention.

[0046] Figure 3 This is a schematic diagram of the model training process in one embodiment of the present invention;

[0047] Figure 4 This is a spectrum information diagram of the fault vibration signal in the first fault data set according to an embodiment of the present invention;

[0048] Figure 5 This is a schematic diagram of the structure of an intelligent fault diagnosis device proposed in an embodiment of the present invention. Detailed Implementation

[0049] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0050] like Figures 1 to 2 As shown, the first embodiment of the present invention proposes an intelligent fault diagnosis method, the method comprising the following steps:

[0051] S10. Collect the vibration signal of the device under test, and use a data processing algorithm to extract the vibration characteristic information of the vibration signal;

[0052] S20. Process the vibration feature information using a pre-trained machine learning classification model to determine whether a fault has occurred in the rotor shaft system. If so, proceed to step S30; otherwise, proceed to step S40.

[0053] S30, Cause of output rotor shaft system failure;

[0054] S40. Process the vibration signal using a pre-trained deep learning classification model to determine whether the bearing has failed. If so, proceed to step S50; otherwise, proceed to step S60.

[0055] S50, Cause of output bearing failure;

[0056] S60, the fault diagnosis result is no fault.

[0057] This embodiment receives real-time vibration signals from the device under test and filters out the vibration feature information needed for diagnosis. A trained machine learning classification model is then used to diagnose the vibration feature information. If a fault is found, it indicates a problem with the device's rotor shaft system, and fault-related information is output. If no fault is found, the system jumps to a deep learning classification model. The deep learning classification model uses the real-time vibration signals for diagnosis. If a fault is found, it indicates a problem with the bearing, and the cause of the fault is output. If no fault is found, no fault is output. This invention is applicable not only to bearing-related fault diagnosis but also to rotor shaft system-related fault diagnosis. Furthermore, the pre-processing machine learning classification model and post-processing deep learning classification model result in low requirements for computer resources.

[0058] In one embodiment, the sampling frequency of the vibration signal is greater than or equal to the natural frequency of the bearing and rotor, and greater than or equal to twice the maximum rotational frequency of the faulty equipment.

[0059] It should be noted that since the frequency of fault characteristics is often closely related to the rotor frequency, and mostly in a multiple relationship, fault information may not be collected if the sampling frequency is too low; the reason for sampling at a frequency higher than the natural frequency is that it is for faults caused by rotor resonance, and the fault frequency is the rotor's natural frequency, so it must be higher than the natural frequency.

[0060] In one embodiment, step S10: collecting vibration signals from the faulty equipment and extracting vibration feature information from the vibration signals using a data processing algorithm, specifically includes the following steps:

[0061] S11. Perform fast Fourier transform and wavelet analysis on the vibration signal to obtain the time domain and frequency domain information of the vibration signal;

[0062] S12. The vibration feature information is extracted from the time domain and frequency domain information of the vibration signal. The vibration feature information includes the vibration passband value and the vibration amplitude of each harmonic.

[0063] It should be noted that this embodiment acquires the real-time vibration signal of the device under test, and obtains the time-frequency domain information of the real-time vibration signal through fast Fourier transform and normalization processing. Then, vibration characteristic information such as vibration pass frequency value and vibration amplitude of each harmonic can be obtained from the time domain and frequency domain information, and the vibration characteristic information required for diagnosis can be selected from them. Among them, the peak-to-peak value calculated based on the vibration signal is used as the vibration pass frequency value, which can be the peak-to-peak value, half-peak value and effective value of the vibration signal.

[0064] It should be noted that the data types of the collected field vibration signals can be vibration displacement, vibration velocity, and vibration acceleration. Furthermore, vibration displacement, vibration velocity, and vibration acceleration can be converted using integration or differentiation. The appropriate data types for fault diagnosis can be found through expert experience and machine learning algorithms.

[0065] In one embodiment, the machine learning classification model employs machine learning models including, but not limited to, decision tree models, support vector machine models, naive Bayes models, or random forest models.

[0066] In one embodiment, the deep learning classification model employs deep learning models including, but not limited to, feedforward neural network models, recurrent neural network models, convolutional neural network models, or long short-term memory neural network models.

[0067] In one embodiment, such as Figure 3 As shown, the method further includes pre-training the machine learning classification model, specifically including the following steps:

[0068] (1) Collect the fault vibration signal of the faulty equipment, establish a first fault dataset based on the fault vibration signal, and establish fault labels according to the actual fault type on site;

[0069] It should be noted that the sampling frequency of the fault vibration signal must be greater than or equal to the natural frequency of the bearing and rotor, and greater than or equal to twice the maximum rotational frequency of the faulty equipment.

