Coal mill powder accumulation fault identification method, device, electronic equipment and system
By using the CNN network model to extract and classify the running images of coal mills, the problem of automatic prediction of powder accumulation faults of coal mills is solved, and efficient fault identification and combustion optimization are achieved.
Patent Information
- Application Number
- CN202411673718.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-21
- Publication Date
- 2025-03-18
AI Technical Summary
The prior art cannot realize automated prediction of powder accumulation failure of coal mill, resulting in low operating efficiency of powder making system and difficulty in identifying faults.
The CNN network model is used to combine the preprocessing technology of real-time running images to extract the local features of the coal mill running images, and the powder accumulation fault classification and prediction are used to generate powder accumulation fault prediction results.
It realizes rapid automated prediction of powder accumulation faults of coal mills, improves the accuracy and reliability of fault identification, optimizes the combustion process, improves the utilization efficiency of coal powder, and reduces environmental pollution.
Smart Images
Figure CN119649097A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of fault diagnosis of pulverizing systems, and in particular to the technical field of coal mill fault identification, and specifically to a method, device, electronic equipment and system for identifying coal mill dust accumulation fault. Background Art
[0002] As an important part of the entire thermal power generation system, the pulverizing system is responsible for providing fuel for the boiler. The pulverizing system refers to a system that grinds the raw coal into powder through the pulverizer and then directly or indirectly feeds it into the furnace for combustion. Its principle is: the raw coal enters the raw coal hopper from the coal conveying system, then falls into the coal feeder, and enters the pulverizer after the coal feeder adjusts the amount of coal given. The pulverizer grinds the raw coal into coal powder with preliminary qualified fineness. This part of the coal powder then enters the coal powder separator, which separates the coal powder into coarse and fine parts. The coarse powder returns to the pulverizer for re-grinding, and the fine powder is carried to the burner through the primary air in the pulverized coal pipe. The pulverizing system of a thermal power plant is prone to various faults due to its harsh operating environment, complex system structure, and long-term continuous high load of the coal feeder and pulverizer, which seriously affects its use effect and cannot meet the current actual development needs. Therefore, it is necessary to carry out early fault diagnosis and scientific prediction of the pulverizing system of a thermal power plant.
[0003] Because the related machinery in the pulverizing system has hierarchical characteristics, the faults in the pulverizing system also have hierarchical characteristics. When the coal mill, as the core equipment in the pulverizing system, fails, the cause of the failure may come from a low-level failure, and it may also cause a higher-level failure. Therefore, the fault diagnosis of the coal mill is also the key content of the fault diagnosis of the pulverizing system.
[0004] There are many types of coal mill failures, such as coal pile failure, coal powder accumulation failure, insufficient coal powder crushing, high coal powder temperature, etc. Among them, high coal powder temperature can easily cause coal powder spontaneous combustion in coal mills. At present, through preliminary research, it is found that neural network models driven by big data are increasingly used in early fault diagnosis of pulverizing systems. However, no relevant solutions that can realize automatic prediction of coal mill powder accumulation failures have been proposed. Summary of the invention
[0005] In view of this, the purpose of the embodiments of the present invention is to provide a method, device, electronic equipment and system for identifying coal mill dust accumulation faults, so as to overcome the technical problem that the existing solutions cannot realize automatic prediction of coal mill dust accumulation faults.
[0006] In order to achieve the above-mentioned object, a first aspect of an embodiment of the present invention provides a method for identifying a coal mill dust accumulation fault, the method comprising: acquiring a real-time operation image of the coal mill;
[0007] Inputting the preprocessed operation image into a trained prediction model to generate a powder accumulation fault prediction result, wherein the powder accumulation fault prediction result is a normal operation or a specific powder accumulation fault type;
[0008] The prediction model is used to extract local features of the running image, and classify the powder accumulation fault according to the local features, and output the powder accumulation fault prediction result after classification, and the local features characterize the powder accumulation state.
[0009] Optionally, the training samples constructed during the prediction model training include coal mill operation images under different lighting conditions.
[0010] Optionally, the local characteristics include shape characteristics, texture characteristics, particle size distribution characteristics, coal quality characteristics, moisture characteristics, density characteristics, adhesion characteristics and coal powder composition characteristics.
