Cloud-edge collaborative intelligent fault diagnosis method and system for rotating machinery

Abstract

The invention discloses an intelligent fault diagnosis method and system for rotating machinery based on cloud-side cooperation, belongs to the field of mechanical fault diagnosis, and aims to distribute tasks such as data acquisition and signal processing to an edge end and deploy tasks such as a convolutional neural network model and fault alarm to a cloud end by utilizing the advantages of edge computing and cloud computing. The edge end uses a sensor to collect a running state signal of the rotating machine in real time, then uses a microcontroller to perform feature extraction, sends feature data obtained by the edge end to the cloud end, and uses a pre-trained convolutional neural network model at the cloud end to perform intelligent classification of feature data with labels. Different parts and different fault modes of the rotating machine can be distinguished according to the classification result, and therefore data-driven mechanical fault intelligent diagnosis under cloud-edge cooperation is achieved. The method is excellent in fault diagnosis accuracy and real-time performance, efficient and accurate fault diagnosis of the rotating machine with the rotor-bearing system as the core is achieved, and the method has wide application prospects.

Description

Technical Field

[0001] The present invention belongs to the field of mechanical fault diagnosis, relates to the field of rotating machinery fault diagnosis, and specifically relates to a cloud-edge collaborative rotating machinery intelligent fault diagnosis method and system. Background Art

[0002] The rotor-bearing system is the core component of rotating machinery. Its operating status directly affects the performance and life of the equipment, and may even cause serious safety accidents. Therefore, using the new generation of artificial intelligence technology to realize fault diagnosis of rotating machinery is of great significance to industrial safety production and reliable operation of mechanical equipment.

[0003] The development of cloud computing and edge computing has provided new solutions for intelligent fault diagnosis. Cloud computing has powerful data storage and processing capabilities, and can perform in-depth analysis of massive data. With the development of intelligent sensing technology, the number of production equipment and its monitoring sensors has become increasingly large. The process of uploading the raw data obtained from monitoring to the cloud will generate a huge amount of data, which can easily lead to problems such as high latency, data loss, and network congestion. As a supplement to cloud computing, edge computing can process raw data in a timely manner near the device, reducing data transmission delays and bandwidth consumption. However, the limited computing power of edge devices sometimes makes it difficult to support complex data processing. The cloud-edge collaborative architecture combines the advantages of both and provides a more scientific and efficient solution for intelligent fault diagnosis.

[0004] At present, the research on mechanical fault diagnosis technology is mainly focused on algorithm development, and there is a lack of systematic application solutions. In addition to algorithm development, the application of mechanical fault diagnosis technology also involves algorithm deployment, which is generally divided into edge deployment, cloud deployment, and cloud-edge collaborative deployment. Among them, cloud deployment has made great progress in recent years thanks to the advancement of cloud computing technology, and has begun to take shape; complete edge deployment has not been able to form a reliable solution for the application of mechanical fault diagnosis technology due to the development constraints of hardware resources; cloud-edge collaborative deployment has gradually become the preferred solution for the application of mechanical fault diagnosis technology because it takes into account the advantages of cloud computing and edge computing. At present, there is a lack of effective solutions for the application of mechanical fault diagnosis technology based on cloud-edge collaboration to achieve signal processing and joint deep learning. Summary of the invention

[0005] The purpose of the present invention is to provide a cloud-edge collaborative intelligent fault diagnosis method and system for rotating machinery to overcome the defects of the prior art. The present invention provides a new solution for fault diagnosis of mechanical equipment, which utilizes the advantages of edge computing and cloud computing respectively to make the fault diagnosis process safer and more efficient.