[0070] (2) Extract the fault vibration feature information of the fault vibration signal, and establish a second fault dataset based on the fault vibration feature information;

[0071] It should be noted that this embodiment uses data processing algorithms such as Fast Fourier Transform and Wavelet Analysis to process the vibration signal, obtain the time domain and frequency domain information of the vibration signal, and then obtain vibration characteristic information such as vibration passband value and vibration amplitude of each harmonic from the time domain and frequency domain information.

[0072] In addition, vibration characteristic information includes, but is not limited to, vibration passband value, first harmonic amplitude, first harmonic amplitude percentage, second harmonic amplitude, and second harmonic amplitude percentage. The vibration passband value can be the peak-to-peak value, half-peak value, and effective value of the vibration signal. The collected field vibration data types can be vibration displacement, vibration velocity, and vibration acceleration. Vibration displacement, vibration velocity, and vibration acceleration can be converted between each other using integration or differentiation.

[0073] (3) Use expert experience to determine the cause of the fault, and use the fault type with a success rate of judgment greater than a set threshold as the fault label of the second fault dataset;

[0074] Specifically, in this embodiment, a second fault dataset is established based on vibration characteristic information, and the cause of the fault is determined by expert experience. The results of the expert experience judgment should be verified on-site, and only fault types with a success rate of more than 90% can be used as fault classification labels for the second fault dataset.

[0075] It should be understood that the threshold value of 90% in this embodiment is only for illustrative purposes, and those skilled in the art can set other specific values ​​according to actual circumstances.

[0076] (4) The machine learning classification model is trained using the second fault dataset with fault labels to obtain the pre-trained machine learning classification model.

[0077] It should be noted that, based on experience and actual algorithm testing, it has been found that the spectral signal characteristics are relatively accurate in judging rotor shaft system faults and have a relatively obvious relationship. The correct results can be obtained through relatively simple machine learning. Bearing faults generally manifest as high-frequency characteristics, and the spectral information is relatively messy. In this case, deep learning training can be carried out by using time-domain signals as samples, and the calculation model can be assisted by using relatively common and mature datasets in the industry.

[0078] Specifically, by collecting fault data in the X and Y directions of the faulty equipment, a first fault dataset is established, and fault labels are established according to the actual fault types on site. No fault is labeled with "0", and other fault types are labeled with 1, 2, 3, etc.

[0079] By obtaining the vibration frequency value at a certain moment from the first fault dataset, and then using Fast Fourier Transform and normalization, the spectral information (vibration amplitude corresponding to each frequency) at the current moment can be obtained. The spectral information is graphically displayed as follows: Figure 4As shown. Then, using expert experience, the vibration data in the first fault dataset is replaced with vibration characteristic information such as the X and Y direction passband values, first harmonic amplitude, first harmonic amplitude percentage, second harmonic amplitude, and second harmonic amplitude percentage, to establish a second fault dataset corresponding to the first fault dataset. The fault causes of each fault data are given by expert experience, and the accuracy of the expert experience is judged based on the actual fault causes on site. Fault types with a success rate greater than 90% are selected as fault labels for the second fault dataset, and fault types with a success rate less than 90% are considered as fault-free in the second fault dataset.

[0080] The second fault dataset with fault labels is shown in Table 1 below. The fault data is divided into training set, test set and validation set, and the SVM machine learning algorithm is used for learning and training to ensure that the diagnosis success rate is greater than 95%.

[0081] Table 1 Second Fault Dataset

[0082]

[0083] It should be noted that by analyzing time-frequency data based on expert experience, selecting fault types with more obvious fault characteristics, and establishing a second fault dataset with fault labels based on vibration characteristic information, and using this dataset to train a machine learning classification model, the number of fault samples required for training the machine learning classification model can be significantly reduced.

[0084] In one embodiment, the method further includes pre-training a deep learning classification model, specifically comprising the following steps:

[0085] The deep learning classification model is trained using a public dataset and a first fault dataset with fault labels to obtain the pre-trained deep learning classification model.

[0086] Specifically, this embodiment uses publicly available datasets (such as the CWRU bearing dataset) and a first fault dataset with fault labels as samples. Similarly, the fault data samples are divided into training set, test set and validation set. A convolutional neural network is used for learning and training until the diagnosis success rate is greater than 95% after training. At the same time, the smaller the loss function value, the better, and it should be less than the set value.

[0087] Furthermore, in this embodiment, the loss function of the convolutional neural network is specifically as follows:

[0088]

[0089] Where r is the actual value, is the predicted value, and n is the number of samples.

[0090] It should be noted that the deep learning classification model trained in this embodiment has a good effect on diagnosing bearing-related faults, especially those with less obvious fault characteristics in time-frequency data; and it uses a public dataset to make up for the small sample size of actual operating data.

