Ship bearing fault diagnosis method based on multichannel data
Through multi-channel data processing and deep learning technology, a ship bearing fault diagnosis model is built, which solves the problem of insufficient diagnostic accuracy in the existing technology, and achieves more efficient fault identification and prediction.
Patent Information
- Application Number
- CN202510499716.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-04-21
- Publication Date
- 2025-08-01
AI Technical Summary
The existing ship bearing fault diagnosis methods based on single vibration or temperature signals have insufficient feature extraction depth and weak fault recognition capabilities, which makes it difficult to improve the diagnostic accuracy.
Multi-channel data is used to build a ship bearing fault diagnosis model. By collecting and processing vibration signals and temperature signals, using convolutional neural networks and long and short-term memory networks to extract images and timing features, combined with local polynomial fitting to process temperature signals, a fully connected layer is built for fault judgment.
It improves the accuracy and comprehensiveness of fault diagnosis, can better identify early weak faults, and enhances the ability to predict faults in complex environments.
Smart Images
Figure CN120408309A_ABST
Abstract
Description
Technical Field
[0001] This application belongs to the technical field of shipbuilding industry, and more specifically, relates to a ship bearing fault diagnosis method based on multi-channel data. Background Art
[0002] In modern shipbuilding industry, bearings, as key components in the ship power system, their operating status directly affects the safety and reliability of ships. Affected by high loads, high vibrations and high humidity and salt environments, bearings are extremely prone to early fatigue and lubrication failure.
[0003] Existing fault diagnosis methods based on single vibration or temperature signals still have deficiencies in the depth of feature extraction and the ability to identify weak faults, which limits the further improvement of the diagnosis accuracy. Summary of the Invention
[0004] In view of the above defects or improvement requirements of the prior art, this application provides a ship bearing fault diagnosis method based on multi-channel data, aiming to solve the technical problem that it is difficult to further improve the accuracy of existing ship bearing fault diagnosis methods.
[0005] To achieve the above object, in the first aspect, this application provides a ship bearing fault diagnosis model training method based on multi-channel data, including: Collect vibration signals and temperature signals of the bearing under various operating conditions; Perform envelope detection on the vibration signal to obtain the first-channel data; use the vibration signal as the second-channel data; perform local polynomial fitting on the temperature signal to obtain the third-channel data; Construct a convolutional neural network to extract image features from the three-channel data, construct a long short-term memory network to extract temporal features from the image features, and finally integrate the features through a fully connected layer to obtain a fault judgment result; Use the difference between the fault judgment result and the actual fault as a loss function to train each network parameter, and finally obtain the ship bearing fault diagnosis model.
[0006] Preferably, constructing a convolutional neural network to extract image features from the three-channel data specifically includes: Align the three-channel data in time series to obtain 3 one-dimensional data with dimensions of where is the time series length; Stack the 3 one-dimensional data to obtain 1 two-dimensional data with dimensions of ; Construct a convolutional neural network to extract image features from the two-dimensional data.
[0007] Preferably, constructing a long short-term memory network to extract temporal features from the image features specifically includes: Use a flattening layer to reduce the dimensionality of the image features to one dimension; Input the one-dimensional image features into a long short-term memory network to extract temporal features.
[0008] Preferably, perform envelope detection on the vibration signal to obtain the first-channel data, specifically: Perform band-pass filtering on the vibration signal; Perform low-noise amplification on the filtered signal; Perform Hilbert transform on the amplified signal to obtain an envelope signal.
[0009] Preferably, perform local polynomial fitting on the temperature signal to obtain the third-channel data, specifically: Preset the window length and polynomial order; Gradually slide the window over the temperature signal in time series; For each window sliding, perform polynomial fitting on the temperature data points within the window and replace the original temperature data with the fitted data.
[0010] Preferably, during the fault diagnosis process of the ship bearing fault diagnosis model, continuously collect the vibration signal, temperature signal of the bearing and the corresponding state of the bearing, and at preset time intervals, mix the newly collected data and the original data set in a preset ratio and retrain the model to update the model parameters.