[0006] In order to achieve the above object, the present invention adopts the following technical scheme: A cloud-edge collaborative rotating machinery intelligent fault diagnosis method comprises the following steps: Step 1) Build a cloud-edge collaboration framework Using the cloud and edge collaborative working mechanism, the edge is responsible for collecting data, extracting features and transmitting data to the cloud, and the cloud is responsible for performing data classification on the feature data from the edge, obtaining fault diagnosis results and generating fault alarm information; Step 2) Build an edge signal processing platform The edge signal processing platform includes sensors, conditioning circuits, analog-to-digital conversion units, microcontrollers, and wireless transmission modules. Sensors, conditioning circuits, and analog-to-digital conversion units are used to collect the original signals of rotating mechanical equipment. The microcontroller is used to extract the features of the original signals, and the features are spliced ​​to obtain feature samples. The wireless transmission module is used to upload the feature samples from the edge to the cloud. Step 3) Build a cloud-based fault diagnosis platform The cloud-based fault diagnosis platform consists of a data management module, a rule engine module, and a decision-making recommendation module. The data management module stores and calls feature samples from the edge. The rule engine module consists of label data, classification models, and fault alarms. The label data is mapped to feature samples, and the feature samples are retrieved and sent to the classification model for inference to obtain the predicted labels. The fault alarm results are then given based on the predicted labels. The decision-making recommendation module sends fault handling decision recommendations to the production site based on the fault alarm results of the rule engine.

[0007] Furthermore, in the step 2) of constructing an edge signal processing platform, the sensors are various sensors capable of collecting one-dimensional time series signals, and the multi-source physical quantities collected by the sensors are converted into voltage analog signals. The conditioning circuit filters the voltage analog signals and removes DC bias. The analog-to-digital conversion unit converts the voltage analog signals processed by the conditioning circuit into digital output signals at a preset sampling frequency. The microcontroller has double-precision floating-point operation capabilities, and can realize feature extraction and feature splicing of the original signals.

[0008] Furthermore, the preset sampling frequency is 5kHz-25kHz.

[0009] Furthermore, the step 2) extracting the features of the original signal comprises the following steps: Step 2.1) Construct a fault data set: divide the collected original signal into segments, that is, divide the data of each type of fault state into N segments, where each segment is a sample of length n, and divide the data set into data set A and data set B according to a preset ratio based on the sample as a unit; Step 2.2) Construct an empirical mode decomposition algorithm suitable for edge computing: decompose the sample data of dataset A and dataset B into several IMFs and a residual component. Each IMF represents a different frequency component in the sample, and the residual component represents the trend of the sample. The decomposition calculation formula is:

[0010] In the formula, For samples; For time; is the component of the empirical mode decomposition result, ; is the number of empirical mode decomposition levels; is the residual component; Step 2.3) Construct a fast Fourier transform algorithm suitable for edge computing: perform fast Fourier transform on the IMF components and residual components after empirical mode decomposition, and use the transformation results to calculate frequency domain feature indicators; Step 2.4) Construct a feature extraction algorithm suitable for edge computing: extract time domain features for each IMF component and residual component and frequency domain characteristics , and the time domain features and frequency domain characteristics Concatenate into feature vectors , the eigenvectors of all IMF components and residual components Sequentially stitch together into feature samples After feature extraction, the samples of dataset A and dataset B are used to obtain the feature sample set and ; Step 2.5) Construct a normalization algorithm suitable for edge computing: perform maximum and minimum normalization operations on feature samples to eliminate the magnitude differences between different features in feature samples. The maximum and minimum normalization calculation formula is:

[0011] In the formula, , is the number of features extracted for a single IMF component, is the number of levels of empirical mode decomposition; is the normalized feature sample; is the element in the feature sample; is the maximum value of the feature sample; is the minimum value of the feature sample; first, Perform normalization operation and then use the feature sample set The maximum value vector MAX and the minimum value vector MIN for the feature sample set during normalization operation Perform normalization operation, and After normalization, we get the data set and .

[0012] Furthermore, the sample with a length of n in step 2.1) contains information of at least one change cycle of the rotating machinery, and the preset ratio is 4:1.

[0013] Furthermore, in step 2.4), the extracted time domain features can reflect the changes of the signal over time and are used to detect instantaneous changes or shocks in the signal.

