Fault diagnosis method and system based on part characteristic frequency, medium and equipment

Abstract

The invention relates to the field of mechanical equipment fault diagnosis, and discloses a fault diagnosis method and system based on part characteristic frequency, a medium and equipment, and the method comprises the steps: obtaining the part fault characteristic frequency from an original fault data set of mechanical equipment, and constructing guidance information of physical knowledge based on the part fault characteristic frequency; splicing the generated fault feature frequency information with fault data to form dual-channel data, and inputting the dual-channel data into a multi-output deep convolutional neural network for training to extract fault features; and performing multi-output classification and diagnosis on the fault features by a multi-output deep convolutional neural network, each output corresponding to a fault type of one mechanical device, extracting the fault features corresponding to the fault types from the global features, and completing diagnosis of whether the fault exists in the mechanical device. According to the method, the composite fault can be decoupled, and prediction of a single fault type is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of mechanical equipment fault diagnosis, and in particular to a fault diagnosis method, system, medium and equipment based on the characteristic frequency of a part. Background Art

[0002] Traditional fault diagnosis techniques typically rely on theoretical analysis, combining the equipment's operating principles and physical characteristics to extract characteristic parameters associated with the fault. For rotating machinery, for example, the frequency characteristics of the vibration signal are a key analytical metric. When a bearing fault occurs, specific characteristic frequency components will appear in its vibration spectrum. Identifying these frequency components can effectively determine the equipment's health. However, in actual engineering applications, due to factors such as manufacturing errors, mechanical wear during operation, and speed fluctuations, the characteristic frequencies in the measured signal may exhibit slight deviations. This, combined with the influence of background noise and unique vibration patterns specific to the equipment, requires detailed analysis and identification by experienced experts.

[0003] However, in the field of deep neural network fault diagnosis, when faced with complex and changeable fault signals, it is impossible to accurately predict single-category faults; and due to the lack of practical physical knowledge, the model cannot correctly extract fault features on smaller fault data sets. Summary of the Invention

[0004] In view of the above problems, the purpose of the present invention is to provide a fault diagnosis method, system, medium and equipment based on the characteristic frequency of parts, which can decouple complex faults and realize the prediction of single fault types.

[0005] To achieve the above-mentioned purpose, in the first aspect, the technical solution adopted by the present invention is: a fault diagnosis method based on part characteristic frequency, which includes: obtaining part fault characteristic frequency from the original fault data set of mechanical equipment, and constructing guidance information of physical knowledge based on the part fault characteristic frequency; splicing the generated fault characteristic frequency information with the fault data to form dual-channel data, and inputting the multi-output deep convolutional neural network for training to extract fault characteristics; the multi-output deep convolutional neural network performs multi-output classification and diagnosis on the fault characteristics, each output corresponds to a fault type of a mechanical equipment, and extracts the fault characteristics corresponding to the fault type from the global features to complete the diagnosis of whether the fault exists in the mechanical equipment.

[0006] Furthermore, the characteristic frequencies of component failures are obtained from the original fault dataset of the mechanical equipment, and guidance information based on the characteristic frequencies of component failures is constructed, including:

[0007] Based on the failure principle and parameter information of the part, the characteristic frequency of the part failure is calculated, and the left and right frequency spreading is performed with the characteristic frequency of the part failure as the center to form a characteristic frequency band;

[0008] The vibration data of the collected mechanical equipment is band-pass filtered using a characteristic frequency band, and the resulting fault characteristic frequency serves as guiding information for physical knowledge.

[0009] Furthermore, the generated fault characteristic frequency information is spliced with the fault data to form dual-channel data, including:

[0010] According to the differences and needs of the parts, the guidance information is spliced with the original collected vibration data, and then the spectrum is converted into the frequency domain or time-frequency domain to form dual-channel data.

[0011] Furthermore, the multi-output deep convolutional neural network adopts a parallel convolutional network, and the number of parallel convolutional networks is determined according to the part data or spectrum signal type.