[0011] Optionally, the powder accumulation failure types include accumulation blockage, insufficient crushing, dust backflow, crushing machinery failure, improper operating parameters, excessive humidity and increased adhesion.
[0012] Optionally, the prediction model is a CNN network model, which includes a first convolutional layer, a second convolutional layer, a third convolutional layer, a fourth convolutional layer, a fifth convolutional layer, an activation layer, a pooling layer, a fully connected layer and an output layer connected in sequence.
[0013] Optionally, the preprocessing includes segmenting the powder accumulation area from the running image based on a threshold segmentation algorithm, normalization processing and image enhancement.
[0014] Optionally, the running image is a visible light and infrared fusion image, and the method further includes:
[0015] identifying the pulverized coal temperature from the visible light and infrared fusion image;
[0016] The coal powder temperature is used to determine the coal mill dust accumulation fault level.
[0017] Optionally, determining the dust accumulation fault level of the coal mill by using the coal powder temperature includes:
[0018] If it is determined according to the prediction result of the dust accumulation fault that the coal mill has a dust accumulation fault, and the coal powder temperature exceeds the first preset value, it is determined that the coal mill has a first level of dust accumulation fault;
[0019] If it is determined according to the prediction result of the dust accumulation fault that the coal mill has a dust accumulation fault, and the coal powder temperature does not exceed the first preset value, it is determined that the coal mill has a second level of dust accumulation fault;
[0020] Among them, the severity of the first level of powder accumulation failure is higher than the second level of powder accumulation failure.
[0021] A second aspect of an embodiment of the present invention provides a coal mill dust accumulation fault identification device, the device comprising:
[0022] A real-time operation image acquisition module is used to obtain the real-time operation image of the coal mill;
[0023] A dust accumulation fault prediction module is used to input the preprocessed operation image into a trained prediction model to generate a dust accumulation fault prediction result, wherein the dust accumulation fault prediction result is normal operation or a specific dust accumulation fault type;
[0024] The prediction model is used to extract local features of the running image, and classify the powder accumulation fault according to the local features, and output the powder accumulation fault prediction result after classification, and the local features characterize the powder accumulation state.
[0025] A third aspect of an embodiment of the present invention provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, a method for identifying coal mill dust accumulation fault according to the first aspect of an embodiment of the present invention is implemented.
[0026] A fourth aspect of an embodiment of the present invention provides a storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, a method for identifying a coal mill dust accumulation fault according to the first aspect of an embodiment of the present invention is implemented.
[0027] The beneficial effects brought by the above technical solution mainly include:
[0028] 1) End-to-end dust accumulation fault identification is achieved through prediction models such as CNN network models. Dust accumulation fault classification is performed directly on the acquired coal mill operation images without manual design and feature extraction, which simplifies the identification process and realizes rapid and automatic prediction of dust accumulation faults, which is conducive to identifying coal mill problems and optimizing the combustion process, so as to improve the utilization efficiency of coal powder and reduce environmental pollution;
[0029] 2) By collecting coal mill operation images under different lighting conditions to construct training samples for the prediction model, combined with post-training parameter adjustment, the generalization ability of the prediction model is improved and can adapt to different lighting environments, thereby improving the accuracy and reliability of dust accumulation fault identification.
[0030] Other features and advantages of the embodiments of the present invention will be described in detail in the subsequent detailed description. BRIEF DESCRIPTION OF THE DRAWINGS
[0031] The accompanying drawings are used to provide a further understanding of the embodiments of the present invention and constitute a part of the specification. Together with the following specific implementations, they are used to explain the embodiments of the present invention, but do not constitute a limitation on the embodiments of the present invention. In the accompanying drawings:
[0032] Figure 1 A schematic flow chart of a method for identifying a coal mill dust accumulation fault implemented in an embodiment of the present invention;
[0033] Figure 2 It is a network structure diagram of the CNN network model;
[0034] Figure 3 A block diagram of a coal mill dust accumulation fault identification device implemented in an embodiment of the present invention;
[0035] Figure 4 The present invention is a block diagram of a coal mill dust accumulation fault identification system implemented in an embodiment of the present invention. DETAILED DESCRIPTION
[0036] The specific implementation of the embodiment of the present invention is described in detail below in conjunction with the accompanying drawings. It should be understood that the specific implementation described here is only used to illustrate and explain the embodiment of the present invention, and is not used to limit the embodiment of the present invention.
