Training method of pulverized coal flame temperature measurement model and pulverized coal frame temperature measurement method

Abstract

To provide a training method of a pulverized coal flame temperature measurement model and pulverized coal flame temperature measurement method.SOLUTION: A method includes the steps of: acquiring a sample flame radiation image; pre-processing the sample flame radiation image to obtain a true temperature corresponding to the sample flame radiation image, and constructing a sample set using the sample flame radiation image and the temperature corresponding to the sample flame radiation image as samples; inputting a training flame radiation image into a pre-constructed initial pulverized coal flame temperature measurement mode and outputting a first temperature corresponding to the training flame radiation image; and calculating an error between the first temperature and the true training temperature, and updating parameters of the initial pulverized coal flame temperature measurement model with the error until the error converges, thereby obtaining a target pulverized coal flame temperature measurement model.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to the technical field of radiographic imaging, and more particularly to a method for training a pulverized coal flame temperature measurement model and a method for measuring pulverized coal flame temperature. [Background technology]

[0002] Combustion is a chemical reaction that emits light and heat. Pulverized coal combustion remains the primary source of energy production and a major method of resource utilization. Temperature is a key parameter in the combustion process because it can accurately reflect the state of the combustion process. Compared with single-point or global temperature measurements, three-dimensional temperature field distribution can directly characterize the three-dimensional structure and flow characteristics of the flame. It also provides known parameters for predicting the chemical reactions, reaction rates, and component concentration distributions occurring during the combustion process. For example, in industries such as power generation, high-temperature and high-pressure burners are often involved, and the safety and service life of the equipment are both affected by combustion stability. The temperature image data provided can accurately reflect the local and global operating conditions of the equipment, allowing for real-time monitoring of flame combustion stability. This not only allows for temperature-based determination of pollutant generation such as SOx and NOx, but also for improving resource utilization and thermal efficiency. Furthermore, non-contact temperature measurements can provide insight into instantaneous high-energy diffusion and energy efficiency during blasting operations and explosive weapon testing. Therefore, the real-time measurement technology of high temperature flame temperature image is of great significance to scientific research and production.

[0003] In related technologies, due to the complexity and unwell-posedness of the problem, it is almost impossible to obtain a direct analytical solution using the first temperature of a radiographic image. Furthermore, due to the radiation characteristics of the flame itself, it is quite difficult to obtain a numerical solution. First, the sensitivity of the device to flame radiation signals and the sensitivity of temperature to flame radiation intensity are high, so even a slight error or perturbation can significantly interfere with the measurement results. Second, the optical parameters, temperature, and radiation intensity are mutually influencing each other, and the three jointly cause nonlinearity in the problem to be solved. Finally, due to the high resolution of the device itself, the number of equations in the problem to be solved is very large, and the coefficient matrix is ​​often a sparse matrix, which results in high computational costs and long operating times. Summary of the Invention

[0004] In view of this, the present invention provides a method for training a pulverized coal flame temperature measurement model and a method for measuring pulverized coal flame temperature, which solve the problems of high calculation costs and long operation times in the real-time measurement of conventional high-temperature flame temperature images.

[0005] In a first aspect, the present invention provides a method for manufacturing a semiconductor device comprising: acquiring a sample flame radiographic image; pre-processing the sample flame radiation images to obtain true temperatures corresponding to the sample flame radiation images, and constructing a sample set using the sample flame radiation images and the temperatures corresponding to the sample flame radiation images as samples, the sample set including a training set, the training set including the training flame radiation images and the true training temperatures corresponding to the training flame radiation images; inputting the training flame radiation image into a pre-constructed initial pulverized coal flame temperature measurement model, and outputting a first temperature corresponding to the training flame radiation image; a step of calculating an error between the first temperature and a true training temperature, and updating parameters of an initial pulverized coal flame temperature measurement model based on the error until the error converges, thereby obtaining a target pulverized coal flame temperature measurement model, wherein the target pulverized coal flame temperature measurement model is used to measure the pulverized coal flame temperature.

[0006] In the present invention, flame radiation images of high-temperature flames are preprocessed, and the flame radiation images and the temperatures corresponding to the flame radiation images are constructed as a training set. A neural network model is trained using the training set to obtain a pulverized coal flame temperature measurement model, thereby realizing real-time measurement of pulverized coal flame temperature images, which can significantly reduce the real-time measurement time and ensure real-time measurement of flame temperature images with high measurement accuracy. This contributes to understanding the temperature distribution pattern of combustion flames, and by exploring the physical action process and chemical reaction mechanism, it can evaluate combustion efficiency and provide effective data support for burner evaluation and improvement.