[0012] Furthermore, a multi-output deep convolutional neural network is used to perform multi-output classification on the fault features, including:

[0013] Use a flattening layer to convert the extracted multi-dimensional composite fault features into one-dimensional features, or use a global average pooling layer for dimensionality reduction;

[0014] Connect multiple output branches, each branch corresponds to a fault type; in each branch, select N layers of fully connected layers as needed, and finally use a fully connected layer with only one output and a sigmoid activation function to output the probability of the fault to determine whether the fault exists.

[0015] Furthermore, after multi-output classification of faults, the following constraint steps are also included:

[0016] The single fault probabilities of all outputs are spliced together to perform multi-output constraints to further constrain the output of each independent branch.

[0017] Furthermore, the multi-output constraint is: the F1 score and cross entropy loss in the convolutional network are used to further constrain the output of each independent branch.

[0018] In a second aspect, the technical solution adopted by the present invention is: a fault diagnosis system based on the characteristic frequency of a part, comprising:

[0019] The physical knowledge information acquisition module obtains the characteristic frequency of part failure from the original fault data set of mechanical equipment and constructs guidance information of physical knowledge based on the characteristic frequency of part failure;

[0020] The feature extraction module combines the generated fault feature frequency information with the fault data to form dual-channel data, which is then fed into a multi-output deep convolutional neural network for training to extract fault features.

[0021] The fault decoupling classification and diagnosis module uses a multi-output deep convolutional neural network to perform multi-output classification and diagnosis on fault features. Each output corresponds to a fault type of mechanical equipment. The fault features corresponding to the fault type are extracted from the global features to complete the diagnosis of whether the fault exists in the mechanical equipment.

[0022] In a third aspect, the technical solution adopted by the present invention is: a computer-readable storage medium storing one or more programs, wherein the one or more programs include instructions, and when the instructions are executed by a computing device, the computing device executes any one of the above methods.

[0023] In a fourth aspect, the technical solution adopted by the present invention is: a computing device, comprising: one or more processors, a memory and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, and the one or more programs include instructions for executing any of the above methods.

[0024] The present invention has the following advantages due to the adoption of the above technical solution:

[0025] 1. The present invention extracts the characteristic frequency of part failures from the original fault data set of mechanical equipment, and constructs guidance information of physical knowledge based on the characteristic frequency of part failures, which effectively solves the problem that the model cannot correctly extract fault features on smaller fault data sets due to the lack of actual physical knowledge.

[0026] 2. The present invention uses a multi-output deep convolutional neural network to perform multi-output classification and diagnosis on composite fault features, and combined with multi-output constraints, it can decouple composite faults. Even if the corresponding composite fault data has not been seen in the training phase, it can achieve accurate diagnosis of the operating status of mechanical equipment. When there are multiple faults in the system, they can be decomposed and diagnosed into multiple separate faults, and overfitting of single fault features can be effectively avoided. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 This is an overall flow chart of a fault diagnosis method based on a part characteristic frequency according to an embodiment of the present invention;

[0028] Figure 2 is a detailed flow chart of a fault diagnosis method based on a component characteristic frequency in an embodiment of the present invention;

[0029] Figure 3is the vibration signal envelope spectrum in the composite fault characteristic of the bearing outer ring fault and the flywheel imbalance in the embodiment of the present invention;

[0030] Figure 4 It is characteristic frequency information generated based on the envelope spectrum in the composite fault feature of the bearing outer ring fault and the flywheel imbalance in the embodiment of the present invention;

[0031] Figure 5 This is a flow chart of composite fault classification and diagnosis using bearings as an example in an embodiment of the present invention;

[0032] Figure 6 This is a schematic diagram of fault diagnosis using a deep neural network based on customized wavelet convolution kernel in solution 1 of the prior art;

[0033] Figure 7 This is a schematic diagram of fault diagnosis using a deep learning composite fault diagnosis model in solution 2 of the prior art;

[0034] Figure 8 This is a fault diagnosis diagram of Solution 3 in the existing technology, which uses a deep learning composite fault diagnosis model. DETAILED DESCRIPTION

