Carrier roller fault detection system and method based on multi-source information

Abstract

The invention relates to a multi-source information-based carrier roller fault detection system and method, which can quickly and accurately diagnose the fault type and position of a carrier roller by using a plurality of sensors to collect data, training a data fusion model and a carrier roller fault diagnosis model through a deep neural network and combining a temperature detection device. The invention solves the problems that the current fault detection of the port belt conveyor carrier roller mainly depends on manual regular inspection, the workload is large, the efficiency is low, the working environment is dangerous, and the dust on the inspection site seriously harms the body health of people. Carrier roller faults can be found in time, equipment shutdown time can be shortened, and loading and unloading efficiency and production capacity are improved. And the risk of accidents in equipment operation is reduced, and the safety of a working place is improved.

Description

Technical Field

[0001] The invention belongs to the technical field of port equipment fault detection, and in particular to a roller fault detection system and method based on multi-source information. Background Art

[0002] Port belt conveyors can load and unload coal quickly and efficiently, increasing the throughput of ports. To adapt to the port environment, port belt conveyors are usually designed with a large load-bearing capacity. Port belt conveyors can usually achieve continuous operation, which improves the efficiency of loading and unloading coal. Rollers are rolling support components in the process of coal loading, unloading and transportation. However, since ports usually face marine climates, rollers may be exposed to humid environments, and the loading and unloading of coal will also introduce a lot of dust. Port belt conveyors usually need to withstand heavy loads and frequent use of coal, which places high demands on rollers. Port belt conveyors working outdoors may experience large temperature differences between day and night, which will also affect the working environment of rollers. Therefore, the stability of rollers is particularly important. The following are common reasons for the failure of port belt conveyor rollers: long-term use or heavy loads may cause wear on the roller surface, leading to failure; if the roller lacks proper lubrication, friction will increase, which may cause failure; if the quality of the roller itself is unqualified, it may also cause failure; improper installation or loose fixings may cause roller failure. If the roller failure is not discovered in time, the roller failure will cause the conveying system to stop working, resulting in interruption of loading and unloading production, affecting the coal throughput efficiency of the port; it may cause coal to spill or other parts of the conveying system to be damaged, posing a safety hazard; it may increase maintenance costs and affect the normal operation of the port belt conveyor.

[0003] At present, the fault detection of port belt conveyor rollers mainly relies on regular manual inspections, which has a large workload, low efficiency and a dangerous working environment. At the same time, the dust at the inspection site is highly dusty and seriously endangers human health. Therefore, a roller fault detection system is needed that can collect information through sensors and analyze and diagnose faults through computers. Summary of the invention

[0004] The purpose of the present invention is to overcome the shortcomings of the prior art and propose a roller fault detection system and method based on multi-source information, which can detect roller faults in time, reduce equipment downtime, improve loading and unloading efficiency and production capacity, reduce the risk of accidents during equipment operation, and improve workplace safety.

[0005] The present invention solves the technical problem by adopting the following technical solutions:

[0006] A roller fault detection system based on multi-source information includes a data acquisition unit, a data processing unit, a data fusion unit, and a fault diagnosis unit, wherein the data acquisition unit, the data processing unit, the data fusion unit, and the fault diagnosis unit are connected in sequence, the data acquisition unit is used to collect data, the data processing unit is used to process the data collected by the data acquisition unit, the data fusion unit is used to fuse the data processed by the data processing unit, and the fault diagnosis unit is used to perform fault diagnosis based on the data fused by the data fusion unit.

[0007] Moreover, the data acquisition unit includes multiple vibration sensors, sound sensors and temperature sensors. The sensors are arranged on the side of the port belt conveyor and are distributed at intervals along the transportation direction. The multiple sensors send the collected data to the terminal for use by the data processing unit.

[0008] Moreover, the data processing unit includes a data storage device and a terminal, wherein the data storage device is divided into a corresponding number of storage spaces according to the number of sensors, and the data processing unit first stores the data collected by the data acquisition unit into the corresponding space, and then processes the noise data, vibration data and temperature data respectively.

[0009] Moreover, the data fusion unit fuses the vibration signal and the sound signal processed by the data processing unit through a multi-attention network.

[0010] A detection method of a roller fault detection system based on multi-source information comprises the following steps:

[0011] Step 1, the data acquisition unit collects data;

[0012] Step 2: The data processing unit processes the data collected by the data collection unit;

[0013] Step 3: The data fusion unit fuses the data processed by the data processing unit;

[0014] Step 4: The fault diagnosis unit performs fault diagnosis based on the data fused by the data fusion unit.

