Deep convolutional neural network learning-based hearth water-cooled wall anomaly detection and identification method

Abstract

The invention discloses a hearth water wall anomaly detection and identification method based on deep convolutional neural network learning, and relates to the field of hearth water wall anomaly detection. In order to solve the technical problems of inaccurate identification and inaccurate classification of detection, identification and classification of the water cooling wall of the cooling pipe in the prior art, the technical scheme provided by the invention is as follows: the method comprises the following steps of: obtaining original pipeline data shot at each pipeline position in a hearth water cooling wall system of a thermal power plant; preprocessing the original pipeline data, and dividing the original pipeline data into a training data sample and a test data sample; respectively carrying out hybrid enhancement processing on the training data sample and the test data sample; and inputting the training data sample subjected to hybrid enhancement processing into a deep convolutional neural network for model training to obtain a trained model. The method can be applied to abnormal detection and identification of the hearth water wall.

Description

Technical Field

[0001] The present invention relates to the field of furnace water-cooled wall anomaly detection, and in particular to a furnace water-cooled wall anomaly detection and recognition algorithm based on a deep convolutional neural network. Background Art

[0002] Thermal power generation is the main force of power development in modern society. A thermal power plant is a factory that uses combustibles (such as coal) as fuel to produce electricity. A modern thermal power plant is a large and complex factory that produces electricity and heat energy. It consists of the following five systems: ① Fuel system. ② Combustion system. ③ Steam-water system. ④ Electrical system. ⑤ Control system. Among the above systems, the most important equipment is the boiler, steam turbine and generator, which are installed in the main plant of the power plant.

[0003] The boiler heating in thermal power plants needs timely cooling. Due to the heating caused by coal combustion and the accumulation of ash and slagging in the boiler, the cooling tube wall inside the furnace will be damaged. This phenomenon of water-cooled wall damage not only causes a waste of resources, but also seriously causes the boiler to stop operating and the thermal power plant to stop working.

[0004] In the current prior art, the detection, identification and classification of cooling tube water-cooled walls have problems such as inaccurate identification and classification. How to accurately detect and classify various abnormal conditions of the furnace water-cooled wall is an urgent problem to be solved. Summary of the invention

[0005] In order to solve the technical problems existing in the prior art, such as inaccurate identification and classification of the detection, identification and classification of the cooling tube water wall, the technical solution provided by the present invention is as follows:

[0006] A method for constructing a furnace water-cooled wall anomaly detection and recognition model based on deep convolutional neural network learning, comprising:

[0007] The steps of obtaining the original pipeline data photographed at each pipeline position in the water-cooled wall system of the furnace of the thermal power plant;

[0008] Preprocessing the original pipeline data into training data samples and testing data samples;

[0009] The step of performing mixed enhancement processing on the training data samples and the test data samples respectively;

[0010] The training data samples that have been processed by hybrid enhancement are input into the deep convolutional neural network for model training to obtain the steps of the trained model.

[0011] Furthermore, a preferred embodiment is provided, wherein the preprocessing includes manually classifying and marking the original pipeline data, and dividing the data into normal data, wear data and mechanical damage data.

[0012] Furthermore, a preferred embodiment is provided, wherein the deep convolutional neural network is a YOLOv8 model.

[0013] Furthermore, a preferred implementation is provided, which includes the step of constructing a leakage data label index when performing hybrid enhancement processing.

[0014] Based on the same inventive concept, the present invention also provides a furnace water wall anomaly detection and recognition model construction device based on deep convolutional neural network learning, comprising:

[0015] A module for obtaining the original pipeline data captured at each pipeline position in the water-cooled wall system of the furnace of a thermal power plant;

[0016] Preprocessing the raw pipeline data into modules of training data samples and testing data samples;

[0017] A module for performing mixed enhancement processing on the training data samples and the test data samples respectively;

[0018] The training data samples that have undergone hybrid enhancement processing are input into the deep convolutional neural network for model training to obtain the modules of the trained model.

[0019] Based on the same inventive concept, the present invention also provides a furnace water-cooled wall anomaly detection and identification method based on deep convolutional neural network learning, comprising:

[0020] A step of acquiring pipeline data in real time and preprocessing the data acquired in real time;

[0021] The step of inputting the preprocessed pipeline data into the obtained model to obtain a prediction result;

[0022] The step of comparing the prediction result with the preset leakage data label index to determine the abnormal condition of the pipeline.