[0091] This embodiment mainly combines the advantages of deep learning algorithms based on public datasets for diagnosing bearing-related faults and the advantages of machine learning based on expert experience for adapting well to situations with fewer fault samples. In addition, machine learning based on expert experience has a good diagnostic effect on common fault types, and the processing of expert experience can reduce the number of parameters, significantly reduce the number of fault samples required to train the machine learning model, and significantly reduce the time required for intelligent fault diagnosis.

[0092] In addition, such as Figure 5 As shown, the second embodiment of the present invention proposes a fault intelligent diagnosis device, the device comprising:

[0093] The acquisition module 10 is used to acquire the vibration signal of the device under test and to extract the vibration feature information of the vibration signal using a data processing algorithm;

[0094] The first judgment module 20 is used to process the vibration feature information using a pre-trained machine learning classification model to determine whether a fault has occurred in the rotor shaft system.

[0095] The fault result output module 30 is used to output the cause of the rotor shaft system fault when the output result of the first judgment module is yes;

[0096] The second judgment module 40 is used to process the vibration signal using a pre-trained deep learning classification model when the output result of the first judgment module is negative, and to determine whether the bearing has failed.

[0097] The fault result output module 30 is used to output the bearing fault cause when the output result of the second judgment module is yes; and to output the fault diagnosis result as no fault when the output result of the second judgment module is no.

[0098] This embodiment receives real-time vibration signals from the device under test and filters out the vibration feature information needed for diagnosis. A trained machine learning classification model is then used to diagnose the vibration feature information. If a fault is found, it indicates a problem with the device's rotor shaft system, and fault-related information is output. If no fault is found, the system jumps to a deep learning classification model. The deep learning classification model uses the real-time vibration signals for diagnosis. If a fault is found, it indicates a problem with the bearing, and the cause of the fault is output. If no fault is found, no fault is output. This invention is applicable not only to bearing-related fault diagnosis but also to rotor shaft system-related fault diagnosis. Furthermore, the pre-processing machine learning classification model and post-processing deep learning classification model result in low requirements for computer resources.

[0099] In one embodiment, the sampling frequency of the vibration signal is greater than or equal to the natural frequency of the bearing and rotor, and greater than or equal to twice the maximum rotational frequency of the faulty equipment.

[0100] In one embodiment, the acquisition module 10 includes:

[0101] The data processing unit is used to perform fast Fourier transform and wavelet analysis on the vibration signal to obtain the time domain and frequency domain information of the vibration signal.

[0102] The feature extraction unit is used to extract the vibration feature information from the time domain and frequency domain information of the vibration signal. The vibration feature information includes the vibration passband value and the vibration amplitude of each harmonic.

[0103] In one embodiment, the machine learning classification model employs machine learning models including, but not limited to, decision tree models, support vector machine models, naive Bayes models, or random forest models.

[0104] In one embodiment, the deep learning classification model employs deep learning models including, but not limited to, feedforward neural network models, recurrent neural network models, convolutional neural network models, or long short-term memory neural network models.

[0105] In one embodiment, the device further includes a first training module, comprising:

[0106] The acquisition unit is used to acquire the fault vibration signal of the faulty equipment, establish a first fault dataset based on the fault vibration signal, and establish fault labels according to the actual fault type on site.

[0107] An extraction unit is used to extract fault vibration feature information from the fault vibration signal and establish a second fault dataset based on the fault vibration feature information.

[0108] The fault type determination unit is used to use expert experience to determine the cause of the fault and to use the fault type with a success rate greater than a set threshold as the fault label of the second fault dataset.

[0109] The first training unit is used to train the machine learning classification model using the second fault dataset with fault labels to obtain the pre-trained machine learning classification model.

[0110] In one embodiment, the apparatus further includes a second training module for:

[0111] The deep learning classification model is trained using a public dataset and a first fault dataset with fault labels to obtain the pre-trained deep learning classification model.

[0112] This embodiment mainly combines the advantages of deep learning algorithms based on public datasets for diagnosing bearing-related faults and the advantages of machine learning based on expert experience for adapting well to situations with fewer fault samples. In addition, machine learning based on expert experience has a good diagnostic effect on common fault types, and the processing of expert experience can reduce the number of parameters, significantly reduce the number of fault samples required to train the machine learning model, and significantly reduce the time required for intelligent fault diagnosis.

[0113] It should be noted that other embodiments or implementation methods of the intelligent fault diagnosis device described in this invention can refer to the above-described method embodiments, and will not be repeated here.