[0037] Method Embodiment
[0038] See also Figure 1 to Figure 2 The present invention provides a method for identifying coal mill dust accumulation fault, which specifically includes the following implementation steps:
[0039] S100. Acquire a real-time operating image of the coal mill. For example, the real-time operating image may be a real-time image collected by an imaging device that monitors the operating status of the coal powder discharge port of the coal mill, or real-time frame data collected by a video device.
[0040] S200. Input the preprocessed real-time operation image into the trained prediction model to generate a dust accumulation fault prediction result, which is a normal operation or a specific type of dust accumulation fault. It should be understood that if the dust accumulation fault prediction result is a normal operation, it indicates that the current dust accumulation state of the coal mill is normal and no dust accumulation fault occurs. The prediction model is used to extract the local features of the real-time operation image input to the model, and classify the dust accumulation fault according to the local features, and output the dust accumulation fault prediction result after classification. Local features refer to features that can characterize the dust accumulation state.
[0041] Among them, the preprocessing can adopt the preprocessing process in the common embodiment, such as denoising, image cropping to adapt to the input image size of the prediction model, etc. The trained prediction model means: after selecting the basic network framework of the prediction model and setting the initial network parameters, the prediction model is trained using training samples, and the model parameters are gradually adjusted during training through the back propagation algorithm, so that the model prediction results are close to the true label, thereby determining the final network parameters, and using the determined network parameters to adjust the prediction model to obtain the prediction model after parameter optimization, and the prediction model after parameter optimization is tested by the constructed test samples, and the evaluation indicators may include recognition accuracy, recall rate, etc. When the evaluation indicators are not up to standard, the parameters are adjusted again, and the final prediction model is determined after repeated multiple times. The prediction model at this time meets the evaluation indicator requirements, and the prediction model at this time can be called a trained prediction model. Preferably, when performing dust accumulation fault classification analysis based on local features, a k-means clustering algorithm or the like can be used. When training the prediction model, the dust accumulation fault classification loss function based on k-means clustering can be expressed by minimizing the sum of squared distances between each sample and the center of its cluster, that is, the intra-cluster sum of squares.
[0042] It can be known that different powder accumulation states can correspond to different types of powder accumulation faults. Therefore, different types of powder accumulation faults can be identified by combining one or more local features. For example, by extracting the particle size distribution characteristics of the real-time operation image, if the particle size distribution characteristics indicate that the coal powder particles are large, it reflects that the coal mill cannot completely crush the coal blocks. At this time, the powder accumulation fault type corresponding to the powder accumulation state can be judged as insufficient crushing.
[0043] It should be understood that the dust accumulation fault type can be combined with the common dust accumulation faults of the coal mill to define the specific dust accumulation fault identification requirements, and then define the local features that need to be extracted.
[0044] For example, in one embodiment, dust accumulation fault types include accumulation blockage, insufficient pulverization, dust backflow, pulverization machinery failure, improper operating parameters, excessive humidity, and increased adhesion.
[0045] Accumulation blockage refers to the accumulation of coal powder inside the coal mill, which blocks the normal discharge of coal powder. Through the analysis of this type of fault, it can be seen that accumulation blockage is mainly caused by improper internal design of the coal mill, incorrect operating parameters during coal grinding, or excessive humidity of coal powder. Insufficient crushing refers to the coal mill not completely crushing the coal block, resulting in some coal powder being unable to be discharged through the crushing system. Through the analysis of this type of fault, it can be seen that insufficient crushing is mainly caused by wear or excessive load of the coal mill. Dust reflux refers to the reflux of coal powder in the dust system, returning to the coal mill to form accumulated powder. Through the analysis of this type of fault, it can be seen that insufficient crushing is mainly caused by poor exhaust of the dust system or pipe blockage. Crushing machinery failure refers to the damage or failure of mechanical parts inside the coal mill, resulting in the inability to properly crush and discharge coal powder. Improper operating parameters refer to the unreasonable setting of the operating parameters of the coal mill, such as too fast supply speed, too slow cutting speed, or insufficient air volume, which leads to coal powder accumulation, and thus the required coal powder discharge state cannot be achieved. Too high humidity of coal powder will increase the adhesion of coal powder, which will easily lead to coal powder accumulation, making it impossible to achieve the desired coal powder discharge state. The increased adhesion of coal powder refers to the increased mutual adhesion between coal powders, which can easily lead to the failure of coal powder to be discharged smoothly. Through the analysis of this type of fault, it can be seen that the increased adhesion of coal powder is mainly caused by changes in coal quality or additives.