[0011] Preferably, during the training process of the ship bearing fault diagnosis model, gradually reduce the learning rate using cosine annealing and perform gradient descent using the mini-batch gradient descent method.
[0012] In a second aspect, the present application provides a ship bearing fault diagnosis method based on multi-channel data, including: Collect the vibration signal and temperature signal of the bearing in real time; Perform envelope detection on the vibration signal to obtain the first-channel data; use the vibration signal as the second-channel data; perform local polynomial fitting on the temperature signal to obtain the third-channel data; Input the three-channel data into the ship bearing fault diagnosis model to obtain the operating state of the ship bearing; The ship bearing fault diagnosis model is trained by any of the training methods in the first aspect.
[0013] In a third aspect, the present application provides an electronic device, including: at least one memory for storing a program; at least one processor for executing the program stored in the memory. When the program stored in the memory is executed, the processor is used to execute any of the methods described in the first aspect, or the processor is used to execute the method described in the second aspect.
[0014] Fourthly, the present application provides a computer-readable storage medium storing a computer program, which, when running on a processor, causes the processor to execute any of the methods described in the first aspect or the method described in the second aspect.
[0015] Generally speaking, compared with the prior art, the above technical solutions conceived by the present application have the following beneficial effects: (1) The present application constructs three-channel data based on multi-modal data during the operation of the bearing for fault prediction, and it is easier to extract more comprehensive, deep and subtle fault features from the three-channel data; the first-channel data is the envelope signal based on the vibration signal, which is more suitable for feature extraction of early weak vibration sensitive faults; the second-channel data is based on the vibration signal, which is more suitable for feature extraction of conventional vibration sensitive faults, and the third-channel data is based on the temperature signal, which is more suitable for feature extraction of temperature sensitive faults. The fault prediction based on the above three-channel data is more comprehensive and accurate.
[0016] (2) The present application uses a convolutional neural network to obtain image features from the three-channel data, and then uses a long short-term memory network to further extract the temporal features ignored by the convolutional neural network, so that the fault can be more comprehensively characterized by features, thereby improving the fault diagnosis accuracy of the fault diagnosis model.
[0017] (3) In the present application, the three-channel data is converted from one-dimensional data to two-dimensional data through time series alignment and stacking. The two-dimensional data is similar to image data and is more suitable for feature extraction by a convolutional neural network; and this stacking method retains the continuity of the data in time series and does not destroy the temporal features of the data, which is convenient for the subsequent long short-term memory network to extract the temporal features.
[0018] (4) During the operation of the ship, various complex factors in the ship's engine room will affect the acquisition of the bearing temperature, resulting in noisy and chaotic bearing temperature data collected, which in turn affects the fault prediction based on the temperature data. In the present application, local polynomial fitting is performed on the original signal of the collected temperature, which can effectively remove the noise signal in the temperature signal and retain the original features of the temperature signal, providing a real temperature data basis for subsequent fault prediction. Description of the Drawings
[0019] Figure 1 It is a schematic diagram of the training process of the ship bearing fault diagnosis model provided by the embodiment of the present application.
[0020] Figure 2 It is a schematic diagram of the original vibration data provided by the embodiment of the present application.
[0021] Figure 3 It is a schematic diagram of the original temperature data provided by the embodiment of the present application.
[0022] Figure 4 It is a schematic diagram of the 3-modal data construction process provided by an embodiment of the present application.
[0023] Figure 5 It is a schematic diagram of the structure of a ship bearing fault diagnosis model provided by an embodiment of the present application.
[0024] Figure 6 It is a schematic diagram of the structure of an electronic device provided by an embodiment of the present application. Detailed implementation manners
[0025] In order to make the objectives, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application, but not to limit the present application.
[0026] The terms "first" and "second" etc. in the description and claims of this article are used to distinguish different objects, rather than to describe a specific order of the objects. For example, the first channel data and the second channel data etc. are used to distinguish different channel data, rather than to describe the specific order of the channel data.
[0027] In the embodiments of the present application, words such as "exemplary" or "for example" are used to represent examples, illustrations or explanations. Any embodiment or design solution described as "exemplary" or "for example" in the embodiments of the present application should not be construed as being more preferred or having more advantages than other embodiments or design solutions. Rather, the use of words such as "exemplary" or "for example" is intended to present relevant concepts in a specific manner.
