Airport luggage system gear motor fault diagnosis method and system, and medium
By constructing and training the WDCNN model, the fault diagnosis of shafts, bearings and gearboxes in the gear reduction motor system can be solved, and the problem that the existing technology cannot judge these components at the same time is solved, and the advance perception and prevention of gear reduction motor failure in the luggage system is achieved. .
Patent Information
- Application Number
- CN202510000348.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-01-02
- Publication Date
- 2025-02-07
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
The existing technical solutions can only diagnose bearings, and cannot judge the shaft, bearing and gearbox in the gear reduction motor system.
By obtaining the operating data of the reducer motor, preprocessing and building the original data set, building the WDCNN model, and training the model through the original data set, a multi-dimensional fault diagnosis algorithm model is obtained. The operating data of the gearbox to be diagnosed is input into the multi-dimensional fault diagnosis algorithm model for fault diagnosis to obtain fault diagnosis results.
The fault diagnosis of shafts, bearings and gear boxes in the gear reduction motor system is realized, and the fault of the luggage system gear reduction motor can be sensed in advance, avoid unexpected downtime, and reduce economic losses.
Smart Images
Figure CN119397405A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of reduction motor fault diagnosis, and in particular to a reduction motor fault diagnosis method, system and medium for an airport baggage system. Background Art
[0002] As an important component of large baggage systems, the stability and safety of the shafts, bearings and gearboxes of reduction motors are of great concern. If they are at risk of failure due to faults or aging and are not discovered early, causing single-point or continuous shutdown of the conveyor line, it will seriously affect the system's transportation efficiency and cause unnecessary economic losses.
[0003] At present, the bearing fault diagnosis method and system based on WDCNN-LSTM mainly obtain the vibration signal of the bearing; obtain the bearing fault diagnosis result according to the vibration signal and the trained bearing fault diagnosis model; wherein the bearing fault diagnosis model includes a spatial feature extraction path and a temporal feature extraction path; the temporal feature extraction path includes a convolutional layer and an LSTM connected in sequence; the vibration signal is respectively input into the temporal feature extraction path and the spatial feature extraction path to extract the temporal feature and the spatial feature; the temporal feature and the spatial feature are fused to obtain the fused feature; the fused feature is identified to obtain the bearing fault diagnosis result; without preprocessing the vibration noise, the bearing fault can be accurately judged by identifying the vibration noise; however, the existing technical scheme can only perform fault diagnosis on the bearing, and cannot judge the heavy shaft, bearing and gearbox of the reduction motor system. Summary of the invention
[0004] The present invention provides a method, system and medium for diagnosing faults of a reduction motor of an airport baggage system, so as to solve the problem that the existing technical solutions can only diagnose faults of bearings but cannot diagnose the heavy shafts, bearings and gearboxes of the reduction motor system.
[0005] According to an embodiment of the present invention, a method for diagnosing a fault of a reduction motor of an airport baggage system comprises the following steps: Obtaining operation data of several groups of reduction motors; wherein the operation data includes: motor body data, shaft data, bearing data and gearbox damage data; Preprocess the running data and construct the original data set according to the rules; Construct a WDCNN model, and train the WDCNN model until convergence using the original data set to obtain a multi-dimensional fault diagnosis algorithm model; The operating data of the gear motor to be diagnosed is input into the multi-dimensional fault diagnosis algorithm model to perform fault diagnosis and obtain the fault diagnosis result.
[0006] As an embodiment of the present invention, the operation data is preprocessed and an original data set is constructed according to a rule, including: Convert the running data into a decimal vector group according to a preset protocol; Match the vector groups with the fault labels and construct the original dataset according to the rules; Among them, the fault tags include: motor body data tag, shaft damage data tag, bearing damage motor tag and gearbox damage motor tag.