[0046] It should be understood that low-level local feature extraction can be achieved by constructing different convolution filters, or even high-level local feature extraction containing more abstract and semantic information.
[0047] For example, in one embodiment, in order to achieve the prediction of the above-mentioned seven types of dust accumulation faults, including accumulation blockage, insufficient crushing, dust reflux, crushing machine failure, improper operating parameters, excessive humidity and enhanced adhesion, the extracted local features include shape features, texture features, particle size distribution features, coal quality features, humidity features, density features, adhesion features and coal powder composition features. Among them, shape features and texture features are low-level features, and particle size distribution features, coal quality features, humidity features, density features, adhesion features and coal powder composition features are high-level features. By extracting the above eight local features, the feature expression ability of the prediction model is enriched, thereby improving the accuracy of the prediction model in predicting dust accumulation faults.
[0048] Particle size distribution refers to the size distribution of coal powder particles in the accumulated powder, that is, the percentage of particles of different particle sizes in the total amount of coal powder particles. The particle shape characterization parameters of coal powder usually include roundness and the aspect ratio of the circumscribed rectangle of the particle. Roundness indicates the similarity between the particle shape and a circle. The size of the roundness is related to the projected area and circumference of the particle. The larger the roundness, the more similar the particle shape is to a circle. The size of the roundness can usually be calculated by the following formula: In the formula, R represents the roundness, S represents the particle projection area, and P represents the particle perimeter. The aspect ratio of the particle circumscribed rectangle is usually calculated using the following formula: Where M represents the aspect ratio of the circumscribed rectangle of the particle, L represents the length of the circumscribed rectangle of the particle, and W represents the width of the circumscribed rectangle of the particle. Particle size distribution can usually be divided into interval distribution and cumulative distribution. Interval distribution represents the percentage of particles within a certain particle size range in the total amount of particles, and cumulative distribution represents the percentage of particles smaller than or larger than a certain particle size in the total amount. It can be seen that the particle size distribution characteristics can reflect the proportion of coal powder of different particle sizes in the accumulated powder, and whether there is too fine or too coarse. In order to extract the particle size distribution characteristics, analysis and evaluation can be carried out through sampling, laboratory testing and online monitoring, such as particle size analysis, so as to determine the correspondence between the coal mill operation image and the particle size distribution characteristics. For example, a convolution filter and a convolution subnetwork are constructed according to the different image grayscale characteristics corresponding to different particle size distributions, and then training and parameter adjustment are performed to determine the parameters of the final convolution filter and the convolution subnetwork, so that the feature map extracted by the convolution subnetwork can characterize the particle size distribution characteristics of the accumulated powder in the coal mill operation image.
[0049] Coal quality includes ash, volatile matter, fixed carbon, etc. in the accumulated coal powder. Different coal quality characteristics can represent different combustion performances of the accumulated coal powder. The differences caused by coal quality differences cause differences in the structure and plastic deformation of the residual carbon formed, which affects the number and size of the generated coal powder. Therefore, based on the following understanding: the more vitrinite and exinite content in the coal, the less coarse-grained fly ash is formed in the coal ash, and the smaller the average volume particle size of the coal ash; the inertinite group is not easy to break and is easy to form coarse-grained coal ash; the higher the average reflectivity of the vitrinite group and the coal with large coal rank differences will lead to a decrease in crushing reactivity, so that the particles experience weaker plastic deformation and are easy to form coarser particles. What we can know is that the ash, volatile matter and fixed carbon of the coal powder in the accumulated coal powder can be further analyzed based on the particle size distribution in the coal powder, that is, the coal quality characteristics can be characterized by the particle size distribution characteristics.