[0035] Due to their ability to adaptively learn and extract features end-to-end, deep neural networks have been extensively and intensively studied in the field of mechanical parts fault diagnosis. The current development of deep learning technology relies heavily on the support of large-scale annotated data, especially in areas such as image classification and natural language processing, where mature data annotation systems have been established. However, in the field of mechanical fault diagnosis, the amount of available annotated data is extremely limited due to the complex equipment operating environment and the high cost and long cycle of fault sample collection. This data scarcity makes overfitting prone to occur during model training. The model tends to capture non-essential features in the training samples (such as noise interference or accidental states of equipment operation). Although it can achieve high accuracy on specific test sets, its generalization ability across operating conditions and equipment scenarios is significantly reduced. This learning model based on surface features not only limits the practical application value of the diagnostic model, but also raises questions about the reliability of its diagnostic results.

[0036] In the actual operation of equipment, compound faults have become a common occurrence, characterized by high uncertainty and low predictability. Equipment failures not only lead to reduced economic benefits but can also impact personnel and the environment. Therefore, developing universal and efficient multi-fault diagnosis techniques for mechanical parts is crucial for ensuring production safety and preventing and avoiding accidents. Numerous scholars have conducted research and exploration into compound fault diagnosis and treatment. Some early studies used the softmax activation function as the output of convolutional network models, treating compound faults as a new fault mode rather than a coupled mode of single faults. Using softmax as a classifier only treats compound faults as a new fault mode, rather than decoupling them into single faults. This results in a dramatic increase in classification labels. Furthermore, since model training struggles to cover all combinations of compound faults, practical application is unfeasible.

[0037] As research deepens, in response to the shortcomings of softmax, many studies have begun to use multi-label classifiers to decouple compound faults. Unlike multi-classification classifiers using softmax, multi-label classifiers use the sigmoid activation function as the activation function of the last layer classifier. Using the sigmoid activation function, each output probability of the model is independent of each other, so the fault classification label can be output according to the number of single faults in the compound fault, thereby achieving the decoupling of compound faults. Compared with softmax, the multi-label classifier using sigmoid can complete the decoupling of compound faults without causing an increase in classification labels. However, due to the limitations of the network structure, the network model using the sigmoid multi-label classifier cannot predict compound fault combinations that have not been seen in the training phase, even if the characteristics of all single faults have been learned in the training phase. Therefore, when training a model using a sigmoid classifier, it is necessary to provide the model with all the compound fault data that are expected to be predicted. Therefore, in existing research, there are mainly the following solutions:

[0038] Solution 1: Integrating physical knowledge into deep neural networks is currently a mainstream direction, that is, using physical knowledge to guide the learning of neural network models to prevent the models from learning non-essential features. The main idea of improving the feature learning ability of fault diagnosis models by improving the CNN network architecture is to replace traditional convolution kernels by embedding kernels with physical meanings. Typical methods include customized wavelet convolution kernels, such as Figure 6 For example, replacing the first convolutional layer in a CNN model with a continuous wavelet convolution layer is equivalent to adding some constraints to the waveform of the convolution kernel. This allows for the extraction of features with clear physical meaning from the data. This mechanism enables efficient feature selection in CNN models and reveals the correlation between the inner product of the random feature space and the signal space.

[0039] However, a crucial aspect of wavelet transform in this scheme is the choice of wavelet basis. Different wavelet bases have different properties, and their selection directly impacts signal decomposition and denoising. Therefore, in practical applications, the adaptability and versatility of the wavelet convolution kernel remains to be improved when faced with complex and variable fault signals.

[0040] Option 2: If Figure 7 As shown in the figure, the softmax activation function is used in the final fully connected layer of the fault diagnosis and classification module. The softmax activation function is commonly used in multi-classification scenarios, where there are multiple classification types, a sample belongs to only one classification, and different classifications are mutually exclusive. Softmax maps the outputs of multiple neurons to the range [0, 1] according to the following formula, so that the sum of all outputs is 1.

[0041]

[0042] Where z i Represents the output value of the i-th node; j is the number of output nodes, that is, the number of classification categories.