[0007] As an embodiment of the present invention, the WDCNN model includes: Input layer, used to input operation data; The first convolutional layer, connected to the input layer, is used to capture the low-frequency features of the running data; The second convolution layer, the third convolution layer, and the fourth convolution layer are arranged side by side, and are all used to realize high-frequency feature extraction of running data; The input ends of the first pooling layer, the second pooling layer, the third pooling layer, and the fourth pooling layer are connected to the output ends of the first convolutional layer, the second convolutional layer, the third convolutional layer, and the fourth convolutional layer, respectively, and the output ends of the first pooling layer, the second pooling layer, and the third pooling layer are connected to the input ends of the second convolutional layer, the third convolutional layer, and the fourth convolutional layer, respectively, all of which are used to significantly reduce the magnitude of network parameters; A Relu activation layer is respectively arranged between the output end of the first pooling layer and the input end of the second convolutional layer, between the output end of the second pooling layer and the input end of the third convolutional layer, and between the output end of the third pooling layer and the input end of the fourth convolutional layer; BN normalization layers are respectively arranged between the output end of the first convolutional layer and the input end of the second pooling layer, between the output end of the second convolutional layer and the input end of the third pooling layer, and between the output end of the third convolutional layer and the input end of the fourth pooling layer; The fully connected layer is connected to the fourth pooling layer to integrate and classify the features extracted from each layer to obtain the extraction result; The Softmax activation layer is connected to the fully connected layer and is used to generate fault diagnosis results based on the extraction results; Dropout layer, located between the fourth pooling layer and the fully connected layer; The output layer is used to output fault diagnosis results.
[0008] As an embodiment of the present invention, the first convolutional layer is a wide convolutional layer; wherein the convolution kernel size of the first convolutional layer is 16*1 and the step size is 8*1.
[0009] As an embodiment of the present invention, the loss function used in the process of training the WDCNN model is as follows:
[0010] in, is the one-hot encoding of the sample target value, For the observed sample Belongs to category The predicted probability of is the sample size, .
[0011] On the other hand, according to an embodiment of the present invention, there is also provided a system for diagnosing faults of a reduction motor of an airport baggage system, comprising: Agile data acquisition module, used to obtain the operating data of several groups of reduction motors; The feature vector reorganization module is used to pre-process the operation data and construct the original data set according to the rules; The convolution model training module is used to build the WDCNN model and train the WDCNN model until convergence using the original data set to obtain a multi-dimensional fault diagnosis algorithm model; The multidimensional fault diagnosis module is used to input the operating data of the reduction motor to be diagnosed into the multidimensional fault diagnosis algorithm model to perform fault diagnosis and obtain the fault diagnosis result.
[0012] As an embodiment of the present invention, the data agile collection module includes: Gear motor; Vibration sensor, used to collect vibration data of reduction motor, shaft, bearing and reduction box damage respectively, and obtain motor body data, shaft data, bearing data and gear box damage data; The motor monitoring DCU box is electrically connected to the vibration sensor and the vibration sensor respectively, and is used to record and manage the motor body data, shaft data, bearing data and gearbox damage data.
[0013] On the other hand, the present invention further provides a computer-readable storage medium for storing computer instructions. When the computer instructions are executed by a processor, the steps described in the airport baggage system reduction motor fault diagnosis method disclosed in an embodiment of the present invention are completed.
[0014] Compared with the prior art, the present invention has the following beneficial effects: through the WDCNN model that can perform fault diagnosis on shafts, bearings and gearboxes, it is possible to effectively perceive the failure of the reduction motor of the baggage system in advance, which helps the operation and maintenance management personnel to grasp the operating status of the reduction motor at the first time, and make arrangements for replacement or purchase of spare motors in advance, thereby avoiding unexpected shutdown of the baggage system and unnecessary economic losses. BRIEF DESCRIPTION OF THE DRAWINGS
[0015] Figure 1 A schematic diagram of a method according to an embodiment of the present invention; Figure 2 This is a schematic diagram of the configuration of a data agile acquisition module according to another embodiment of the present invention; Figure 3 It is a structural schematic diagram of a multi-dimensional fault diagnosis algorithm model according to another embodiment of the present invention. DETAILED DESCRIPTION
[0016] The technical solution of the present invention is further described below in conjunction with the accompanying drawings and embodiments.