[0023] Based on the same inventive concept, the present invention also provides a furnace water-cooled wall anomaly detection and identification device based on deep convolutional neural network learning, comprising:

[0024] A module that obtains pipeline data in real time and pre-processes the data obtained in real time;

[0025] A module for inputting the preprocessed pipeline data into the obtained model to obtain a prediction result;

[0026] A module that compares the prediction results with the preset leakage data label index to determine the abnormal situation of the pipeline.

[0027] Based on the same inventive concept, the present invention also provides a computer storage medium for storing a computer program. When the computer program is read by a computer, the computer executes the method described.

[0028] Based on the same inventive concept, the present invention also provides a computer, including a processor and a storage medium. When the processor reads a computer program stored in the storage medium, the computer executes the method described.

[0029] Based on the same inventive concept, the present invention also provides a computer program product, which is a computer program. When the computer program is executed, the method described above is implemented.

[0030] Compared with the prior art, the technical solution provided by the present invention is beneficial in that:

[0031] The present invention provides a furnace water wall anomaly detection and recognition method based on deep convolutional neural network learning, which improves the quality and diversity of data, so that the model can better identify and classify different abnormal situations during the training process. Hybrid enhancement processing overcomes the problem of insufficient original data by increasing the amount and complexity of data, and improves the generalization ability of the model.

[0032] The present invention provides a furnace water wall anomaly detection and recognition method based on deep convolutional neural network learning. The YOLOv8 model has powerful feature extraction and classification capabilities and can efficiently process complex image data. After the model is fully trained, it can accurately identify the wear, mechanical damage and normal state of the furnace water wall pipe, greatly improving the accuracy and efficiency of detection.

[0033] The present invention provides a furnace water wall anomaly detection and identification method based on deep convolutional neural network learning. Real-time data processing ensures the timeliness of detection, can quickly respond to abnormal situations, and avoid safety hazards caused by delays. Through real-time prediction, the system can continuously monitor the status of the furnace water wall, provide instant feedback, and enhance the reliability of the system.

[0034] The present invention provides a furnace water wall anomaly detection and identification method based on deep convolutional neural network learning. Accurate anomaly detection can distinguish different types of anomalies such as wear and mechanical damage, and provide accurate reference information for maintenance and repair. By comparing label indexes, the accuracy and confidence of detection are improved, and the possibility of false positives and false negatives is reduced.

[0035] The present invention provides a furnace water wall anomaly detection and recognition method based on deep convolutional neural network learning. Through hybrid enhancement and preprocessing technology, the quality of training data is greatly improved and the recognition ability of the model is enhanced. The YOLOv8 model is used for deep learning training to achieve efficient and accurate anomaly detection. Real-time processing and prediction ensure the response speed and reliability of the system, which is significantly better than traditional methods.

[0036] The present invention provides a furnace water-cooled wall anomaly detection and identification method based on deep convolutional neural network learning, which can be applied to furnace water-cooled wall anomaly detection and identification work. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 The present invention is a flowchart of an embodiment of a method for detecting and classifying furnace water-cooled wall anomalies based on a deep convolutional neural network. DETAILED DESCRIPTION

[0038] In order to make the advantages and benefits of the technical solution provided by the present invention more clearly reflected, the technical solution provided by the present invention is now further described in detail with reference to the accompanying drawings, specifically:

[0039] Embodiment 1: This embodiment provides a method for constructing a furnace water wall anomaly detection and recognition model based on deep convolutional neural network learning, including:

[0040] The steps of obtaining the original pipeline data photographed at each pipeline position in the water-cooled wall system of the furnace of the thermal power plant;

[0041] Preprocessing the original pipeline data into training data samples and testing data samples;

[0042] The step of performing mixed enhancement processing on the training data samples and the test data samples respectively;

[0043] The training data samples that have been processed by hybrid enhancement are input into the deep convolutional neural network for model training to obtain the steps of the trained model.

[0044] Implementation method 2. This implementation method is a further limitation of the method for constructing a furnace water-cooled wall anomaly detection and identification model based on deep convolutional neural network learning provided in implementation method 1. The preprocessing includes manually classifying and labeling the original pipeline data, and dividing the data into normal data, wear data, and mechanical damage data.