[0114] Furthermore, the third embodiment of the present invention also proposes a fault intelligent diagnosis device, the device including a memory and a processor; wherein, the processor reads executable program code stored in the memory to run a program corresponding to the executable program code, so as to implement the fault intelligent diagnosis method as described above.

[0115] Furthermore, the fourth embodiment of the present invention also proposes a computer-readable storage medium storing a computer program thereon, characterized in that, when the computer program is executed by a processor, it implements the intelligent fault diagnosis method described above.

[0116] It should be noted that the logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include: an electrical connection having one or more wires (electronic device), a portable computer disk drive (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Alternatively, the computer-readable medium may be paper or other suitable media on which the program can be printed, since the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in a computer memory.

[0117] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0118] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0119] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "a plurality of" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0120] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A fault intelligent diagnosis method, characterized in that, The method includes the following steps: The vibration signal of the device under test is collected, and the vibration feature information of the vibration signal is extracted using a data processing algorithm; The vibration feature information is processed using a pre-trained machine learning classification model to determine whether a fault has occurred in the rotor shaft system. If so, output the cause of the rotor shaft system fault; If not, the vibration signal is processed using a pre-trained deep learning classification model to determine whether the bearing has failed. If so, then output the cause of the bearing failure; If not, the fault diagnosis result is no fault; The pre-training process of the machine learning classification model includes: Collect the fault vibration signal of the faulty equipment, establish a first fault dataset based on the fault vibration signal, and establish fault labels according to the actual fault type on site; Extract the fault vibration feature information of the fault vibration signal, and establish a second fault dataset based on the fault vibration feature information; The cause of the fault is determined by expert experience, and the fault type with a success rate greater than a set threshold is used as the fault label of the second fault dataset. The machine learning classification model is trained using the second fault dataset with fault labels to obtain the pre-trained machine learning classification model. The pre-training process of the deep learning classification model includes: The deep learning classification model is trained using a public dataset and a first fault dataset with fault labels to obtain the pre-trained deep learning classification model.

2. The intelligent fault diagnosis method as described in claim 1, characterized in that, The sampling frequency of the vibration signal is greater than or equal to the natural frequency of the bearing and rotor, and greater than or equal to twice the maximum rotational frequency of the faulty equipment.

3. The intelligent fault diagnosis method as described in claim 1, characterized in that, The process of collecting vibration signals from faulty equipment and extracting vibration characteristic information from these signals using data processing algorithms includes: The vibration signal is subjected to fast Fourier transform and wavelet analysis to obtain the time domain and frequency domain information of the vibration signal; The vibration feature information is extracted from the time and frequency domain information of the vibration signal. The vibration feature information includes the vibration passband value and the vibration amplitude of each harmonic.

4. The intelligent fault diagnosis method as described in claim 1, characterized in that, The machine learning classification model adopts a decision tree model, a support vector machine model, a Naive Bayes model, or a random forest model.

5. The intelligent fault diagnosis method as described in claim 1, characterized in that, The deep learning classification model adopts a feedforward neural network model, a recurrent neural network model, a convolutional neural network model, or a long short-term memory neural network model.

6. A fault intelligent diagnostic device, characterized in that, The device includes: The acquisition module is used to acquire vibration signals from the device under test and to extract vibration characteristic information from the vibration signals using data processing algorithms. The first judgment module is used to process the vibration feature information using a pre-trained machine learning classification model to determine whether a fault has occurred in the rotor shaft system. The fault result output module is used to output the cause of the rotor shaft system fault when the output result of the first judgment module is yes; The second judgment module is used to process the vibration signal using a pre-trained deep learning classification model when the output result of the first judgment module is negative, and to determine whether the bearing has failed. The fault result output module is used to output the bearing fault cause when the output result of the second judgment module is yes; and to output the fault diagnosis result as no fault when the output result of the second judgment module is no. The pre-training process of the machine learning classification model includes: Collect the fault vibration signal of the faulty equipment, establish a first fault dataset based on the fault vibration signal, and establish fault labels according to the actual fault type on site; Extract the fault vibration feature information of the fault vibration signal, and establish a second fault dataset based on the fault vibration feature information; The cause of the fault is determined by expert experience, and the fault type with a success rate greater than a set threshold is used as the fault label of the second fault dataset. The machine learning classification model is trained using the second fault dataset with fault labels to obtain the pre-trained machine learning classification model. The pre-training process of the deep learning classification model includes: The deep learning classification model is trained using a public dataset and a first fault dataset with fault labels to obtain the pre-trained deep learning classification model.

7. A fault intelligent diagnostic device, characterized in that, The device includes a memory and a processor; wherein the processor runs a program corresponding to the executable program code by reading executable program code stored in the memory, so as to implement the method as described in any one of claims 1-5.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method as described in any one of claims 1-5.

Images