[0017] like Figures 1 to 3 As shown, the embodiment of the present invention proposes a method for diagnosing a fault of a reduction motor of an airport baggage system, comprising the following steps: Acquire operation data of several groups of reduction motors; wherein the operation data includes: motor body data, shaft data, bearing data and gearbox data; Preprocess the running data and construct the original data set according to the rules; Construct a WDCNN model, and train the WDCNN model until convergence using the original data set to obtain a multi-dimensional fault diagnosis algorithm model; Inputting the operating data of the gear motor to be diagnosed into the multi-dimensional fault diagnosis algorithm model to perform fault diagnosis and obtain the fault diagnosis result; The working principle of the above technical solution is as follows: in actual use, the motor body data, shaft data, bearing data and gearbox data of the reduction motor are collected respectively; then the operation data is processed to generate the original data set; the WDCNN model is trained with the original data to obtain a multi-dimensional fault diagnosis algorithm model that can perform fault diagnosis on the shaft, bearing and gearbox; Beneficial effects of the above technical solution: Through the above technical solution, through the multi-dimensional fault diagnosis algorithm model that can diagnose faults of shafts, bearings and gearboxes, it is possible to effectively perceive the failure of the reduction motor of the baggage system in advance, which helps the operation and maintenance management personnel to grasp the operating status of the reduction motor at the first time, and make arrangements for replacement or purchase of spare motors in advance, thereby avoiding unexpected shutdown of the baggage system and unnecessary economic losses.
[0018] In one embodiment, the operation data is preprocessed and an original data set is constructed according to a regular pattern, including: Convert the running data into a decimal vector group according to a preset protocol; Match the vector groups with the fault labels and construct the original dataset according to the rules; Among them, the fault labels include: no damage data label, motor body damage data label, shaft damage data label, bearing damage motor label and gearbox damage motor label; The working principle and beneficial effects of the above technical solution: In actual use, a one-dimensional feature vector group is formed by collecting the vibration signal of the reduction motor (operation data); the real-time operation vibration information of the motor body and shaft, bearing, and gearbox damage is collected and converted into a decimal vector group through a protocol, and the original data set is constructed according to the corresponding labels; in the process of training the WDCNN model with the original data set, the WDCNN model is repeatedly trained with the original data set. To facilitate the verification of this case, 20% of the original data set is randomly selected as a test set, 70% as a training set, and 10% as a test set. The test set is used as the validation set; the test set is input into the trained multidimensional fault diagnosis algorithm model, and the result is the gear motor fault diagnosis result; compared with the traditional CNN model, the simulation results show that WDCNN has converged at 200 iterations, and the accuracy rate is 95.2%, while CNN converges at nearly 500 iterations, and the accuracy rate is 88.3%, which is significantly weaker than the multidimensional fault diagnosis algorithm model; among them, when matching fault labels according to the vibration data of the gear motor, manual judgment is generally performed through Al technology or experts.
[0019] In one embodiment, the WDCNN model includes: Input layer, used to input operation data; The first convolutional layer, connected to the input layer, is used to capture the low-frequency features of the running data; The second convolution layer, the third convolution layer, and the fourth convolution layer are arranged side by side, and are all used to realize high-frequency feature extraction of running data; The input ends of the first pooling layer, the second pooling layer, the third pooling layer, and the fourth pooling layer are connected to the output ends of the first convolutional layer, the second convolutional layer, the third convolutional layer, and the fourth convolutional layer, respectively, and the output ends of the first pooling layer, the second pooling layer, and the third pooling layer are connected to the input ends of the second convolutional layer, the third convolutional layer, and the fourth convolutional layer, respectively, all of which are used to significantly reduce the magnitude of network parameters; A Relu activation layer is respectively arranged between the output end of the first pooling layer and the input end of the second convolutional layer, between the output end of the second pooling layer and the input end of the third convolutional layer, and between the output end of the third pooling layer and the input end of the fourth convolutional layer; BN normalization layers are respectively arranged between the output end of the first convolutional layer and the input end of the second pooling layer, between the output end of the second convolutional layer and the input end of the third pooling layer, and between the output end of the third convolutional layer and the input end of the fourth pooling layer; The fully connected layer is connected to the fourth pooling layer to integrate and classify the features extracted from each layer to obtain the extraction result; The Softmax activation layer is connected to the fully connected layer and is used to generate fault diagnosis results based on the extraction results; Dropout layer, located between the fourth pooling layer and the fully connected layer; The output layer is used to output fault diagnosis results.