[0050] The humidity of coal powder in the accumulated powder refers to the percentage of water contained in the coal powder. Different humidity characteristics can represent different coal powder flow properties and combustion performance. In order to extract humidity characteristics, analysis and evaluation can be carried out through sampling, laboratory testing and online monitoring to determine the correspondence between the coal mill operation image and humidity characteristics. For example, convolution filters and convolution sub-networks are constructed according to the grayscale characteristics of different images corresponding to different humidity levels, and then training and parameter adjustment are performed to determine the parameters of the final convolution filters and convolution sub-networks, so that the feature map extracted by the convolution sub-network can characterize the humidity characteristics of the accumulated powder in the coal mill operation image.
[0051] The density of coal powder in the accumulated powder refers to the mass per unit volume, and different density characteristics can represent different coal powder flow properties and combustion performance. In order to extract density features, analysis and evaluation can be carried out through sampling, laboratory testing and online monitoring to determine the correspondence between the coal mill operation image and the density features. For example, convolution filters and convolution subnetworks are constructed according to the grayscale, shape and texture characteristics of different images corresponding to different densities, and then training and parameter adjustment are performed to determine the parameters of the final convolution filters and convolution subnetworks, so that the feature map extracted by the convolution subnetwork can characterize the density characteristics of the accumulated powder in the coal mill operation image.
[0052] The adhesion of coal powder in the accumulated powder refers to the mutual adhesion between coal powders. The adhesion characteristics can reflect the adhesion level of coal powder in the accumulated powder. It is known that coal powder with high adhesion is more likely to form coal powder accumulation than coal powder with low adhesion, which affects the operation of the coal mill and the transportation of coal powder. In order to extract the adhesion characteristics, analysis and evaluation can be carried out through sampling, laboratory testing and online monitoring to determine the correspondence between the coal mill operation image and the adhesion characteristics. For example, convolution filters and convolution subnetworks are constructed according to the grayscale, shape and texture characteristics of different images corresponding to different adhesions, and then training and parameter adjustment are performed to determine the parameters of the final convolution filters and convolution subnetworks, so that the feature map extracted by the convolution subnetwork can characterize the adhesion characteristics of the accumulated powder in the coal mill operation image.
[0053] The coal powder composition in the accumulated powder refers to the content of various elements contained in the coal powder, such as the content of different elements such as Car, Har, Oar, Nar and Sar. It is known that different coal powder compositions can represent different coal powder qualities and combustion performances. In order to extract the characteristics of coal powder composition, analysis and evaluation can be carried out through sampling, laboratory testing and online monitoring, such as coal powder chemical composition analysis, so as to determine the correspondence between the coal mill operation image and the coal powder composition characteristics. For example, convolution filters and convolution sub-networks are constructed according to the grayscale characteristics of different images corresponding to different coal powder components, and then training and parameter adjustment are carried out to determine the parameters of the final convolution filters and convolution sub-networks, so that the feature map extracted by the convolution sub-network can characterize the coal powder composition characteristics of the accumulated powder in the coal mill operation image.
[0054] It should be understood that the present invention does not record the composition of the above-mentioned convolutional filters and the hierarchical structure of the convolutional sub-networks. However, for ordinary technicians in this field, based on the explanation of the extraction principles of the above-mentioned eight local features, they can know and implement the design of corresponding convolutional filters and convolutional sub-networks, that is, they can determine one or more networks for extracting the above-mentioned eight local features.
[0055] Exemplarily, in order to achieve the above-mentioned extraction of local features and classification of dust accumulation fault types, the prediction model can adopt a CNN network model. The CNN network model is constructed based on a convolutional neural network. For example, it may include a convolutional neural network, an activation layer, a pooling layer, a fully connected layer and an output layer connected in sequence. The convolutional neural network is used to extract features of the input coal mill operation image to obtain local features. The activation layer is used to increase the nonlinear expression ability of the model so that it can fit complex relationships. The pooling layer is used to downsample, thereby reducing the complexity of the model and preventing overfitting, such as using average pooling. The fully connected layer is used to determine whether there is a dust accumulation fault based on the value of the input feature vector, and the output layer is used to output the dust accumulation fault prediction result.