[0014] Furthermore, in step 2.4), the extracted frequency domain features can reflect the changes of the signal in the spectrum and are used to analyze the changes of the periodicity, harmonic components and frequency components of the signal.

[0015] Furthermore, in the step 3) of building a cloud fault diagnosis platform, the classification model is constructed by: Divide into training set and test set, train the classification model, first take different hyperparameter values ​​to train the classification model, combine the stability and convergence speed of the training process and the accuracy of the test set to determine the optimal value of the hyperparameter, then use the determined hyperparameter to train the classification model to obtain the data classification model, use the classification accuracy of the data classification model on the test set as the evaluation index, select the classification model with the highest classification accuracy as the final classification model, the classification accuracy calculation formula is:

[0016] In the formula, Indicates the classification accuracy of the sample; Indicates the number of correctly classified samples in the test set; Represents the total number of test set samples.

[0017] Furthermore, the dataset The dataset is divided into training set and test set according to the ratio of 7:3.

[0018] A cloud-edge collaborative rotating machinery intelligent fault diagnosis system, comprising: Cloud-edge collaboration framework: Build a collaborative working mechanism between the cloud and edge. The edge is responsible for collecting data, extracting features, and transmitting data to the cloud. The cloud is responsible for performing data classification on the feature data from the edge, obtaining fault diagnosis results, and generating fault alarm information. Edge signal processing platform: including sensors, conditioning circuits, analog-to-digital conversion units, microcontrollers and wireless transmission modules. Sensors, conditioning circuits and analog-to-digital conversion units are used to collect the original signals of rotating mechanical equipment. Microcontrollers are used to extract the features of the original signals, and the features are spliced ​​to obtain feature samples. The wireless transmission module is used to upload the feature samples from the edge to the cloud. Cloud-based fault diagnosis platform: The cloud-based fault diagnosis platform consists of data management, rule engine and decision-making recommendation modules. The data management module stores and calls feature samples from the edge. The rule engine module consists of label data, classification models and fault alarms. The label data is mapped to feature samples, and the feature samples are fed into the classification model for inference to obtain the predicted labels. The fault alarm results are then given based on the predicted labels. The decision-making recommendation module sends fault handling decision recommendations to the production site based on the fault alarm results of the rule engine.

[0019] Compared with the prior art, the present invention has the following beneficial technical effects: 1) In response to the problem of intelligent fault diagnosis of rotating machinery, a cloud-edge collaborative fault diagnosis system and an intelligent fault diagnosis algorithm suitable for the cloud-edge collaborative system were proposed, which significantly improved the efficiency, accuracy and reliability of rotating machinery fault diagnosis, gave full play to the advantages of edge computing and cloud computing, exceeded the diagnostic capabilities of a single application, and formed a synergistic effect.

[0020] 2) Designed a set of edge devices, including embedded hardware and onboard programs. Designed a cloud-based fault diagnosis visualization interface that can display fault diagnosis information in a personalized way. Used embedded devices for data acquisition, signal processing, and feature extraction to achieve edge private processing of raw data. Deployed a convolutional neural network model on the cloud platform, and used powerful cloud computing capabilities to achieve efficient and accurate fault diagnosis, which can be extended to more fields of rotating machinery fault diagnosis applications.

[0021] 3) For fault diagnosis methods, a signal processing and feature extraction algorithm library suitable for edge computing has been developed. Signal processing and feature extraction of real-time collected data are realized on the edge, providing high-quality feature data for cloud-based fault diagnosis. A deep learning algorithm suitable for cloud computing has been developed, which can accurately classify feature data on the cloud, with excellent diagnostic accuracy and efficiency, providing a new feasible approach for intelligent fault diagnosis of rotating machinery. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] The drawings in the specification are used to provide further understanding of the present invention and constitute a part of the present invention. The schematic embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0023] Figure 1This is a flow chart of a cloud-edge collaborative rotating machinery intelligent fault diagnosis method; Figure 2 Schematic diagram of the cloud-edge collaboration framework constructed for the present invention; Figure 3 It is a visualization diagram of edge processing of the present invention; Figure 4 It is a schematic diagram of the cloud diagnosis result of the present invention. DETAILED DESCRIPTION

[0024] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in 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 embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0025] It should be understood that the numbers used in this way are interchangeable where appropriate, so that the embodiments of the invention described herein can be implemented in sequences other than those illustrated or described herein. In addition, the terms "comprises" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product, or apparatus comprising a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units that are not explicitly listed or inherent to these processes, methods, products, or apparatus.