[0015] Moreover, the specific implementation method of step 1 is that the vibration sensor, sound sensor and temperature sensor in the data acquisition unit respectively collect noise data, vibration data and temperature data.

[0016] Moreover, the specific implementation method of step 2 is as follows: the data processing unit processes the sound signal by performing noise reduction processing on the sound signal through an extended Kalman filter, preliminarily filters out the environmental noise, and then performs empirical mode decomposition, fuses the decomposed data through dimensionality reduction technology, normalizes the fused data, and represents the data as a wavelet time-frequency image through continuous wavelet transform, and the wavelet time-frequency image maps the signal's dependence on time into the image feature space, so that the original feature information is retained in the time-frequency diagram; the data processing unit processes the vibration signal by selecting a vibration acceleration signal, denoising it through wavelet transform, and obtaining a power spectrum image using the fast Fourier transform method.

[0017] Moreover, the specific implementation method of step 3 is: the processed wavelet time-frequency image and power spectrum image are fused through a multi-attention network, which consists of two attention models, a position attention model and a channel attention model. First, the local features of the input image generated by the residual network are used to draw the global content, and then the features of the extended residual network are input into two parallel attention modules to generate a fused image.

[0018] Moreover, the specific implementation method of step 4 is: perform a preliminary diagnosis through the temperature curve of the roller, then input the generated fusion image into the transformer network, and diagnose the roller fault in combination with the above results. When only one of the two indicators is diagnosed as a roller fault, an alarm will be issued. At the same time, a shutdown temperature threshold is set. When the threshold is exceeded, the system will be forced to shut down to ensure the safety of equipment and personnel.

[0019] The advantages and positive effects of the present invention are:

[0020] The present invention collects data by applying a variety of sensors (such as temperature sensors, vibration sensors and sound sensors), and trains data fusion models and roller fault diagnosis models through deep neural networks. In combination with a temperature detection device, it can quickly and accurately diagnose the fault type and location of the roller. The present invention solves the problem that the current fault detection of port belt conveyor rollers mainly relies on regular manual inspections, which has a large workload, low efficiency and a dangerous working environment. At the same time, the dust at the inspection site is large and seriously endangers human health. Timely detection of roller faults can reduce equipment downtime, improve loading and unloading efficiency and production capacity. Reduce the risk of accidents during equipment operation and improve workplace safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 is a system structure diagram of the present invention;

[0022] Figure 2 This is a diagram of the multi-attention network structure of the present invention;

[0023] Figure 3This is a structural diagram of the position attention model of the present invention;

[0024] Figure 4 This is a structural diagram of the channel attention model of the present invention. DETAILED DESCRIPTION

[0025] The present invention is further described in detail below with reference to the accompanying drawings.

[0026] A roller fault detection system based on multi-source information, such as Figure 1 As shown, it includes a data acquisition unit, a data processing unit, a data fusion unit, and a fault diagnosis unit, wherein the data acquisition unit, the data processing unit, the data fusion unit, and the fault diagnosis unit are connected in sequence, the data acquisition unit is used to collect data, the data processing unit is used to process the data collected by the data acquisition unit, the data fusion unit is used to fuse the data processed by the data processing unit, and the fault diagnosis unit is used to perform fault diagnosis according to the data fused by the data fusion unit.

[0027] The data acquisition unit includes multiple vibration sensors, sound sensors and temperature sensors. The sensors are arranged on the side of the port belt conveyor and are distributed at intervals along the transportation direction. The multiple sensors send the collected data to the terminal for use by the data processing unit.

[0028] The data processing unit includes a data storage device and a terminal, wherein the data storage device is divided into a corresponding number of storage spaces according to the number of sensors. The data processing unit first stores the data collected by the data acquisition unit into the corresponding space, and then processes the noise data, vibration data and temperature data respectively.

[0029] The data fusion unit fuses the vibration signal and sound signal processed by the data processing unit through a multi-attention network.

[0030] A detection method of a roller fault detection system based on multi-source information comprises the following steps:

[0031] Step 1: The data acquisition unit acquires data. The vibration sensor, sound sensor and temperature sensor in the data acquisition unit acquire noise data, vibration data and temperature data respectively.

[0032] Step 2: The data processing unit processes the data collected by the data collection unit.