[0045] Implementation method three: This implementation method is a further limitation of the method for constructing a furnace water-cooled wall anomaly detection and recognition model based on deep convolutional neural network learning provided in implementation method one, and the deep convolutional neural network is a YOLOv8 model.

[0046] Implementation method 4: This implementation method is a further limitation of the method for constructing a furnace water-cooled wall anomaly detection and recognition model based on deep convolutional neural network learning provided in implementation method 1. When performing hybrid enhancement processing, it also includes the step of constructing a leakage data label index.

[0047] Embodiment 5: This embodiment provides a furnace water wall anomaly detection and recognition model construction device based on deep convolutional neural network learning, including:

[0048] A module for obtaining the original pipeline data captured at each pipeline position in the water-cooled wall system of the furnace of a thermal power plant;

[0049] Preprocessing the raw pipeline data into modules of training data samples and testing data samples;

[0050] A module for performing mixed enhancement processing on the training data samples and the test data samples respectively;

[0051] The training data samples that have undergone hybrid enhancement processing are input into the deep convolutional neural network for model training to obtain the modules of the trained model.

[0052] Embodiment 6: This embodiment provides a furnace water wall anomaly detection and identification method based on deep convolutional neural network learning, including:

[0053] A step of acquiring pipeline data in real time and preprocessing the data acquired in real time;

[0054] The step of inputting the preprocessed pipeline data into the model obtained in the first embodiment to obtain a prediction result;

[0055] The step of comparing the prediction result with the preset leakage data label index to determine the abnormal condition of the pipeline.

[0056] Specifically,

[0057] The technical solution provided in this embodiment includes:

[0058] Step S10: Obtain pipeline data

[0059] Input: Pipeline data captured at each pipe location in the water-cooled wall system of the furnace of a thermal power plant. Process:

[0060] Acquire image data of each pipe location in the water-cooled wall system of the furnace of a thermal power plant.

[0061] These data are used for subsequent preprocessing and model training.

[0062] Output: original furnace water wall pipe data.

[0063] Step S20: Data preprocessing

[0064] Input: Original furnace water wall pipe data. Process:

[0065] The pre-collected pipeline data is divided into training data samples and test data samples.

[0066] The training data samples and the test data samples are mixed and enhanced respectively.

[0067] Sub-steps:

[0068] Z1: Perform manual classification and labeling to divide the data into normal data, wear data, and mechanical damage data.

[0069] Z2: Build a label index for the leaked data, including the type and location of the outliers.

[0070] Output: training data samples and test data samples after hybrid enhancement.

[0071] Step S30: Model training

[0072] Input: training data samples after hybrid enhancement. Process:

[0073] The preprocessed data is input into a deep convolutional neural network (YOLOv8) for training.

[0074] After training, the model can recognize wear and tear, mechanical damage, and normal conditions.

[0075] The trained model is tested using the test data samples after hybrid enhancement.

[0076] Output: Trained YOLOv8 model.

[0077] Step S40: Real-time data processing

[0078] Input: Real-time furnace water wall pipe data and trained model. Process:

[0079] Acquire pipeline data in real time and pre-process the data.

[0080] Input the preprocessed real-time data into the trained model.

[0081] Obtain prediction results and perform analysis.

[0082] Output: analysis results and prediction results.

[0083] Step S50: Anomaly detection

[0084] Input: prediction results and leaked data label index. Process:

[0085] Compare the prediction results with the missing data label index.

[0086] Determine abnormal situations.

[0087] Output: The final judged abnormal situation.

[0088] Embodiment 7: This embodiment provides a furnace water wall abnormality detection and identification device based on deep convolutional neural network learning, including:

[0089] A module that obtains pipeline data in real time and pre-processes the data obtained in real time;

[0090] A module for inputting the preprocessed pipeline data into the model obtained in the first embodiment to obtain a prediction result;

[0091] A module that compares the prediction results with the preset leakage data label index to determine the abnormal situation of the pipeline.

[0092] Embodiment 8: This embodiment provides a computer storage medium for storing a computer program. When the computer program is read by a computer, the computer executes the method provided in embodiment 1.