[0026] Embodiment 1 The present invention proposes a cloud-edge collaborative intelligent fault diagnosis framework for rotating machinery with a rotor-bearing system as the core component, gives an intelligent fault diagnosis model suitable for the cloud-edge collaborative framework, introduces edge computing devices and a cloud service platform suitable for rotating machinery fault diagnosis, and decomposes the algorithm according to function and deploys it to the edge device and cloud server respectively, to realize cloud-edge collaborative intelligent fault diagnosis of rotating machinery.

[0027] The specific steps include: 1) Build a cloud-edge collaborative framework: Use the cloud and edge collaborative working mechanism. The edge is responsible for collecting data, extracting features, and transmitting data to the cloud. The cloud is responsible for performing data classification on the feature data from the edge, obtaining fault diagnosis results, and generating fault alarm information. 2) Build an edge signal processing platform: including sensors, conditioning circuits, analog-to-digital conversion units, microcontrollers and wireless transmission modules. Use sensors, conditioning circuits and analog-to-digital conversion units to collect the original signals of rotating mechanical equipment, use microcontrollers to extract the features of the original signals, splice the features to obtain feature samples, and use wireless transmission modules to upload feature samples from the edge to the cloud. The specific steps for building the edge signal processing platform in step 2) are as follows: 2.1) Constructing fault data set: The collected original signal is segmented, that is, the data of each type of fault state is divided into N segments, where each segment is a sample of length n, and the data set is divided into data set A and data set B according to a preset ratio based on the sample as a unit; 2.2) Construct an empirical mode decomposition algorithm suitable for edge computing: Decompose the samples of data set A and data set B into several intrinsic mode functions (IMFs) and a residual component. Each IMF represents a different frequency component in the sample, and the residual component represents the trend of the sample. The decomposition calculation formula is:

[0028] In the formula, For samples; For time; is the component of the empirical mode decomposition result, ; is the number of empirical mode decomposition levels; is the residual component; 2.3) Construct a fast Fourier transform algorithm suitable for edge computing: Perform fast Fourier transform on the IMF components and residual components after empirical mode decomposition, and use the transformation results to calculate frequency domain feature indicators; 2.4) Construct a feature extraction algorithm suitable for edge computing: extract time domain features for each IMF component and residual component and frequency domain characteristics , and the time domain features and frequency domain characteristics Concatenate into feature vectors , the eigenvectors of all IMF components and residual components Sequentially stitch together into feature samples After feature extraction, the samples of dataset A and dataset B are used to obtain the feature sample set and ; 2.5) Construct a normalization algorithm suitable for edge computing: Perform maximum and minimum normalization operations on feature samples to eliminate the magnitude differences between different features in feature samples. The maximum and minimum normalization calculation formula is:

[0029] In the formula, , is the number of features extracted for a single IMF component, is the number of levels of empirical mode decomposition; is the normalized feature sample; is the element in the feature sample; is the maximum value of the feature sample; is the minimum value of the feature sample; first, Perform normalization operation and then use the feature sample set The maximum value vector (MAX) and the minimum value vector (MIN) of the feature sample set during normalization operation Perform normalization operation, and After normalization, we get the data set and .