[0033] The data processing unit processes the sound signal by performing noise reduction processing on the sound signal through the extended Kalman filter, preliminarily filtering out the environmental noise, and then performing empirical mode decomposition. The decomposed data are fused through dimensionality reduction technology, the fused data are normalized and represented as wavelet time-frequency images through continuous wavelet transform. The wavelet time-frequency image maps the signal's dependence on time into the image feature space, so that the original feature information is retained in the time-frequency diagram.

[0034] Empirical mode decomposition (EMD) is essentially a means of stabilizing non-stationary signals. The result is to decompose the fluctuations and trends of different scales in the signal step by step, generating a series of data sequences with different characteristic scales, each of which is called an intrinsic mode function IMF. Empirical mode decomposition has 6 steps:

[0035] Step 2.1: Find all extreme points of the signal, connect the local maximum points into an upper envelope through a cubic spline curve, and connect the local minimum points into a lower envelope. The upper and lower envelopes contain all data points.

[0036] Step 2.2: Take the average value of the upper and lower envelopes

[0037] Step 2.3: Subtract the mean envelope from the original signal to obtain the intermediate signal.

[0038] x(t)-m1(t)=C 1,1 (t)

[0039] Step 2.4: Determine the intermediate signal C 1,1 (t) Whether it meets the two conditions of IMF. If it does, the signal is an IMF component. If not, based on the signal, repeat the analysis of 1 to 4. That is, continue to "screen" until the signal obtained after k decompositions meets the IMF conditions and obtains the first IMF component I1(t) of the original signal.

[0040] r1(t)=x(t)-I1(t)

[0041] Step 2.5: Perform the above “screening” process on r1(t).

[0042] r1(t)-I2(t)=r2(t)

[0043] …

[0044] r n-1 (t)-I n (t) = r n (t)

[0045] Step 2.6: When r nAfter (t) becomes a monotonic function, the sum of all IMF components and residual components is the original signal

[0046]

[0047] Continuous wavelet transform is to use a series of wavelets obtained by shifting and scaling the mother wavelet to represent a signal. A series of wavelets can be used as basis functions to represent some functions.

[0048]

[0049] Wavelet transform is the sum of the product of the signal f(t) and the scaled and translated wavelet function ψ over the entire period of the signal's existence. The result of continuous wavelet transform is a number of wavelet coefficients C, which are functions of the scaling factor (scale) and position (position).

[0050] The data processing unit processes the vibration signal to select the vibration acceleration signal, denoise it through wavelet transform and use the fast Fourier transform method to obtain the power spectrum image. This image is the characteristic of the vibration signal. Suppose the vibration acceleration signal is x i , N is the sampling length, then the power spectrum of the vibration acceleration signal is:

[0051]

[0052] The fast Fourier transform, its mathematical expression is:

[0053]

[0054] Step 3: The data fusion unit fuses the data processed by the data processing unit. The data fusion unit fuses the processed vibration signal and sound signal through Figure 2 The DualAttentionNetwork shown in Figure 1 is fused and consists of two attention models, such as Figure 3 The position attention model shown in Figure 4 The channel attention model shown in Figure 1 first draws the global content by passing the local features generated by the residual network through the input image, and then inputs the features of the extended residual network into two parallel attention modules to generate a fused image.

[0055] Position Attention Model We first use a convolutional layer to obtain dimension-reduced features, and then input the features into the position attention module to generate new features of spatial long-range context information through the following three steps. The first step is to generate a spatial attention matrix, which models the spatial relationship between any two pixels of the feature. Next, we perform matrix multiplication between the attention matrix and the original feature. Its expression is:

[0056]

[0057] Third, we perform an element-wise sum operation on the result matrix of the above multiplication and the original features to obtain the final representation reflecting the above, which is expressed as (α is taken as 0.5):

[0058]

[0059] The channel attention model captures long-range feature information on the channel dimension through the channel attention module. We directly calculate the channel attention map X from the original feature A. Specifically, we reshape A and then perform matrix multiplication of A and the transpose of A. Finally, we apply a softmax layer to get the channel attention map X, which is expressed as:

[0060]

[0061] In addition, we perform matrix multiplication between the transpose of X and A and reshape the result. Then we multiply the result by a scale parameter β (taken as 0.5) and perform an element-wise sum operation with A to obtain the final output E, which is expressed as:

[0062]

[0063] Before fault diagnosis, the transformer network needs to be trained, with an initial learning rate of 0.01, a momentum of 0.9, a weight decay of 0.0001, a batch size of 8, a poly learning rate strategy, and an epoch of 1000. The data of different parts are divided into different training sets and test sets in a ratio of 7:3. The training set is used to train the network model, and the test set is used to evaluate the accuracy of the diagnosis model.