[0093] Embodiment 9: This embodiment provides a computer, including a processor and a storage medium. When the processor reads a computer program stored in the storage medium, the computer executes the method provided in embodiment 1.

[0094] Embodiment 10: This embodiment provides a computer program product, which is a computer program. When the computer program is executed, the method provided in embodiment 1 is implemented.

[0095] Implementation Method XI: Combination Figure 1 This embodiment is described in detail by using specific examples to further describe the technical solution provided above. Specifically:

[0096] like Figure 1 As shown in the flow chart, the technical solution provided by this embodiment includes:

[0097] Pre-fetching data

[0098] Steps: Obtain various damage data of various positions and areas of the furnace water-cooled wall in advance.

[0099] Effect: Ensure that the acquired data can fully cover various possible damage conditions, and the data features are obvious and unified.

[0100] Data preprocessing

[0101] Steps: Preprocess the pre-collected furnace water-cooled wall data.

[0102] The pre-collected furnace water-cooled wall pipe data is divided into training data samples and test data samples.

[0103] The steps of dividing the pre-collected furnace water wall pipe data into training data samples and test data samples include:

[0104] Sub-step Z1: First, perform simple manual classification on the pre-collected data, use image feature labeling software to simply label and classify the collected abnormal data of the furnace water-cooled wall, and divide the labeled image data into training data samples and test data samples.

[0105] Sub-step Z2: The types of furnace water wall pipe data include normal data, wear data and mechanical damage data

[0106] Sub-step Z3: constructing a leakage data label index based on the labels of each type of abnormal data of the furnace water-cooled wall pipeline data, each leakage data label index includes the abnormal point type and the abnormal point position.

[0107] Use labeling software to label wear and mechanical damage in the training data.

[0108] The training data samples and test data samples are mixed and enhanced (Mixup, CutMix) respectively.

[0109] The training data sample and the test data sample are respectively subjected to mixed enhancement processing, specifically comprising the following sub-steps:

[0110] Sub-step A1: Select two random training samples:

[0111] Sub-step A2: Randomly sample the mixing parameters

[0112] Sub-step A3: Generate mixed samples:

[0113] Sub-step A4: Save the mixed training data

[0114] Effect: Improve data quality and diversity, and provide more comprehensive input for model training.

[0115] Model Training

[0116] Steps: Input the preprocessed furnace water-cooled wall data into the improved YOLOv8 model for training.

[0117] Use the Context and Spatial Feature Calibration module to improve YOLOv8 and enhance the semantic segmentation capability of the model.

[0118] Define the loss function and optimizer for the improved YOLOv8 model.

[0119] Designing and improving the YOLOv8 model includes the following small steps:

[0120] Small step 1, first of all, you need to deeply understand the working principle and function of the CSFCN module. CSFCN is mainly used to enhance the ability of semantic segmentation models, especially in processing contextual information and spatial features. It usually includes some specific convolutional layers, attention mechanisms or multi-scale feature fusion methods.

[0121] Small step 2, choose a suitable backbone network as the basis of YOLOv8. The recommended choice can be Darknet or other well-performing lightweight convolutional neural network architecture to ensure that the backbone network can effectively extract image features.

[0122] Small step 3, embed the CSFCN module into the YOLOv8 architecture. The CSFCN module is usually used in the feature extraction stage after the backbone network.

[0123] Step 4: Adjust the parameters and hyperparameters of the CSFCN module according to the actual dataset and task requirements. The design of the CSFCN module can be optimized based on the performance during training to ensure that it can effectively enhance the performance of the YOLOv8 model.

[0124] Small step 5, prepare the training data set and perform necessary data preprocessing and enhancement. Make sure the data format meets the input requirements of the YOLOv8 model and apply appropriate data enhancement methods during training.

[0125] In small step 6, select a loss function (such as the loss function commonly used in YOLOv8) and optimizer suitable for the target detection task, and prepare the model for training.

[0126] In step 7, start training the YOLOv8 model and monitor the performance during training. Adjust the parameters of the model and CSFCN module based on the evaluation results of the validation set or test set to optimize the detection accuracy and generalization ability of the model.

[0127] Small step 8: Analyze the training results and select the best model for deployment. In actual applications, monitor the performance of the model and continuously optimize it to adapt to different scenarios and data changes.