[0056] The convolutional neural network is used to extract features from the operating image of the input coal mill to obtain local features. For example, in one embodiment, the convolutional neural network includes five convolutional layers, namely a first convolutional layer, a second convolutional layer, a third convolutional layer, a fourth convolutional layer and a fifth convolutional layer connected in sequence.
[0057] It is known to those skilled in the art that the prediction model may also adopt other deep learning models, and this embodiment does not describe other deep learning models in detail.
[0058] As an improvement of the above embodiment, the training samples constructed before training the prediction model include coal mill operation images under different lighting conditions to improve the robustness of the model after parameter adjustment to different lighting environments.
[0059] Optionally, after obtaining a certain number of raw data of coal mill operation images, the number of samples can be increased through data augmentation. Coal mill operation images under different lighting conditions can be obtained by using brightness adjustment operations during data augmentation. Image samples under different lighting environments can be generated by increasing or decreasing image brightness.
[0060] Optionally, to improve the robustness of the prediction model to different light environments, an improved training process includes the following implementation steps:
[0061] A1. Select the parameters of the image classification model that has been pre-trained on a large-scale dataset as the initial parameters of the prediction model. The image classification model can be ResNet, Inception, etc.
[0062] A2. Fine-tune the parameters based on the constructed training samples containing coal mill operation images under different lighting conditions, and optimize the model parameters through methods such as cross-validation.
[0063] As another improvement of the above embodiment, the preprocessing process also includes: segmenting the powder accumulation area from the running image based on the threshold segmentation algorithm, normalization processing and image enhancement. Through the above preprocessing process, the accuracy of the powder accumulation fault prediction result is improved, and the powder accumulation area is segmented from the running image based on the threshold segmentation algorithm, which can compress the data required for storage, reduce the storage capacity, and greatly simplify the subsequent powder accumulation fault prediction process.
[0064] Among them, the normalization processing and image enhancement can adopt the processes in the common embodiment.
[0065] Optionally, a first specific implementation process of segmenting the powder accumulation area from the running image based on the threshold segmentation algorithm is as follows:
[0066] Convert the running image into a grayscale image;
[0067] Determine the segmentation threshold t;
[0068] The grayscale image is segmented using the segmentation threshold t to obtain the powder accumulation area image g(x,y). or,
[0069] Among them, Hf(x,y) represents the pixel grayscale in the grayscale image f(x,y).
[0070] Optionally, a second specific implementation process of segmenting the powder accumulation area from the running image based on the threshold segmentation algorithm is as follows:
[0071] Convert the running image into a grayscale image;
[0072] Determine the segmentation threshold range |t 1 -t 2 |;
[0073] Using the segmentation threshold range |t 1 -t 2 |Segment the grayscale image to obtain the powder accumulation area image g(x,y),
[0074]
[0075] Among them, Hf(x,y) represents the pixel grayscale in the grayscale image f(x,y), t 1 represents the lower limit of the segmentation threshold, t 2 Indicates the upper limit of the split threshold.
[0076] Optionally, a third specific implementation process of segmenting the powder accumulation area from the running image based on the threshold segmentation algorithm is as follows:
[0077] Convert the running image into a grayscale image;
[0078] Determine the segmentation threshold t;
[0079] The grayscale image is segmented using the segmentation threshold t to obtain the powder accumulation area image g(x,y).
[0080]
[0081] Among them, Hf(x,y) represents the pixel grayscale in the grayscale image f(x,y).
[0082] In the above three implementation processes, the segmentation threshold t preferably adopts a variable threshold, that is, an adaptive threshold. The extraction principle of the adaptive threshold is as follows: if for each pixel point, its gray value is represented by f(i, j), the adaptive threshold can be calculated by the following formula: t(i, j) = m(i, j) + kδ(i, j), where m(i, j) represents the gray mean of the pixel point in the neighborhood of the pixel point, δ(i, j) represents the gray variance of the pixel point in the neighborhood of the pixel point, and k represents the deviation coefficient, and k is preferably 0.2.
[0083] As another improvement of the above embodiment, the real-time operation image of the coal mill obtained is a visible light and infrared fusion image. Exemplarily, it can be collected by a visible light and infrared fusion device. Accordingly, the coal mill dust accumulation fault identification method provided by the present invention also includes:
[0084] S300. Identify the coal powder temperature from the visible light and infrared fusion image;
[0085] S400. Determine the level of coal powder accumulation fault of the coal mill using the coal powder temperature.