[0030] 3) Build a cloud-based fault diagnosis platform: The cloud-based fault diagnosis platform consists of data management, rule engine, and decision-making recommendation modules. The data management module stores and calls feature samples from the edge. The rule engine module consists of label data, classification models, and fault alarms. Label data is mapped to feature samples. The feature samples are retrieved and sent to the classification model for inference to obtain predicted labels. Fault alarm results are then given based on the predicted labels. The decision-making recommendation module sends fault handling decision recommendations to the production site based on the fault alarm results of the rule engine. The details of building a cloud fault diagnosis platform in step 3) are as follows: The dataset The classification model is trained by dividing the dataset into training set and test set according to the ratio of 7:3. First, different hyperparameter values ​​are used to train the classification model. The optimal value of the hyperparameter is determined by combining the stability and convergence speed of the training process and the accuracy of the test set. Then, the classification model is trained using the determined hyperparameters to obtain the classification model. The classification accuracy of the classification model on the test set is used as the evaluation index. The classification model with the highest classification accuracy is selected as the final classification model. The accuracy calculation formula is:

[0031] In the formula, Indicates the classification accuracy of the sample; Indicates the number of correctly classified samples in the test set; Represents the total number of test set samples.

[0032] Embodiment 2 See attached Figure 1The present invention provides a cloud-edge-coordinated rotating machinery intelligent fault diagnosis method, which mainly comprises the following steps: 1) Build a cloud-edge collaborative framework, that is, use the cloud and edge collaborative working mechanism. The schematic diagram of the cloud-edge collaborative framework is as follows: Figure 2 As shown in the figure, the edge is responsible for collecting data, executing signal processing algorithms and transmitting feature data to the cloud. The cloud is responsible for performing data classification on the feature data from the edge, obtaining fault diagnosis results, and sending fault alarm information. Specifically: 1.1) The edge signal processing platform includes sensors, conditioning circuits, analog-to-digital conversion units, microcontrollers and wireless transmission modules to realize signal acquisition, edge computing and data transmission functions. Signal acquisition is realized by sensors, conditioning circuits and analog-to-digital conversion units. Sensors include piezoelectric sensors and eddy current sensors, which respectively collect vibration acceleration signals and displacement signals. The conditioning circuit is mainly composed of coupling capacitors to eliminate the DC bias caused by the excitation source required for the sensor to work. The analog-to-digital conversion unit is composed of an analog-to-digital conversion chip (AD7606) and peripheral circuits to realize the conversion of analog signals to digital signals. Edge computing is realized by a microcontroller unit (MCU; STM32F407) to complete the execution of signal processing algorithms and feature extraction algorithms. AD7606 and MCU are connected using a serial peripheral interface (SPI) communication method. Data transmission uses a WIFI module to realize two-way transmission of edge and cloud data. 1.2) The cloud-based fault diagnosis platform consists of data management, rule engine and decision-making recommendation modules. The data management module stores, calls and displays waveforms of data from the edge, performs spectral analysis on the original data and displays edge features. The rule engine module retrieves feature samples and feeds them into the classification model for inference to obtain prediction labels. The classification model is a pre-trained convolutional neural network model. The decision-making recommendation module gives fault results based on the prediction labels obtained by the rule engine, and then sends disposal decision recommendations to the production site, thereby realizing cloud-edge collaborative fault diagnosis.

[0033] 2) Build an edge signal processing platform, the steps are as follows: 2.1) Construct a rotor-bearing system fault dataset, set the motor speed to 1400r / min, and set the magnetic powder brake to 0A and 0.5A currents to simulate no-load and 25N·m load. Collect 5 sets of data for each type of fault under each working condition. The length of a single set of data is 2048, so the number of samples for each type of fault under each working condition is 250, and the total number of samples is 2×6×250=3000. Adopt a load-insensitive strategy to divide the 3000 samples into dataset A and dataset B. Dataset A contains 400 samples of each type of fault, for a total of 2400 samples. Dataset B contains 100 samples of each type of fault, for a total of 600 samples. Divide the samples in dataset A into training sets and test sets for model training and testing. Use dataset B only for testing to simulate data collected in real time at the edge; 2.2) Construct an empirical mode decomposition algorithm suitable for edge computing, and decompose the samples of data set A and data set B into several IMFs and a residual component. Each IMF represents a different frequency component in the signal, and the residual component represents the trend of the signal. The decomposition calculation formula is:

[0034] In the formula, For samples; For time; is the component of the empirical mode decomposition result, ; is the number of empirical mode decomposition levels; is the residual component; 2.3) Construct a fast Fourier transform algorithm suitable for edge computing, perform fast Fourier transform on the IMF component and residual component after empirical mode decomposition, and use it to calculate the frequency domain feature index; 2.4) Construct a feature extraction algorithm suitable for edge computing to produce feature samples and extract time domain features for each IMF component ,Right now

[0035] in, is the maximum value; is the maximum absolute value; is the minimum value; is the mean; is the peak-to-peak value; is the absolute average value; is the root mean square value; is the square root amplitude; is the standard deviation; is the kurtosis value; is the skewness value; is the margin indicator; It is a waveform indicator; It is a pulse indicator; is the peak index; extract frequency domain index ,Right now

[0036] is the center of gravity frequency; is the average frequency; is the root mean square frequency; is the frequency variance; is the power spectrum entropy; and the time domain features and frequency domain characteristics Concatenate into feature vectors , that is, the eigenvector of a single component can be expressed as , the eigenvectors of all IMF components and residual components Sequentially stitch together into feature samples After feature extraction, the samples of dataset A and dataset B are used to obtain the feature sample set and ; 2.5) Construct a normalization algorithm suitable for edge computing, perform maximum and minimum normalization operations on feature samples, and eliminate the magnitude differences between different features in feature samples. The maximum and minimum normalization calculation formula is:

[0037] In the formula, , is the number of features extracted, is the number of levels of empirical mode decomposition; is the normalized feature sample; is the element in the feature sample; is the maximum value of the feature sample; is the minimum value of the feature sample; first, Perform normalization operation and then use the feature sample set The maximum value vector (MAX) and the minimum value vector (MIN) of the feature sample set during normalization operation Perform normalization operation, and After normalization, we get the data set and .

[0038] 3) Build a cloud-based fault diagnosis platform. The convolutional neural network model is constructed as follows: The dataset The classification model is trained by dividing the dataset into training set and test set according to the ratio of 7:3. First, different hyperparameter values ​​are used to train the classification model. The optimal value of the hyperparameter is determined by combining the stability and convergence speed of the training process and the accuracy of the test set. Then, the classification model is trained using the determined hyperparameters to obtain the classification model. The specific parameter settings of each layer of the final model are shown in Table 1. The classification accuracy of the classification model on the test set is used as the evaluation index, and the classification model with the highest classification accuracy is selected as the final classification model. The accuracy calculation formula is:

[0039] In the formula, Indicates the classification accuracy of the sample; Indicates the number of correctly classified samples in the test set; Represents the total number of test set samples.

[0040] Table 1 Parameter settings of each layer of the convolutional neural network model

[0041] Embodiment 3 A cloud-edge collaborative rotating machinery intelligent fault diagnosis system, comprising: Cloud-edge collaboration framework: used to build a collaborative working mechanism between the cloud and edge. The edge is responsible for collecting data, extracting features, and transmitting data to the cloud. The cloud is responsible for performing data classification on the feature data from the edge, obtaining fault diagnosis results, and generating fault alarm information. Edge signal processing platform: including sensors, conditioning circuits, analog-to-digital conversion units, microcontrollers and wireless transmission modules. Sensors, conditioning circuits and analog-to-digital conversion units are used to collect the original signals of rotating mechanical equipment. Microcontrollers are used to extract the features of the original signals, and the features are spliced ​​to obtain feature samples. The wireless transmission module is used to upload the feature samples from the edge to the cloud. Figure 3 Provides a visual interface for edge processing; Cloud fault diagnosis platform: The cloud fault diagnosis platform consists of data management, rule engine and decision suggestion modules. The data management module stores and calls feature samples from the edge. The rule engine module consists of label data, classification model and fault alarm. Label data is mapped to feature samples. The feature samples are called and sent to the classification model for inference to obtain the predicted label, and then the attached information is given according to the obtained predicted label. Figure 4 As shown in the fault alarm results, the decision suggestion module sends fault handling decision suggestions to the production site based on the fault alarm results of the rule engine.