[0028] In the description of the embodiments of the present application, unless otherwise specified, the meaning of "a plurality" refers to two or more. For example, a plurality of channel data refers to two or more channel data etc.
[0029] The embodiments of the present application will be described below with reference to the accompanying drawings in the embodiments of the present application.
[0030] Embodiment 1 of the present application is a method for training a ship bearing fault diagnosis model based on multi-channel data, as Figure 1 shown, and specifically includes the following steps: S1. Acquisition of the data set: S11. Acquisition of the original signal: Obtain the vibration signal of the propulsion bearing through a vibration sensor deployed in the vibration sensitive area of the ship propulsion bearing; obtain the temperature signal of the bearing through an infrared sensor deployed in the temperature sensitive area of the ship propulsion bearing.
[0031] Obtain the vibration signal and temperature signal of the propulsion bearing in various operating states through the sensor,Figure 2 and Figure 3 The waveforms of the vibration signal and the temperature signal are shown. It can be seen that the signals are relatively messy, and the subsequent steps need to process the original signals.
[0032] S12. Signal processing: First, perform data cleaning on the vibration signal and the temperature signal to remove invalid outliers.
[0033] Then, preprocess the vibration signal and the temperature signal to remove the trend terms and retain the periodic components.
[0034] The processing process after preprocessing is as Figure 4 shown: Perform band-pass filtering on the vibration signal to focus on the fault feature frequency band range and suppress the interference of signals in irrelevant frequency bands. Subsequently, perform low-noise amplification on the filtered signal to amplify the weak fault impact signal and improve the signal-to-noise ratio. Finally, extract the fault feature signal on the basic vibration signal through envelope detection and then perform AD sampling to obtain the data of the first channel. The data of the first channel obtained through envelope detection can better characterize the early weak vibration sensitive faults.
[0035] Perform low-pass filtering on the vibration signal, which can effectively filter out high-frequency noises such as electromagnetic interference and can retain the vibration signal in the fault feature frequency band. Subsequently, perform AD sampling on the filtered vibration signal to ensure obtaining the complete second-channel data of the time series waveform. The second-channel data can characterize the conventional vibration sensitive faults.
[0036] Perform low-pass filtering on the temperature signal, which can effectively filter out the high-frequency noise in the temperature signal. Subsequently, perform AD sampling on the filtered temperature signal to obtain the data of the third channel. The data of the third channel can better characterize the temperature sensitive faults.
[0037] S13. Construct a dataset: Align the data of the first channel, the second channel, and the third channel obtained in the same time period in time series to obtain three one-dimensional data with a size of where is the time series length; Stack the three one-dimensional data to obtain a two-dimensional data of size ; Obtain multiple two-dimensional data from the 3-channel data of multiple time periods, and mark the bearing operating state of this time period as a label on the data. ;
[0038] The dataset is composed of all two-dimensional data with state labels.
[0039] S2. Construction of a ship bearing fault diagnosis model: As Figure 5As shown in the figure, the model of this embodiment includes a convolutional neural network part and a long short-term memory network part, and the convolutional neural network part and the long short-term memory network part are connected in series.
[0040] The convolutional neural network part includes two convolutional-pooling layers, and the long short-term memory network part includes two layers of long short-term memory networks (LSTM) and two fully connected layers.
[0041] The two convolutional-pooling layers can gradually extract the image features that can characterize the running state of the bearing from the two-dimensional data. The convolutional process is as follows:
[0042] Among them, is the output of the neuron in the convolutional layer, represents the activation function, represents the neuron input, represents the neuron weight, represents the neuron bias.
[0043] After convolution, the convolution result needs to be normalized and then pooled:
[0044] Among them is the normalized value, is the value before normalization, and are the maximum and minimum values.
[0045] The image features obtained after two layers of convolutional-pooling are flattened by a flattening layer to reduce the dimensionality of the image features to one dimension; then the one-dimensional image features are sequentially input into two serially connected long short-term memory networks to extract temporal features.