[0020] The working principle and beneficial effects of the above technical solution: In actual use, the multidimensional fault diagnosis algorithm model mainly has 11 layers, mainly input layer, convolution layer, pooling layer, fully connected layer and output layer; among them, the convolution kernel size of the first convolution layer (wide convolution layer-1 in the attached figure) is larger than the subsequent second convolution layer (convolution layer-2 in the attached figure), the third convolution layer (convolution layer-3 in the attached figure) and the fourth convolution layer (convolution layer-4 in the attached figure), which can effectively exclude unclear feature information, thereby optimizing network training; among them, the wide convolution layer is used to capture the low-frequency features of the input signal, and other convolution layers use smaller convolution kernels to achieve High-frequency feature extraction is now performed; specifically, the convolution layer is designed to be 4 layers with 96 neurons, of which the convolution kernel size of the first convolution layer is 16*1 and the step size is 8*1, which can obtain a better receptive field, and the convolution kernel size of the remaining convolution layers is designed to be 8*1 and the step size is 2*1; at the same time, 4 pooling layers are set, namely the first pooling layer (pooling layer-1 in the attached figure), the second pooling layer (pooling layer-2 in the attached figure), the third pooling layer (pooling layer-3 in the attached figure) and the fourth pooling layer (pooling layer-4 in the attached figure), and the number of pooling elements in the pooling layer is designed to be 512; the pooling layer can avoid the overfitting problem by greatly reducing the magnitude of network parameters; Adding a BN normalization layer between the convolutional layer and the pooling layer for batch normalization can enhance the generalization ability of the network and avoid gradient explosion; the activation function is selected as Relu, that is, the Relu activation layer, to further solve the gradient problem. When the input value is greater than 0, its derivative is always 1, which can speed up the convergence of the model; adding a Dropout layer between the pooling layer and the fully connected layer, designed to be 0.5, effectively solves the overfitting problem; adding a softmax classifier after the fully connected layer for output, with a size of 4, corresponding to 4 types of outputs; at the same time, the model uses the Adam optimizer for gradient optimization, sets the learning rate to 0.001, and the total number of iterations to 500.
[0021] Among them, the convolution layer of the WDCNN model, Layer The feature output can be expressed by the following formula:
[0022] in, Indicates Layer The output of features, * indicates convolution calculation, Indicates The convolution weights corresponding to the features, Indicates bias, Indicates the number of channels in the previous layer, represents the activation function; After convolution, the feature size becomes ,in Indicates the number of filters in this layer, It can be calculated by the following formula:
[0023] in, Represents the length and width of the input, Indicates the filling method. represents the size of the convolution kernel, Indicates the step length.
[0024] Preferably, the pooling layer of the WDCNN first divides the features extracted by the convolution layer into multiple regions, and converts the features from large size to small size by means of maximum value or average value, etc., because the network parameters are reduced, the training complexity of the network is reduced, and this operation can also prevent overfitting; after the pooling operation, the feature size becomes , which can be expressed by the following formula:
[0025] in, Represents the length and width of the input, represents the size of the convolution kernel, represents the step length; The fully connected layer of the WDCNN model, that is, the one-dimensional features converted by pooling, are connected to the fully connected layer. In this way, the fault type can be diagnosed, and its forward propagation can be expressed by the following formula:
[0026] in, Indicates Tier The logits value of each neuron, Indicates Tier The neuron and Tier The weight of a neuron, Indicates All neurons in the layer The neuron pairs No. The bias of a neuron, Represents the input of the nth neuron in the tth layer; The activation layer of the WDCNN model usually uses ReLU, which is mainly used to enhance the feature learning ability of the network. The whole process can be expressed by the following formula:
[0027] in, Indicates the output, Represents input.