[0007] In an alternative embodiment, the step of pre-processing the sample flame radiation image to obtain a true temperature corresponding to the sample flame radiation image comprises: The method includes solving the furnace radiative transfer equation for the sample flame radiation image to obtain a true temperature corresponding to the sample flame radiation image.

[0008] In this manner, by solving the furnace radiative transfer equation and determining the true temperatures corresponding to the flame radiation images, subsequent images and the true temperatures corresponding to the images can be used to facilitate training of a pulverized coal flame temperature measurement model.

[0009] In an optional embodiment, the step of calculating an error between the first temperature and the true training temperature, and updating parameters of the initial pulverized coal flame temperature measurement model according to the error until the error converges, to obtain a target pulverized coal flame temperature measurement model, includes: official Calculating the error between the first temperature and the temperature according to TIFF2025165849000002.tif11170, where N is the total number of grids in the training flame radiographic image, and T m is the first temperature, T r is the true training temperature, L is the error, and official TIFF2025165849000003.tif14170, updating the parameters of the initial pulverized coal flame temperature measurement model until the error converges, to obtain a target pulverized coal flame temperature measurement model, wherein: TIFF2025165849000004.tif5170 is the weight of the previous iteration, TIFF2025165849000005.tif8170 is the updated weight, TIFF2025165849000006.tif3170 is the learning rate, L is the error, k, j represent the kth layer network and jth neurons of the initial pulverized coal flame temperature measurement model, h is the activation function, and w and b are the weight and offset of the initial pulverized coal flame temperature measurement model, respectively.

[0010] In this form, by training the initial pulverized coal flame temperature measurement model, the initial pulverized coal flame temperature measurement model can find the relationship between input and output from a priori knowledge, and realizes direct solution without the need to actually model the flame physicochemical rules in the combustion process, thus having the advantages of very strong large-volume data processing ability, high efficiency, high robustness, and high adaptability under different conditions.

[0011] In an alternative embodiment, the sample set further includes a test set, the test set including test flame radiographic images and true test temperatures corresponding to the test flame radiographic images; The method further comprises: inputting the test flame radiographic image into a target pulverized coal flame temperature measurement model to obtain a second temperature corresponding to the test flame radiographic image; calculating a second error between the second temperature and the true test temperature; and if the second error is equal to or greater than a preset threshold, updating parameters of the target pulverized coal flame temperature measurement model with the second error and performing a next iteration of training, stopping the iteration until the second error between the second temperature and the true test temperature is less than a preset threshold, wherein the preset threshold is for evaluating the pulverized coal flame temperature measurement model.

[0012] In this model, the test set image data is input into the model, and the second temperature is output and compared with the true temperature to obtain the error. If the error is large, the model parameters are re-adjusted and re-trained. When the error is within a reasonable range, the model construction is complete, thereby further improving the measurement accuracy of the model for real-time flame temperature image measurement.

[0013] In a second aspect, the present invention provides a method for producing a method of manufacturing a semiconductor device comprising: acquiring a flame radiation image of the object to be measured; a step of inputting a flame radiation image of a measurement target into a pulverized coal flame temperature measurement model, and outputting a temperature distribution image corresponding to the flame radiation image of the measurement target, wherein the pulverized coal flame temperature measurement model is obtained by training using the training method for a pulverized coal flame temperature measurement model according to any one of the first aspects.

[0014] In the present invention, the trained pulverized coal flame temperature measurement model realizes real-time measurement of pulverized coal flame temperature images, significantly reduces the real-time measurement time, ensures real-time measurement of flame temperature images, and has high measurement accuracy, which contributes to understanding the temperature distribution pattern of combustion flames, and by exploring the physical action process and chemical reaction mechanism, it is possible to evaluate combustion efficiency and even evaluate and improve burners.

[0015] In a third aspect, the present invention provides a method for producing a method of manufacturing a semiconductor device comprising: a first image acquisition module for acquiring a sample flame radiographic image; a data pre-processing module for pre-processing the sample flame radiation images to obtain true temperatures corresponding to the sample flame radiation images, and constructing a sample set using the sample flame radiation images and the temperatures corresponding to the sample flame radiation images as samples, the sample set including a training set, the training set including the training flame radiation images and the true training temperatures corresponding to the training flame radiation images; a first temperature output module for inputting the training flame radiation image into a pre-constructed initial pulverized coal flame temperature measurement model and outputting a first temperature corresponding to the training flame radiation image; and a model training module for calculating an error between a first temperature and a true training temperature, updating parameters of an initial pulverized coal flame temperature measurement model based on the error until the error converges, and obtaining a target pulverized coal flame temperature measurement model, wherein the target pulverized coal flame temperature measurement model is used to measure the pulverized coal flame temperature.