[0046] The temporal features are input into two fully connected layers for feature integration to obtain the classification results of various states of the bearing, and fault diagnosis can be performed according to the state classification results.
[0047] S3. Training of the ship bearing fault diagnosis model: The training set is divided from the dataset to train the ship bearing fault diagnosis model. The classification results of the output state of the model are compared with the state labels corresponding to the training set, and the model is trained according to the gap to optimize the model parameters. In this embodiment, the mean square error is used as the loss function to measure the gap:
[0048] Among them, is the mean square error, is the true value, is the predicted value, is the number of samples, with the subscript being the sample sequence number.
[0049] During the training process, the method of mini-batch gradient descent is used for gradient descent, which can enable the model to jump out of the local optimal solution.
[0050] During the training process, the cosine annealing is used to gradually reduce the learning rate. This cosine learning rate has a very good effect on jumping out of the local optimal solution, making the trained network model more superior.
[0051] Finally, the trained ship bearing fault diagnosis model is used for fault diagnosis, which specifically includes the following steps: Step 1: Real-time collect the vibration signal and temperature signal of the bearing; Step 2: Perform envelope detection on the vibration signal to obtain the first-channel data; use the vibration signal as the second-channel data; perform local polynomial fitting on the temperature signal to obtain the third-channel data; Step 3: Input the three-channel data into the ship bearing fault diagnosis model to obtain the operating state of the ship bearing.
[0052] During the process of using the ship bearing fault diagnosis model for fault diagnosis, continuously collect the vibration signal, temperature signal of the bearing and the corresponding state of the bearing, and at a preset time interval, such as a 7-day interval in a week, mix the newly collected data and the original data set according to a preset ratio and retrain the model, such as a 1:1 ratio, to update the model parameters. Thus, the ship bearing fault diagnosis model can adapt to the changes in the current production environment and be continuously optimized.
[0053] Experimentally compare the accuracy rates of the embodiments of this application and other fault diagnosis models: Using the same data set, train the fault diagnosis model with different models. The specific experimental data is shown in Table 1: Table 1
[0054] It can be seen that the accuracy rate of the ship bearing fault diagnosis model of this application is higher than that of other models.
[0055] Based on the method in the above embodiments, an embodiment of this application provides an electronic device, as Figure 6 shown. This electronic device includes: a processor, a communications interface, a memory, and a communication bus. Among them, the processor, the communications interface, and the memory complete communication with each other through the communication bus. The processor can call the logical instructions in the memory to execute the method in the above embodiments.
[0056] In addition, when the logical instructions in the above-mentioned memory are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing 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 methods described in various embodiments of this application.
[0057] Based on the method in the above embodiments, an embodiment of this application provides a computer-readable storage medium. The computer-readable storage medium stores a computer program. When the computer program runs on a processor, it causes the processor to execute the method in the above embodiments.
[0058] Based on the method in the above embodiments, an embodiment of this application provides a computer program product. When the computer program product runs on a processor, it causes the processor to execute the method in the above embodiments.
[0059] It can be understood that the processor in the embodiments of this application can be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application specific integrated circuits (ASICs), field programmable gate arrays (FPGAs), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. The general-purpose processor can be a microprocessor or any conventional processor.
[0060] The method steps in the embodiments of the present application can be implemented in a hardware manner or by a processor executing software instructions. The software instructions can be composed of corresponding software modules, and the software modules can be stored in a random access memory (RAM), flash memory, read-only memory (ROM), programmable ROM (PROM), erasable PROM (EPROM), electrically erasable PROM (EEPROM), registers, hard disks, removable hard disks, CD-ROMs, or any other form of storage medium well-known in the art. An exemplary storage medium is coupled to the processor so that the processor can read information from the storage medium and write information to the storage medium. Of course, the storage medium can also be a component of the processor. The processor and the storage medium can be located in an ASIC.