[0043] In this multi-classification scenario, the softmax function predicts the probability of each class, typically selecting the class with the highest predicted value. This means that each prediction outputs only one class. In this scenario, softmax cannot simultaneously predict multiple classes. Therefore, when using this approach, the composite fault classifier treats a composite fault as a single label, essentially treating it as a single fault class.

[0044] This approach has the following problems: 1. The most significant issue is that it treats compound faults as a single fault. Mechanical equipment has numerous parts and a wide variety of fault types, resulting in a vast number of compound fault combinations. Furthermore, given the diverse application scenarios and operating conditions, this approach is not practically applicable. 2. Given the large number of compound fault combinations, data collection for model training presents significant challenges and is practically impossible.

[0045] Option 3: If Figure 8 As shown in the figure, the sigmoid activation function is used in the final fully connected layer of the fault diagnosis and classification module. The sigmoid activation function is commonly used in multi-label classification scenarios, where a sample can belong to multiple label classification types, and the categories are not mutually exclusive. For example, cars come in different types, such as sedans, SUVs, and trucks, and also have different colors. The sigmoid function independently maps the outputs of multiple neurons to the range [0, 1] according to the following formula.

[0046]

[0047] Where x represents the output value of the neuron.

[0048] In this multi-label classification scenario, the sigmoid function predicts the probability of each class. When the probability is greater than a threshold, such as 0.5, the sample is considered to belong to that class. In this case, the model can output multiple classification labels for each prediction. In this scenario, the sigmoid function can simultaneously output multiple classes. Therefore, when a composite fault classifier employs this approach, it can theoretically predict multiple labels for a composite fault, with each label corresponding to a single fault class.

[0049] However, the biggest challenge with this solution is that, in real-world applications, to classify a compound fault as multiple single faults, the model must have seen data with this compound fault during training and complete the learning process based on the labels. If the model has not seen data with a compound fault during training, it will not be able to make accurate classification diagnoses in real-world applications. Furthermore, due to the numerous combinations of compound faults, collecting data for training the model presents a significant challenge, making it practically impossible.

[0050] To address the problems existing in the prior art and the aforementioned solutions, the present invention provides a fault diagnosis method, system, medium, and device based on component characteristic frequencies. By incorporating prior knowledge of the characteristic frequencies of mechanical component failures, the present invention constructs a theoretically guided feature constraint mechanism, guiding the neural network to focus on characteristic frequency components directly related to the fault mode during training, thereby suppressing interference from irrelevant features. Furthermore, a multi-output classification scheme within a convolutional neural network is employed, with each output corresponding to a single fault type of the mechanical device. Based on all features extracted by the neural network, the presence of the characteristics of the specified single fault is searched, and the presence of the corresponding single fault type is output.

[0051] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the described embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field are within the scope of protection of the present invention.

[0052] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is intended to include the plural form. In addition, it should be understood that when the terms "comprise" and / or "include" are used in this specification, they indicate the presence of features, steps, operations, devices, components and / or combinations thereof.

[0053] In one embodiment of the present invention, a fault diagnosis method based on the characteristic frequency of a part is provided. In this embodiment, Figure 1 、 Figure 2 As shown, the method includes the following steps:

[0054] 1) Extracting the characteristic frequencies of component failures from the original mechanical equipment fault dataset, constructing guidance information based on physical knowledge based on the characteristic frequencies of component failures for training the neural network model, and guiding the neural network to focus on the characteristic frequency components directly related to the failure mode during training;

[0055] The guidance information based on physical knowledge refers to the generated fault characteristic frequency. This characteristic frequency is calculated based on the component failure mechanism and its parameter information, such as the pitch diameter and rolling element diameter of a specific bearing model. Based on this calculated characteristic frequency, a certain bandwidth, such as 5 Hz, is then expanded to form a characteristic frequency band. This characteristic frequency band is then used to filter the vibration information collected from the equipment to generate guidance information. This is why it is called guidance information based on physical knowledge, and more specifically, it can also be referred to as guidance information based on the fault characteristic frequency.