[0086] For example, using the pulverized coal temperature to determine the level of pulverized coal accumulation faults in the coal mill includes:
[0087] If it is determined according to the prediction result of the dust accumulation fault that the coal mill has a dust accumulation fault, and the coal powder temperature exceeds the first preset value, it is determined that the coal mill has a first level of dust accumulation fault;
[0088] If it is determined according to the prediction result of the dust accumulation fault that the coal mill has a dust accumulation fault, and the coal powder temperature does not exceed the first preset value, it is determined that the coal mill has a second level of dust accumulation fault;
[0089] The severity of the first level of powder accumulation failure is higher than the severity of the second level of powder accumulation failure.
[0090] In the above embodiment, by combining the coal powder temperature identified in the visible light and infrared fusion image with the integral fault prediction result, the classification of coal powder accumulation faults is achieved, and different warning and response measures can be set according to different classifications.
[0091] Device Embodiment
[0092] See also Figure 3 The present invention provides a coal mill dust accumulation fault identification device, comprising a real-time operation image acquisition module and a dust accumulation fault prediction module which are interconnected.
[0093] The real-time operation image acquisition module is used to obtain the real-time operation image of the coal mill.
[0094] The dust accumulation fault prediction module is used to input the preprocessed real-time operation image into the trained prediction model to generate a dust accumulation fault prediction result, which is normal operation or a specific dust accumulation fault type. The prediction model is used to extract the local features of the operation image and classify the dust accumulation fault according to the local features. After classification, the dust accumulation fault prediction result is output, and the local features represent the dust accumulation state.
[0095] Optionally, the training samples constructed during the prediction model training include coal mill operation images under different lighting conditions.
[0096] Optionally, the local characteristics include shape characteristics, texture characteristics, particle size distribution characteristics, coal quality characteristics, moisture characteristics, density characteristics, adhesion characteristics and coal powder composition characteristics.
[0097] Optionally, powder accumulation fault types include accumulation blockage, insufficient crushing, dust backflow, crushing machinery failure, improper operating parameters, excessive humidity and increased adhesion.
[0098] Optionally, the prediction model is a CNN network model, which includes a first convolutional layer, a second convolutional layer, a third convolutional layer, a fourth convolutional layer, a fifth convolutional layer, an activation layer, a pooling layer, a fully connected layer and an output layer, which are connected in sequence.
[0099] Optionally, the preprocessing includes segmenting the powder accumulation area from the running image based on a threshold segmentation algorithm, normalization processing and image enhancement.
[0100] Optionally, the operating image is a visible light and infrared fusion image, and the device further includes a coal powder temperature recognition module and a coal powder accumulation fault classification module, wherein:
[0101] A pulverized coal temperature recognition module is used to recognize the pulverized coal temperature from a visible light and infrared fusion image;
[0102] The dust accumulation fault classification module is used to determine the dust accumulation fault level of the coal mill using the coal powder temperature.
[0103] Optionally, the pulverized coal temperature is used to determine the level of pulverized coal accumulation fault in the coal mill, including:
[0104] If it is determined according to the prediction result of the dust accumulation fault that the coal mill has a dust accumulation fault, and the coal powder temperature exceeds the first preset value, it is determined that the coal mill has a first level of dust accumulation fault;
[0105] If it is determined according to the prediction result of the dust accumulation fault that the coal mill has a dust accumulation fault, and the coal powder temperature does not exceed the first preset value, it is determined that the coal mill has a second level of dust accumulation fault;
[0106] Among them, the severity of the first level of powder accumulation failure is higher than the second level of powder accumulation failure.
[0107] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.
[0108] System Example
[0109] See also Figure 4 The present invention provides a coal mill dust accumulation fault identification system, comprising any one of the coal mill dust accumulation fault identification devices and fault warning devices described in the above-mentioned device embodiments, wherein the fault warning device is used to provide fault warning based on the dust accumulation fault prediction result generated by the coal mill dust accumulation fault identification device.