[0061] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the processes or functions described in the embodiments of the present application are generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium or transmitted through the computer-readable storage medium. The computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center in a wired manner (such as coaxial cable, optical fiber, digital subscriber line (DSL)) or a wireless manner (such as infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that the computer can access or a data storage device such as a server, data center, etc. that includes one or more integrated available media. The available medium can be a magnetic medium (such as a floppy disk, hard disk, magnetic tape), an optical medium (such as a DVD), or a semiconductor medium (such as a solid state disk (SSD)).
[0062] It can be understood that the various numerical numbers involved in the embodiments of the present application are only for the convenience of description and are not used to limit the scope of the embodiments of the present application.
[0063] Those skilled in the art can easily understand that the above are only preferred embodiments of the present application and are not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present application shall be included within the protection scope of the present application.
Claims
1. A method for training a ship bearing fault diagnosis model based on multi-channel data, characterized in that, Including: Collecting vibration signals and temperature signals of the bearing under various operating conditions; Performing envelope detection on the vibration signal to obtain the first-channel data; using the vibration signal as the second-channel data; performing local polynomial fitting on the temperature signal to obtain the third-channel data; Constructing a convolutional neural network to extract image features from the three-channel data, constructing a long short-term memory network to extract temporal features from the image features, and finally integrating the features through a fully connected layer to obtain a fault judgment result; Taking the gap between the fault judgment result and the actual fault as a loss function to train each network parameter, and finally obtaining the ship bearing fault diagnosis model.
2. The method for training a ship bearing fault diagnosis model according to claim 1, wherein Constructing a convolutional neural network to extract image features from the three-channel data, specifically: Align the three-channel data in time sequence to obtain three one-dimensional data with dimensions of ; is the time sequence length. Stack three one-dimensional data to obtain a two-dimensional data of ; Constructing a convolutional neural network to extract image features of the two-dimensional data.
3. The method for training a ship bearing fault diagnosis model according to claim 1, characterized in that Constructing a long short-term memory network to extract temporal features from the image features, specifically: Using a flattening layer to reduce the dimension of the image features to one dimension; Inputting the one-dimensional image features into the long short-term memory network to extract temporal features.
4. The method for training a ship bearing fault diagnosis model according to claim 1, wherein, Performing envelope detection on the vibration signal to obtain the first-channel data, specifically: Performing band-pass filtering on the vibration signal; Performing low-noise amplification on the filtered signal; Performing Hilbert transform on the amplified signal to obtain an envelope signal.
5. The method for training a ship bearing fault diagnosis model according to claim 1, wherein, Performing local polynomial fitting on the temperature signal to obtain the third-channel data, specifically: Presetting a window length and a polynomial order; Gradually sliding the window over the temperature signal in time series; Each time the window is slid, performing polynomial fitting on the temperature data points within the window and replacing the original temperature data with the fitted data.
6. The method for training a ship bearing fault diagnosis model according to claim 1, wherein, During the process of fault diagnosis by the ship bearing fault diagnosis model, continuously collecting the vibration signal, temperature signal and the corresponding state of the bearing, and at preset time intervals, mixing the newly collected data and the original data set according to a preset ratio and retraining the model to update the model parameters.
7. The method for training a ship bearing fault diagnosis model according to claim 1, characterized in that During the training process of the ship bearing fault diagnosis model, gradually reducing the learning rate by cosine annealing and using the method of mini-batch gradient descent for gradient descent.
8. A method for diagnosing faults of ship bearings based on multi-channel data, characterized in that, Including: Real-time collecting the vibration signal and temperature signal of the bearing; Performing envelope detection on the vibration signal to obtain the first-channel data; using the vibration signal as the second-channel data; performing local polynomial fitting on the temperature signal to obtain the third-channel data; Inputting the three-channel data into the ship bearing fault diagnosis model to obtain the operating state of the ship bearing; The ship bearing fault diagnosis model is trained by the training method as described in any one of claims 1-7.
9. An electronic device, characterized in that, Including: At least one memory for storing a computer program; At least one processor for executing the program stored in the memory. When the program stored in the memory is executed, the processor is used to execute the method as described in any one of claims 1-7, or execute the method as described in claim 8.
10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program runs on the processor, the processor is caused to execute the method as described in any one of claims 1-7, or execute the method as described in claim 8.