[0056] 2) The generated fault feature frequency information is combined with the fault data to form dual-channel data, which is then fed into a multi-output deep convolutional neural network for training to extract fault features. The fault features include composite fault features and single fault features.

[0057] 3) A multi-output deep convolutional neural network is used to perform multi-output classification and diagnosis of fault features. Each output corresponds to a fault type of mechanical equipment. The fault features corresponding to the fault type are extracted from the global features to complete the diagnosis of whether the fault exists in the mechanical equipment.

[0058] In the above step 1), the component failure characteristic frequency is obtained from the original fault data set of the mechanical equipment, and the guidance information of physical knowledge is constructed based on the component failure characteristic frequency, including the following steps:

[0059] 1.1) Based on the failure principle and parameter information of the component, the characteristic frequency of the component failure is calculated, and the left and right frequency spreading is performed with the characteristic frequency of the component failure as the center to form a characteristic frequency band;

[0060] 1.2) The vibration data of the collected mechanical equipment is band-pass filtered using a characteristic frequency band. The resulting fault characteristic frequency is the guiding information of physical knowledge.

[0061] In this embodiment, for example, the pitch diameter and rolling element diameter of a particular bearing model are used to calculate the characteristic frequency of a component failure. Based on this calculated characteristic frequency, a certain bandwidth, such as 5 Hz, is then expanded to form a characteristic frequency band. This characteristic frequency band is then used to filter vibration information collected from the equipment to generate guidance information. This is referred to as guidance information based on physics knowledge, and more specifically, as guidance information based on the characteristic frequency of the failure.

[0062] In the above step 2), the generated fault characteristic frequency information is spliced with the fault data to form dual-channel data. Specifically, according to the different parts and needs, the guidance information is spliced with the original collected vibration data, and then the spectrum is converted into the frequency domain or time-frequency domain to form dual-channel data.

[0063] In this embodiment, the spectrum conversion may be envelope spectrum, Fourier spectrum, wavelet transform in time-frequency domain, short-time Fourier transform, or the like.

[0064] Specifically, since traditional fault diagnosis is mainly based on theoretical analysis, features related to the fault are extracted according to the working principle and physical characteristics of the mechanical equipment. For rotating parts, vibration frequency is an important characteristic parameter. For example, when a bearing fails, its vibration frequency will contain characteristic frequency components related to the bearing failure. The fault state of the bearing can be judged by analyzing these frequency characteristics. Due to factors such as manufacturing errors, wear during the operation of the equipment, and slight fluctuations in the speed, the characteristic frequency may be slightly offset during actual detection. In order to improve the performance of the deep neural network model, effectively extract key features from the original fault data, and use these features for fault diagnosis and classification, the present invention constructs guidance information of physical knowledge based on the characteristic frequency of the parts for model training.

[0065] Taking bearings as an example, the above step 2) is explained in detail:

[0066] (1) First, based on the principle of bearing failure, the characteristic frequencies of the bearing are calculated, including the inner race passing frequency (BPFI), the outer race passing frequency (BPFO), the rolling element spin frequency (BSF), and the cage frequency (FTF). The calculation formulas are as follows: (1) to (4).

[0067]

[0068] Where, f BPFI n is the passing frequency of the inner ring of the rolling element; b is the number of rolling elements; f r is the frequency conversion; d b is the rolling element diameter; D c is the bearing pitch diameter (i.e. the diameter of the circle where the rolling element center is located); α is the contact angle; f BPFOf is the passing frequency of the outer ring of the rolling element; BSF is the rolling body spin frequency; f FTF is the cage frequency.

[0069] (2) Based on the calculated characteristic frequency, the left and right extensions are made into characteristic frequency bands to solve the problem of slight characteristic frequency offset. According to the needs of the specific model and data set, the kth harmonic frequency band is calculated, where k = [1…N].

[0070] (3) Use the generated characteristic frequency band to perform bandpass filtering on the vibration information; and simultaneously convert the vibration information and the information filtered using the characteristic frequency band into the frequency domain. Depending on the different parts, the bearing is converted into the envelope spectrum, and the gears, flywheels, etc. are converted into the Fourier spectrum.