[0028] In one embodiment, the loss function used in the process of training the WDCNN model is as follows:
[0029] in, is the one-hot encoding of the sample target value, For the observed sample Belongs to category The predicted probability of is the sample size, .
[0030] Working principle and beneficial effects of the above technical solution: In order to facilitate the evaluation of the model, the loss function is designed as cross entropy; among them, the category It is one of the damages of shaft, bearing and gearbox; specifically, when collecting several groups of operation data, the collection interval is set to 15s, and the vibration data of 4 types of reduction motors, namely motor body and shaft, bearing and gearbox, are collected at a speed of 1430rpm; 4096 groups of data are collected for each of the 4 types of vibration data, for a total of 16384 groups of data, thus obtaining sample data.
[0031] The present invention also provides a fault diagnosis system for a reduction motor of an airport baggage system, comprising: Agile data acquisition module, used to obtain the operating data of several groups of reduction motors; The feature vector reorganization module is used to pre-process the operation data and construct the original data set according to the rules; The convolution model training module is used to build the WDCNN model and train the WDCNN model until convergence using the original data set to obtain a multi-dimensional fault diagnosis algorithm model; A multi-dimensional fault diagnosis module is used to input the operating data of the gear motor to be diagnosed into the multi-dimensional fault diagnosis algorithm model to perform fault diagnosis and obtain a fault diagnosis result; Agile data collection module, including: The vibration sensor is used to collect vibration data of the reduction motor, shaft, bearing and reduction box respectively, and obtain motor body data, shaft data, bearing data and gear box data; The motor monitoring DCU box is electrically connected to the vibration sensor and is used to record and manage motor body data, shaft data, bearing data and gearbox data.
[0032] The working principle and beneficial effects of the above technical solution: In actual use, a vibration sensor is arranged in the reduction motor system of the baggage system to collect motor body data, motor body data, shaft data, bearing data and gearbox data; for example, the vibration sensor is arranged on the reduction box of the reduction motor, and the threaded bolt at the head of the vibration sensor is screwed into the reserved hole of the reduction box or directly magnetically attracted to the outer wall of the reduction box to achieve full contact and fastening between the sensor and the reduction box; the vibration sensor is mainly responsible for collecting healthy and unhealthy vibration data of the reduction motor of the baggage system; the motor monitoring DCU is arranged next to a conveyor in the monitoring area according to the actual number of connected reduction motors; the motor monitoring DCU includes a 24V power supply, an air switch, a communication module, and several analog input modules. The device is connected to the vibration analog quantity through multiple analog input modules; the analog data is converted into digital quantity and transmitted to the Internet of Things platform deployed in the deep learning server in the computer room through the Ethernet link of the baggage system; When collecting several groups of operating data, the collection interval is set to 15s. Under the condition of a speed of 1430rpm, the vibration data of four types of reduction motors, including the motor body and shaft, bearing, and gearbox, are collected; 4096 groups of data are collected for each of the four types of vibration data, for a total of 16384 groups of data; After data collection is completed, the corresponding fault labels are matched according to the data, that is, the label of the non-damaged data is named "0", the label of the motor body damaged data is named '1', the label of the shaft damaged data is named '2', the label of the bearing damaged motor is named '3', and the label of the gearbox damaged motor is named '4'. The vibration combined label data are combined in sequence to form the original data set.
[0033] The present invention also provides a computer-readable storage medium for storing computer instructions. When the computer instructions are executed by a processor, the steps of the above-disclosed method for diagnosing faults of a reduction motor of an airport baggage system are completed.
[0034] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the purpose and scope of the technical solution of the present invention, which should be included in the scope of the claims of the present invention.
Claims
1. A method for diagnosing faults of a reduction motor in an airport baggage system, characterized in that: The following steps are involved: Acquire operation data of several groups of reduction motors; wherein the operation data includes: motor body data, shaft data, bearing data and gearbox data; Preprocess the running data and construct the original data set according to the rules; Construct a WDCNN model, and train the WDCNN model until convergence using the original data set to obtain a multi-dimensional fault diagnosis algorithm model; The operating data of the gear motor to be diagnosed is input into the multi-dimensional fault diagnosis algorithm model to perform fault diagnosis and obtain the fault diagnosis result.