[0042] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0043] 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 flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, 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 flowchart and / or block diagram. 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.

[0044] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate 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 A function specified in one or more boxes.

[0045] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0046] 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 its protection scope. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the field should understand that after reading the present invention, those skilled in the art can still make various changes, modifications or equivalent substitutions to the specific implementation methods of the invention, but these changes, modifications or equivalent substitutions are all within the protection scope of the pending claims of the invention.

Claims

1. A cloud-edge collaborative rotating machinery intelligent fault diagnosis method, characterized in that: The following steps are involved: Step 1) Build a cloud-edge collaboration framework Using the cloud and edge collaborative working mechanism, the edge is responsible for collecting data, extracting features and transmitting data to the cloud, and the cloud is responsible for performing data classification on the feature data from the edge, obtaining fault diagnosis results and generating fault alarm information; Step 2) Build an edge signal processing platform The edge signal processing platform includes sensors, conditioning circuits, analog-to-digital conversion units, microcontrollers, and wireless transmission modules. Sensors, conditioning circuits, and analog-to-digital conversion units are used to collect the original signals of rotating mechanical equipment. The microcontroller is used to extract the features of the original signals, and the features are spliced ​​to obtain feature samples. The wireless transmission module is used to upload the feature samples from the edge to the cloud. Step 3) Build a cloud-based fault diagnosis platform The cloud-based fault diagnosis platform consists of a data management module, a rule engine module, and a decision-making recommendation module. The data management module stores and calls feature samples from the edge. The rule engine module consists of label data, classification models, and fault alarms. The label data is mapped to feature samples, and the feature samples are retrieved and sent to the classification model for inference to obtain the predicted labels. The fault alarm results are then given based on the predicted labels. The decision-making recommendation module sends fault handling decision recommendations to the production site based on the fault alarm results of the rule engine.

2. The cloud-edge collaborative rotating machinery intelligent fault diagnosis method according to claim 1 is characterized in that: In the step 2) of constructing an edge signal processing platform, the sensors are various sensors capable of collecting one-dimensional time series signals, and the multi-source physical quantities collected by the sensors are converted into voltage analog signals. The conditioning circuit filters the voltage analog signals and removes DC bias. The analog-to-digital conversion unit converts the voltage analog signals processed by the conditioning circuit into digital output signals at a preset sampling frequency. The microcontroller has double-precision floating-point operation capabilities, and can realize feature extraction and feature splicing of the original signals.

3. The cloud-edge collaborative rotating machinery intelligent fault diagnosis method according to claim 2 is characterized in that: The preset sampling frequency is 5kHz-25kHz.

4. The cloud-edge collaborative rotating machinery intelligent fault diagnosis method according to claim 1 is characterized in that: The step 2) extracts the features of the original signal, and the steps are as follows: Step 2.1) Construct a fault data set: divide the collected original signal into segments, that is, divide the data of each type of fault state into N segments, where each segment is a sample of length n, and divide the data set into data set A and data set B according to a preset ratio based on the sample as a unit; Step 2.2) Construct an empirical mode decomposition algorithm suitable for edge computing: decompose the sample data of dataset A and dataset B into several IMFs and a residual component. Each IMF represents a different frequency component in the sample, and the residual component represents the trend of the sample. The decomposition calculation formula is: In the formula, For samples; For time; is the result component of empirical mode decomposition, ; is the number of empirical mode decomposition levels; is the residual component; Step 2.3) Construct a fast Fourier transform algorithm suitable for edge computing: perform fast Fourier transform on the IMF components and residual components after empirical mode decomposition, and use the transformation results to calculate frequency domain feature indicators; Step 2.4) Construct a feature extraction algorithm suitable for edge computing: extract time domain features for each IMF component and residual component and frequency domain characteristics , and the time domain features and frequency domain characteristics Concatenate into feature vectors , the eigenvectors of all IMF components and residual components Sequentially stitch together into feature samples After feature extraction, the samples of dataset A and dataset B are used to obtain the feature sample set and ; Step 2.5) Construct a normalization algorithm suitable for edge computing: perform maximum and minimum normalization operations on feature samples to eliminate the magnitude differences between different features in feature samples. The maximum and minimum normalization calculation formula is: In the formula, , is the number of features extracted for a single IMF component, is the number of levels of empirical mode decomposition; is the normalized feature sample; is the element in the feature sample; is the maximum value of the feature sample; is the minimum value of the feature sample; first, Perform normalization operation and then use the feature sample set The maximum value vector MAX and the minimum value vector MIN for the feature sample set during normalization operation Perform normalization operation, and After normalization, we get the data set and .