[0071] like Figure 3 As shown in the figure, it is the envelope spectrum of a composite fault of bearing outer ring and flywheel imbalance, in which we can see the obvious characteristic frequency of imbalance fault, rotation frequency and its second harmonic, and the characteristic frequency of bearing outer ring fault, BPFO and its second harmonic. Figure 4 The figure shows the characteristic frequency information of a composite fault of a bearing outer ring and flywheel imbalance, including bearing inner ring fault, outer ring fault, rolling element fault and 1 to 5 times of the rotation frequency. Figure 3 、 Figure 4 As can be seen in Figure 2, the main characteristic frequencies of bearing outer race fault and flywheel imbalance fault are retained. These two pieces of information are input into the deep neural network as two channels, guiding the model to focus on the fault characteristics.

[0072] In the above step 2), the multi-output deep convolutional neural network adopts a parallel convolutional network, and the number of parallel convolutional networks is determined according to the part data or spectrum signal type.

[0073] In this embodiment, any conventional convolutional network can be used without limitation. The number of parallel convolutional networks can be determined based on the part data or spectrum signal type. For example, one convolutional network can be used for each part type, or one convolutional network can be used for the envelope spectrum and another for the FFT.

[0074] In step 3) above, the goal of multi-output classification is to classify samples into multiple outputs, each of which can be classification, regression, or multi-label. Multi-output classification methods achieve decoupled classification of complex faults by using multiple output branches, each corresponding to a single fault type. Each branch uses a sigmoid activation function to predict a single fault type.

[0075] Specifically, such as Figure 5 As shown in FIG, a multi-output deep convolutional neural network is used to perform multi-output classification on the fault features, including the following steps:

[0076] 3.1) Use a flattening layer to convert the extracted multi-dimensional composite fault features into one-dimensional features, or use a global average pooling layer for dimensionality reduction;

[0077] 3.2) Connect multiple output branches, each corresponding to a fault type; in each branch, select N layers of fully connected layers as needed. Finally, a fully connected layer with a single output and a sigmoid activation function outputs the probability of the fault to determine whether the fault exists.

[0078] In the above embodiment, after multi-output classification of the fault in step 3.2), a constraint step 3.3) is also included: the single fault probabilities of all outputs are directly spliced together, and multi-output constraints are performed to further constrain each independent branch output to achieve the setting of accuracy and loss function.

[0079] In this embodiment, the multi-output constraint is: the F1 score and cross entropy loss in the convolutional network are used to further constrain the output of each independent branch.

[0080] Specifically, in multi-output fault diagnosis models, when each output branch independently predicts a specific type of fault, it may overfit to noise or background features in the training data. This phenomenon can lead to performance degradation in complex fault scenarios. To address this issue, all output branches of the model are spliced together in parallel and constrained using cross-entropy loss. This encourages the model to learn more general features, improving its robustness and reliability in practical applications and effectively preventing overfitting to single fault features during training.

[0081] In summary, the present invention constructs guidance information of physical knowledge based on the characteristic frequency of parts, which is used for model training to guide the model to learn the correct fault characteristics; the multi-output mode in the deep neural network is used to realize the decomposition and diagnosis of complex faults of mechanical equipment, each output corresponds to the fault type of a mechanical equipment, and the features corresponding to the fault are extracted from the global features to complete the diagnosis of whether the fault exists in the equipment. It can extract the correct fault features and has good generalization ability; at the same time, it can realize accurate diagnosis of the operating status of mechanical equipment. When there are multiple faults in the system, they can be decomposed and diagnosed into multiple separate faults.

[0082] In one embodiment of the present invention, a fault diagnosis system based on a component characteristic frequency is provided, comprising:

[0083] The physical knowledge information acquisition module obtains the characteristic frequency of part failure from the original fault data set of mechanical equipment and constructs guidance information of physical knowledge based on the characteristic frequency of part failure;

[0084] The feature extraction module combines the generated fault feature frequency information with the fault data to form dual-channel data, which is then fed into a multi-output deep convolutional neural network for training to extract fault features.