[0128] Through the above small steps, we can effectively design and implement the process of using the CSFCN module to improve the YOLOv8 model and improve the performance of the model in the target detection task.

[0129] After the test is successful, save the trained YOLOv8 model.

[0130] Effect: The trained deep convolutional neural network model is obtained, which can better distinguish the damage of the furnace water-cooled wall.

[0131] Real-time data acquisition and processing

[0132] Steps: Obtain abnormal data of the furnace water-cooled wall in real time and perform preprocessing.

[0133] New fault image types are collected and hybrid enhancement preprocessing is performed on real-time furnace water-cooled wall data.

[0134] Effect: Real-time monitoring and preprocessing of data to ensure that the system can respond and detect anomalies in a timely manner.

[0135] The pre-processed furnace water-cooled wall data is input into the deep convolutional neural network model for learning. The trained network contains multiple labels, namely wear, mechanical damage and normal. Then the accuracy of the damage data learning is calculated.

[0136] In the specific implementation, this step will input the training data samples after hybrid enhancement processing into the improved YOLOv8 model for training; use the test data samples after hybrid enhancement processing to test the trained improved YOLOv8, and save the trained model after the test is successful.

[0137] Model Optimization

[0138] Steps: Send the preprocessed real-time data into the trained deep convolutional neural network model to optimize the deep convolutional neural network.

[0139] Effect: Continuously optimize the model to improve its accuracy and reliability in anomaly classification detection.

[0140] Promotion and application

[0141] Steps: Use the continuously optimized model to carry out research on the classification and detection of furnace water-cooled wall anomalies using visible light images, and promote the application of the optimized algorithm model in other similar image recognition fields.

[0142] Effect: Expand the application scope of the algorithm and improve the applicability and robustness of the model in different scenarios.

[0143] Detailed Description

[0144] Pre-fetching data

[0145] The original pipeline data captured at each pipeline position in the water-cooled wall system of the furnace of a thermal power plant is obtained, including appearance data, label data, and scale data.

[0146] Ensure that damage data covers different types such as wear and tear, mechanical damage, etc., with obvious and uniform characteristics, providing a good basis for subsequent processing.

[0147] Data preprocessing

[0148] Classification and labeling:

[0149] The pre-collected data is divided into training data samples and test data samples.

[0150] Use labeling software to label wear, mechanical damage, etc. in training data samples in detail.

[0151] Hybrid Enhancement Processing:

[0152] Mixup method: select two images and their corresponding labels, generate new images and labels through linear mixing, increase data diversity and reduce overfitting.

[0153] CutMix method: select two images and their corresponding labels, generate new training samples by cutting and pasting, and improve the model's tolerance to partial occlusion and generalization ability.

[0154] Model Training

[0155] The training data samples that have undergone hybrid enhancement processing are input into the improved YOLOv8 model for training.

[0156] Use the Context and Spatial Feature Calibration module to enhance the semantic segmentation capability of the YOLOv8 model.

[0157] Define the loss function and optimizer of the improved model, and save the model after successful training.

[0158] Model improvements:

[0159] The CSFCN module improves the image processing accuracy and efficiency of the model by capturing and utilizing contextual information and spatial features in images.

[0160] Enhanced the detection capability of small targets, especially the identification of small defects in complex backgrounds.

[0161] Real-time data acquisition and processing

[0162] Acquire abnormal data of the furnace water-cooled wall in real time, and use hybrid enhancement methods to preprocess the real-time data to ensure data quality and consistency.

[0163] Real-time monitoring and analysis of real-time data ensures that the system can quickly respond to abnormal situations and provide timely and accurate detection results.

[0164] Model Optimization

[0165] The preprocessed real-time data is input into the trained deep convolutional neural network model for continuous optimization.

[0166] By continuously training and optimizing new data, the model's anomaly classification detection accuracy and reliability can be improved.

[0167] judge

[0168] Acquire the pipeline data of the furnace water wall in real time, preprocess the pipeline data, send the preprocessed real-time furnace water wall pipeline data to the trained model, obtain the prediction results through the trained model, analyze the results, and select the best model for deployment. In actual applications, monitor the performance of the model and continuously optimize it to adapt to different scenarios and data changes.