[0110] Optionally, the coal mill dust accumulation fault identification system also includes a model optimization device, which is connected to the coal mill dust accumulation fault identification device and is used for parameter setting and system maintenance. The parameter types that can be set by the user through the interactive interface include training parameters of the prediction model, parameters of the image preprocessing process, etc. System maintenance includes updating the prediction network model, adding new image data, etc.
[0111] Optionally, the coal mill dust accumulation fault identification system also includes a data query and analysis device, which is connected to the coal mill dust accumulation fault identification device and is used to query historical dust accumulation fault prediction results, analyze the frequency and regularity of dust accumulation faults, etc.
[0112] On the other hand, the present invention also provides an electronic device, including a memory and a processor, wherein the memory stores a computer program that can be run on the processor, and after loading and executing the computer program, the processor can implement any one of the coal mill powder accumulation fault identification methods described in the above method embodiments.
[0113] On the other hand, the present invention further provides a machine-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, it implements any one of the coal mill dust accumulation fault identification methods described in the above method embodiments.
[0114] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.
[0115] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A method for identifying coal mill dust accumulation fault, characterized in that: include: Get real-time operation images of coal mill; Inputting the preprocessed operation image into a trained prediction model to generate a powder accumulation fault prediction result, wherein the powder accumulation fault prediction result is a normal operation or a specific powder accumulation fault type; The prediction model is used to extract local features of the running image, and classify the powder accumulation fault according to the local features, and output the powder accumulation fault prediction result after classification, and the local features characterize the powder accumulation state.
2. The method according to claim 1, characterized in that The training samples constructed during the prediction model training include coal mill operation images under different lighting conditions.
3. The method according to claim 1, characterized in that The local characteristics include shape characteristics, texture characteristics, particle size distribution characteristics, coal quality characteristics, moisture characteristics, density characteristics, adhesion characteristics and coal powder composition characteristics.
4. The method according to claim 1, characterized in that: The types of powder accumulation failures include accumulation blockage, insufficient crushing, dust backflow, crushing machinery failure, improper operating parameters, excessive humidity and increased adhesion.
5. The method according to claim 1, characterized in that The prediction model is a CNN network model, which includes a first convolutional layer, a second convolutional layer, a third convolutional layer, a fourth convolutional layer, a fifth convolutional layer, an activation layer, a pooling layer, a fully connected layer and an output layer connected in sequence.
6. The method according to claim 1, characterized in that The preprocessing includes segmenting the powder accumulation area from the running image based on a threshold segmentation algorithm, normalization processing and image enhancement.
7. The method for identifying coal mill dust accumulation fault according to claim 1, characterized in that: The running image is a visible light and infrared fusion image, and the method further includes: identifying the pulverized coal temperature from the visible light and infrared fusion image; The coal powder temperature is used to determine the coal mill dust accumulation fault level.
8. The method for identifying coal mill dust accumulation fault according to claim 7, characterized in that: The method of determining the coal powder accumulation fault level of the coal mill by using the coal powder temperature includes: If it is determined according to the prediction result of the dust accumulation fault that the coal mill has a dust accumulation fault, and the coal powder temperature exceeds the first preset value, it is determined that the coal mill has a first level of dust accumulation fault; If it is determined according to the prediction result of the dust accumulation fault that the coal mill has a dust accumulation fault, and the coal powder temperature does not exceed the first preset value, it is determined that the coal mill has a second level of dust accumulation fault; Among them, the severity of the first level of powder accumulation failure is higher than the second level of powder accumulation failure.
9. A coal mill dust accumulation fault identification device, characterized in that: The device comprises: A real-time operation image acquisition module is used to obtain the real-time operation image of the coal mill; A dust accumulation fault prediction module is used to input the preprocessed operation image into a trained prediction model to generate a dust accumulation fault prediction result, wherein the dust accumulation fault prediction result is normal operation or a specific dust accumulation fault type; The prediction model is used to extract local features of the running image, and classify the powder accumulation fault according to the local features, and output the powder accumulation fault prediction result after classification, and the local features characterize the powder accumulation state.
10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, a method for identifying coal mill dust accumulation fault according to any one of claims 1 to 8 is implemented.
11. A storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, a method for identifying coal mill dust accumulation fault according to any one of claims 1 to 8 is implemented.