2. The method for diagnosing a fault of a reduction motor of an airport baggage system according to claim 1, characterized in that: Preprocess the running data and construct the original data set according to the rules, including: Convert the running data into a decimal vector group according to a preset protocol; Match the vector groups with the fault labels and construct the original dataset according to the rules; Among them, the fault labels include: no damage data label, motor body damage data label, shaft damage data label, bearing damage motor label and gearbox damage motor label.
3. The method for diagnosing a fault of a reduction motor of an airport baggage system according to claim 1, characterized in that: The WDCNN model includes: Input layer, used to input operation data; The first convolutional layer, connected to the input layer, is used to capture the low-frequency features of the running data; The second convolution layer, the third convolution layer, and the fourth convolution layer are arranged side by side, and are all used to realize high-frequency feature extraction of running data; The input ends of the first pooling layer, the second pooling layer, the third pooling layer, and the fourth pooling layer are connected to the output ends of the first convolutional layer, the second convolutional layer, the third convolutional layer, and the fourth convolutional layer, respectively, and the output ends of the first pooling layer, the second pooling layer, and the third pooling layer are connected to the input ends of the second convolutional layer, the third convolutional layer, and the fourth convolutional layer, respectively, all of which are used to significantly reduce the magnitude of network parameters; A Relu activation layer is respectively arranged between the output end of the first pooling layer and the input end of the second convolutional layer, between the output end of the second pooling layer and the input end of the third convolutional layer, and between the output end of the third pooling layer and the input end of the fourth convolutional layer; BN normalization layers are respectively arranged between the output end of the first convolutional layer and the input end of the second pooling layer, between the output end of the second convolutional layer and the input end of the third pooling layer, and between the output end of the third convolutional layer and the input end of the fourth pooling layer; The fully connected layer is connected to the fourth pooling layer to integrate and classify the features extracted from each layer to obtain the extraction result; The Softmax activation layer is connected to the fully connected layer and is used to generate fault diagnosis results based on the extraction results; Dropout layer, located between the fourth pooling layer and the fully connected layer; The output layer is used to output fault diagnosis results.
4. A method for diagnosing a fault of a reduction motor of an airport baggage system as claimed in claim 3, characterized in that: The first convolutional layer is a wide convolutional layer; the convolution kernel size of the first convolutional layer is 16*1 and the step size is 8*1.
5. The method for diagnosing a fault of a reduction motor of an airport baggage system according to claim 1, characterized in that: The loss function used in training the WDCNN model as follows: in, is the one-hot encoding of the sample target value, For the observed sample Belongs to category The predicted probability of is the sample size.
6. An airport baggage system reduction motor fault diagnosis system, characterized in that: include: Agile data acquisition module, used to obtain the operating data of several groups of reduction motors; The feature vector reorganization module is used to pre-process the operation data and construct the original data set according to the rules; The convolution model training module is used to build the WDCNN model and train the WDCNN model until convergence using the original data set to obtain a multi-dimensional fault diagnosis algorithm model; The multidimensional fault diagnosis module is used to input the operating data of the reduction motor to be diagnosed into the multidimensional fault diagnosis algorithm model to perform fault diagnosis and obtain the fault diagnosis result.
7. The airport baggage system reduction motor fault diagnosis system according to claim 6, characterized in that: Agile data collection module, including: The vibration sensor is used to collect vibration data of the reduction motor, shaft, bearing and reduction box respectively, and obtain motor body data, shaft data, bearing data and gear box data; The motor monitoring DCU box is electrically connected to the vibration sensor and is used to record and manage motor body data, shaft data, bearing data and gearbox data.
8. A computer-readable storage medium for storing computer instructions, wherein when the computer instructions are executed by a processor, the steps of the method for diagnosing a fault of a reduction motor of an airport baggage system disclosed in any one of claims 1 to 5 are completed.
Citation Information
Patent Citations
Fault diagnosis method and device for hydraulic variable pitch system
CN113027696A