[0169] The prediction result is compared with the leakage data label index of the corresponding leakage point type to determine the abnormal situation.

[0170] In this embodiment, an improved YOLOv8 model is established and trained through the image data of the furnace water-cooled wall pipe. The trained model can judge its abnormal conditions, including wear, mechanical damage and normal. Then, the real-time data obtained later can be analyzed. First, the received real-time data is pre-processed with hybrid enhancement, and then the data after hybrid enhancement pre-processing is sent to the improved YOLOv8 for learning to obtain the final prediction result.

[0171] at last:

[0172] The optimized model is used to study the classification and detection of furnace water-cooled wall anomalies, and the algorithm is applied to other similar image recognition fields, such as industrial inspection and medical image analysis.

[0173] Improve the applicability of the model in different scenarios and enhance the robustness and scalability of the algorithm.

[0174] Through the above steps, the scheme can efficiently and accurately detect and classify abnormal conditions of the furnace water-cooled wall, and has strong real-time and applicability, providing a strong guarantee for the safe operation of thermal power plants.

[0175] The technical solution provided by the present invention is further described in detail above through several specific implementation modes in order to highlight the advantages and benefits of the technical solution provided by the present invention. However, the several specific implementation modes described above are not intended to be used as limitations on the present invention. Any reasonable modification and improvement of the present invention, combination of implementation modes and equivalent substitution within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for constructing a furnace water wall anomaly detection and recognition model based on deep convolutional neural network learning, It is characterized in that include: The steps of obtaining the original pipeline data photographed at each pipeline position in the water-cooled wall system of the furnace of the thermal power plant; Preprocessing the original pipeline data into training data samples and testing data samples; The step of performing mixed enhancement processing on the training data samples and the test data samples respectively; The training data samples that have been processed by hybrid enhancement are input into the deep convolutional neural network for model training to obtain the steps of the trained model.

2. According to claim 1, a method for constructing a furnace water wall anomaly detection and recognition model based on deep convolutional neural network learning, It is characterized in that The preprocessing includes manually classifying and marking the original pipeline data, and dividing the data into normal data, wear data and mechanical damage data.

3. According to claim 1, a method for constructing a furnace water wall anomaly detection and recognition model based on deep convolutional neural network learning, It is characterized in that The deep convolutional neural network is a YOLOv8 model.

4. According to claim 1, a method for constructing a furnace water wall anomaly detection and recognition model based on deep convolutional neural network learning, It is characterized in that When performing hybrid enhancement processing, a step of constructing a leaked data label index is also included.

5. A device for building a furnace water wall anomaly detection and recognition model based on deep convolutional neural network learning, It is characterized in that include: A module for obtaining the original pipeline data captured at each pipeline position in the water-cooled wall system of the furnace of a thermal power plant; Preprocessing the raw pipeline data into modules of training data samples and testing data samples; A module for performing mixed enhancement processing on the training data samples and the test data samples respectively; The training data samples that have undergone hybrid enhancement processing are input into the deep convolutional neural network for model training to obtain the modules of the trained model.

6. A method for detecting and identifying furnace water-cooled wall anomalies based on deep convolutional neural network learning. It is characterized in that include: A step of acquiring pipeline data in real time and preprocessing the data acquired in real time; The step of inputting the pre-processed pipeline data into the model obtained in claim 1 to obtain a prediction result; The step of comparing the prediction result with the preset leakage data label index to determine the abnormal condition of the pipeline.

7. A furnace water wall anomaly detection and recognition device based on deep convolutional neural network learning, It is characterized in that include: A module that obtains pipeline data in real time and pre-processes the data obtained in real time; A module for inputting the preprocessed pipeline data into the model obtained in claim 1 to obtain a prediction result; A module that compares the prediction results with the preset leakage data label index to determine the abnormal situation of the pipeline.

8. Computer storage media for storing computer programs, It is characterized in that When the computer program is read by a computer, the computer executes the method of claim 1 .

9. A computer, including a processor and a storage medium, It is characterized in that When the processor reads the computer program stored in the storage medium, the computer executes the method of claim 1 .

10. Computer program product, being a computer program, It is characterized in that When the computer program is executed, the method of claim 1 is implemented.

Images