[0085] The fault decoupling classification and diagnosis module uses a multi-output deep convolutional neural network to perform multi-output classification and diagnosis on fault features. Each output corresponds to a fault type of mechanical equipment. The fault features corresponding to the fault type are extracted from the global features to complete the diagnosis of whether the fault exists in the mechanical equipment.

[0086] In the above embodiment, the physical knowledge information acquisition module obtains the characteristic frequency of part failure from the original fault data set of the mechanical equipment, and constructs the guidance information of physical knowledge based on the characteristic frequency of part failure, including:

[0087] Based on the failure principle and parameter information of the part, the characteristic frequency of the part failure is calculated, and the left and right frequency spreading is performed with the characteristic frequency of the part failure as the center to form a characteristic frequency band;

[0088] The vibration data of the collected mechanical equipment is band-pass filtered using a characteristic frequency band, and the resulting fault characteristic frequency serves as guiding information for physical knowledge.

[0089] In this embodiment, the generated fault characteristic frequency information is spliced with the fault data to form dual-channel data. Specifically, according to the differences and needs of the parts, the guidance information is spliced with the original collected vibration data, and then the spectrum is converted into the frequency domain or time-frequency domain to form dual-channel data.

[0090] In the above embodiment, the multi-output deep convolutional neural network adopts a parallel convolutional network, and the number of parallel convolutional networks is determined according to the part data or spectrum signal type.

[0091] In the above embodiment, in the fault decoupling classification and diagnosis module, a multi-output deep convolutional neural network is used to perform multi-output classification on the fault features, including:

[0092] Use a flattening layer to convert the extracted multi-dimensional composite fault features into one-dimensional features, or use a global average pooling layer for dimensionality reduction;

[0093] Connect multiple output branches, each branch corresponds to a fault type; in each branch, select N layers of fully connected layers as needed, and finally use a fully connected layer with only one output and a sigmoid activation function to output the probability of the fault to determine whether the fault exists.

[0094] In this embodiment, after multi-output classification is performed on the fault, constraints are further included: the single fault probabilities of all outputs are spliced together to perform multi-output constraints, so as to further constrain each independent branch output.

[0095] Among them, the multi-output constraint is: the F1 score and cross entropy loss in the convolutional network are used to further constrain the output of each independent branch.

[0096] The system provided in this embodiment is used to execute the above-mentioned method embodiments. Please refer to the above-mentioned embodiments for specific processes and detailed contents, which will not be repeated here.

[0097] In one embodiment of the present invention, a computing device is provided. The computing device may be a terminal and may include: a processor, a communications interface, a memory, a display screen, and an input device. The processor, communications interface, and memory communicate with each other via a communications bus. The processor is configured to provide computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system and a computer program. When executed by the processor, the computer program implements the methods described in the above embodiments. The internal memory provides an environment for the operating system and computer program in the non-volatile storage medium to run. The communications interface is configured to communicate with an external terminal via wired or wireless communication. The wireless communication may be achieved via Wi-Fi, a network management service provider, NFC (near field communication), or other technologies. The display screen may be a liquid crystal display or an electronic ink display. The input device may be a touchscreen layer covering the display screen, or may be buttons, a trackball, or a touchpad provided on the computing device housing, or may be an external keyboard, touchpad, or mouse. The processor may invoke logic instructions stored in the memory.

[0098] In addition, the logical instructions in the above-mentioned memory can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when sold or used as an independent product. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0099] In one embodiment of the present invention, a computer program product is provided, which includes a computer program stored on a non-transitory computer-readable storage medium, and the computer program includes program instructions. When the program instructions are executed by a computer, the computer can execute the methods provided by the above-mentioned method embodiments.

[0100] In one embodiment of the present invention, a non-transitory computer-readable storage medium is provided. The non-transitory computer-readable storage medium stores server instructions. The computer instructions enable a computer to execute the methods provided in the above embodiments.