[0064] Step 4: The fault diagnosis unit performs fault diagnosis based on the data fused by the data fusion unit. The generated fusion image is input into the transformer network, and the roller fault is diagnosed based on the above results. When one of the two indicators is diagnosed as a roller fault, an alarm will be issued. At the same time, a shutdown temperature threshold is set. When the threshold is exceeded, the system will be forced to shut down to ensure the safety of equipment and personnel.

[0065] It should be emphasized that the embodiments described in the present invention are illustrative rather than restrictive. Therefore, the present invention includes but is not limited to the embodiments described in the specific implementation manner. Any other implementation manners derived by those skilled in the art based on the technical solution of the present invention also fall within the scope of protection of the present invention.

Claims

1. A roller fault detection system based on multi-source information, characterized in that: It includes a data acquisition unit, a data processing unit, a data fusion unit, and a fault diagnosis unit, wherein the data acquisition unit, the data processing unit, the data fusion unit, and the fault diagnosis unit are connected in sequence, the data acquisition unit is used to collect data, the data processing unit is used to process the data collected by the data acquisition unit, the data fusion unit is used to fuse the data processed by the data processing unit, and the fault diagnosis unit is used to perform fault diagnosis based on the data fused by the data fusion unit.

2. The roller fault detection system based on multi-source information according to claim 1, characterized in that: The data acquisition unit includes multiple vibration sensors, sound sensors and temperature sensors. The sensors are arranged on the side of the port belt conveyor and are distributed at intervals along the transportation direction. The multiple sensors send the collected data to the terminal for use by the data processing unit.

3. The roller fault detection system based on multi-source information according to claim 1 is characterized in that: The data processing unit includes a data storage device and a terminal, wherein the data storage device is divided into a corresponding number of storage spaces according to the number of sensors. The data processing unit first stores the data collected by the data acquisition unit into the corresponding space, and then processes the noise data, vibration data and temperature data respectively.

4. The roller fault detection system based on multi-source information according to claim 1 is characterized in that: The data fusion unit fuses the vibration signal and the sound signal processed by the data processing unit through a multi-attention network.

5. A detection method of a roller fault detection system based on multi-source information according to any one of claims 1 to 4, characterized in that: The following steps are involved: Step 1, the data acquisition unit collects data; Step 2: The data processing unit processes the data collected by the data collection unit; Step 3: The data fusion unit fuses the data processed by the data processing unit; Step 4: The fault diagnosis unit performs fault diagnosis based on the data fused by the data fusion unit.

6. A roller fault detection system and method based on multi-source information according to claim 5, characterized in that: The specific implementation method of step 1 is that the vibration sensor, sound sensor and temperature sensor in the data acquisition unit respectively collect noise data, vibration data and temperature data.

7. The roller fault detection system and method based on multi-source information according to claim 5, characterized in that: The specific implementation method of step 2 is as follows: the data processing unit processes the sound signal by performing noise reduction processing on the sound signal through an extended Kalman filter, preliminarily filters out the environmental noise, and then performs empirical mode decomposition, fuses the decomposed data through dimensionality reduction technology, normalizes the fused data, and represents the data as a wavelet time-frequency image through continuous wavelet transform, and the wavelet time-frequency image maps the signal's dependence on time into the image feature space, so that the original feature information is retained in the time-frequency diagram; the data processing unit processes the vibration signal by selecting a vibration acceleration signal, denoising it through wavelet transform, and obtaining a power spectrum image using the fast Fourier transform method.

8. The roller fault detection system and method based on multi-source information according to claim 5, characterized in that: The specific implementation method of step 3 is: the processed wavelet time-frequency image and power spectrum image are fused through a multi-attention network, which consists of two attention models, a position attention model and a channel attention model. First, the local features of the input image generated by the residual network are used to draw the global content, and then the features of the extended residual network are input into two parallel attention modules to generate a fused image.

9. The roller fault detection system and method based on multi-source information according to claim 5, characterized in that: The specific implementation method of step 4 is: perform a preliminary diagnosis through the temperature curve of the roller, then input the generated fusion image into the transformer network, and diagnose the roller fault in combination with the above results. When only one of the two indicators is diagnosed as a roller fault, an alarm will be issued. At the same time, a shutdown temperature threshold is set. When the threshold is exceeded, the system will be forced to shut down to ensure the safety of equipment and personnel.

Images