5. The cloud-edge collaborative rotating machinery intelligent fault diagnosis method according to claim 4 is characterized in that: The sample of length n in step 2.1) contains at least information of one change cycle of the rotating machinery, and the preset ratio is 4:

1.

6. The cloud-edge collaborative rotating machinery intelligent fault diagnosis method according to claim 4 is characterized in that: In step 2.4), the extracted time domain features can reflect the changes of the signal over time and are used to detect instantaneous changes or shocks in the signal.

7. The cloud-edge collaborative rotating machinery intelligent fault diagnosis method according to claim 4 is characterized in that: In step 2.4), the extracted frequency domain features can reflect the changes in the signal in the spectrum and are used to analyze the changes in the periodicity, harmonic components and frequency components of the signal.

8. The cloud-edge collaborative rotating machinery intelligent fault diagnosis method according to claim 4 is characterized in that: In step 3) of building a cloud fault diagnosis platform, the classification model is constructed as follows: Divide into training set and test set, train the classification model, first take different hyperparameter values ​​to train the classification model, combine the stability and convergence speed of the training process and the accuracy of the test set to determine the optimal value of the hyperparameter, then use the determined hyperparameter to train the classification model to obtain the data classification model, use the classification accuracy of the data classification model on the test set as the evaluation index, select the classification model with the highest classification accuracy as the final classification model, the classification accuracy calculation formula is: In the formula, Indicates the classification accuracy of the sample; Indicates the number of correctly classified samples in the test set; Represents the total number of test set samples.

9. The cloud-edge collaborative rotating machinery intelligent fault diagnosis method according to claim 8 is characterized in that: The dataset The dataset is divided into training set and test set according to the ratio of 7:

3.

10. A cloud-edge collaborative rotating machinery intelligent fault diagnosis system, characterized in that: include: Cloud-edge collaboration framework: Build a collaborative working mechanism between the cloud and edge. The edge is responsible for collecting data, extracting features, and transmitting data to the cloud. The cloud is responsible for performing data classification on the feature data from the edge, obtaining fault diagnosis results, and generating fault alarm information. Edge signal processing platform: including sensors, conditioning circuits, analog-to-digital conversion units, microcontrollers and wireless transmission modules. Sensors, conditioning circuits and analog-to-digital conversion units are used to collect the original signals of rotating mechanical equipment. Microcontrollers are used to extract the features of the original signals, and the features are spliced ​​to obtain feature samples. The wireless transmission module is used to upload the feature samples from the edge to the cloud. Cloud-based fault diagnosis platform: The cloud-based fault diagnosis platform consists of data management, rule engine and decision-making recommendation modules. The data management module stores and calls feature samples from the edge. The rule engine module consists of label data, classification models and fault alarms. The label data is mapped to feature samples, and the feature samples are fed into the classification model for inference to obtain the predicted labels. The fault alarm results are then given based on the predicted labels. The decision-making recommendation module sends fault handling decision recommendations to the production site based on the fault alarm results of the rule engine.

Images