[0101] The above embodiment provides a computer-readable storage medium, whose implementation principle and technical effects are similar to those of the above method embodiment, and will not be repeated here.

[0102] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0103] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0104] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0105] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A fault diagnosis method based on the characteristic frequency of a part, characterized in that: include: Obtain the characteristic frequency of part failures from the original fault data set of mechanical equipment, and construct guidance information based on the characteristic frequency of part failures; The generated fault feature frequency information is spliced with the fault data to form dual-channel data, which is then input into a multi-output deep convolutional neural network for training to extract fault features. A multi-output deep convolutional neural network is used to perform multi-output classification and diagnosis on fault features. Each output corresponds to a fault type of mechanical equipment. The fault features corresponding to the fault type are extracted from the global features to complete the diagnosis of whether the fault exists in the mechanical equipment.

2. The fault diagnosis method based on the characteristic frequency of a part according to claim 1, characterized in that: Obtain part failure characteristic frequencies from the original mechanical equipment failure dataset and construct guidance information based on physical knowledge based on the part failure characteristic frequencies, including: Based on the failure principle and parameter information of the part, the characteristic frequency of the part failure is calculated, and the left and right frequency spreading is performed with the characteristic frequency of the part failure as the center to form a characteristic frequency band; The vibration data of the collected mechanical equipment is band-pass filtered using a characteristic frequency band, and the resulting fault characteristic frequency serves as guiding information for physical knowledge.

3. The fault diagnosis method based on the characteristic frequency of a part according to claim 1, characterized in that: The generated fault characteristic frequency information is spliced with the fault data to form dual-channel data, including: According to the differences and needs of the parts, the guidance information is spliced with the original collected vibration data, and then the spectrum is converted into the frequency domain or time-frequency domain to form dual-channel data.

4. The fault diagnosis method based on the characteristic frequency of a part as claimed in claim 1, characterized in that: The multi-output deep convolutional neural network adopts parallel convolutional networks, and the number of parallel convolutional networks is determined according to the part data or spectrum signal type.

5. The fault diagnosis method based on the characteristic frequency of a part as claimed in claim 1, characterized in that: The multi-output deep convolutional neural network performs multi-output classification on the fault features, including: Use a flattening layer to convert the extracted multi-dimensional composite fault features into one-dimensional features, or use a global average pooling layer for dimensionality reduction; Connect multiple output branches, each branch corresponds to a fault type; in each branch, select N layers of fully connected layers as needed, and finally use a fully connected layer with only one output and a sigmoid activation function to output the probability of the fault to determine whether the fault exists.

6. The fault diagnosis method based on the characteristic frequency of a part as claimed in claim 5, characterized in that: After multi-output classification of faults, the constraint step is also included: The single fault probabilities of all outputs are spliced together to perform multi-output constraints to further constrain the output of each independent branch.

7. The fault diagnosis method based on the characteristic frequency of a part as claimed in claim 6, characterized in that: The multi-output constraint is to use the F1 score and cross entropy loss in the convolutional network to further constrain the output of each independent branch.

8. A fault diagnosis system based on the characteristic frequency of a part, characterized in that: include: The physical knowledge information acquisition module obtains the characteristic frequency of part failure from the original fault data set of mechanical equipment and constructs guidance information of physical knowledge based on the characteristic frequency of part failure; The feature extraction module combines the generated fault feature frequency information with the fault data to form dual-channel data, which is then fed into a multi-output deep convolutional neural network for training to extract fault features. The fault decoupling classification and diagnosis module uses a multi-output deep convolutional neural network to perform multi-output classification and diagnosis on fault features. Each output corresponds to a fault type of mechanical equipment. The fault features corresponding to the fault type are extracted from the global features to complete the diagnosis of whether the fault exists in the mechanical equipment.

9. A computer-readable storage medium storing one or more programs, characterized in that: The one or more programs include instructions that, when executed by a computing device, cause the computing device to perform any one of the methods of claims 1 to 7 .

10. A computing device, characterized in that include: One or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, and the one or more programs include instructions for executing any one of the methods according to claims 1 to 7.

Images