[0016] In a fourth aspect, the present invention provides a method for producing a method of manufacturing a semiconductor device comprising: a second image acquisition module for acquiring a flame radiation image of the object to be measured; and a temperature measurement module for inputting a flame radiation image of a measurement target into a pulverized coal flame temperature measurement model and outputting a temperature distribution image corresponding to the flame radiation image of the measurement target, wherein the pulverized coal flame temperature measurement model is obtained by training using the pulverized coal flame temperature measurement model training device of the third aspect.

[0017] In a fifth aspect, the present invention provides a computer device including a memory and a processor, the memory and the processor being communicatively connected to each other, computer instructions stored in the memory, the processor executing the computer instructions to perform the method for training a pulverized coal flame temperature measurement model of the first aspect or any one of the corresponding embodiments, or to perform the pulverized coal flame temperature measurement method of the second aspect.

[0018] In a sixth aspect, the present invention provides a computer-readable storage medium having stored thereon computer instructions for causing a computer to execute the method for training a pulverized coal flame temperature measurement model according to the first aspect or any one of the corresponding embodiments, or the method for measuring pulverized coal flame temperature according to the second aspect.

[0019] In a seventh aspect, the present invention provides a computer program product comprising computer instructions for causing a computer to execute the method for training a pulverized coal flame temperature measurement model according to the first aspect or any one of the corresponding embodiments, or the method for measuring pulverized coal flame temperature according to the second aspect. [Brief explanation of the drawings]

[0020] In order to more clearly describe the specific embodiments of the present invention or the technical solutions in the prior art, the following will briefly describe the drawings used in the description of the specific embodiments or the prior art. Of course, the drawings described below are part of the embodiments of the present invention, and those skilled in the art can further conceive of other drawings based on these drawings without any creative efforts. [Figure 1] 1 is a flowchart of a method for training a pulverized coal flame temperature measurement model according to an embodiment of the present invention. [Figure 2] FIG. 1 is a schematic diagram illustrating the configuration of one pulverized coal flame temperature measurement model according to an embodiment of the present invention. [Figure 3] 1 is a flowchart of one pulverized coal flame temperature measurement model training according to an embodiment of the present invention. [Figure 4] 1 is a flowchart of one pulverized coal flame temperature image real-time measurement algorithm according to an embodiment of the present invention. [Figure 5] 10 is a flowchart of another method for training a pulverized coal flame temperature measurement model according to an embodiment of the present invention. [Figure 6] 1 is a flowchart of a pulverized coal flame temperature measurement method according to an embodiment of the present invention. [Figure 7] 1 is a block diagram of a pulverized coal flame temperature measurement model training device according to an embodiment of the present invention. FIG. [Figure 8] 1 is a block diagram of a pulverized coal flame temperature measuring device according to an embodiment of the present invention. [Figure 9] FIG. 1 is a schematic diagram illustrating a hardware configuration of a computer device according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0021] In order to make the objectives, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be described clearly and completely below with reference to the drawings in the embodiments of the present invention, but it should be understood that the described embodiments are only a part of the embodiments of the present invention, and are not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without any creative efforts fall within the protection scope of the present invention.

[0022] In related technologies, due to the complexity and unwell-posedness of the problem, it is almost impossible to obtain a direct analytical solution using the first temperature of a radiographic image. Furthermore, due to the radiation characteristics of the flame itself, it is quite difficult to obtain a numerical solution. First, the sensitivity of the device to flame radiation signals and the sensitivity of temperature to flame radiation intensity are high, so even a slight error or perturbation can significantly interfere with the measurement results. Second, the optical parameters, temperature, and radiation intensity are mutually influencing each other, and the three jointly cause nonlinearity in the problem to be solved. Finally, due to the high resolution of the device itself, the number of equations in the problem to be solved is very large, and the coefficient matrix is ​​often a sparse matrix, which results in high computational costs and long operating times.

[0023] To solve the above problems, an embodiment of the present invention provides a method for training a pulverized coal flame temperature measurement model for use in a computer device. It should be noted that the execution entity may be a pulverized coal flame temperature measurement model training device, which may be implemented as part or all of a computer device in the form of software, hardware, or a combination of software and hardware. Here, the computer device may be a terminal, a client terminal, or a server. The server may be a single server or a server cluster consisting of multiple servers. The terminal in the embodiments of the present application may be other smart hardware devices such as a smartphone, a personal computer, or a tablet computer. In the following method embodiments, the execution entity is described as a computer device.

[0024] The computer device in this embodiment is applied to the use scenario of monitoring the combustion stability of pulverized coal flames. The training method for a pulverized coal flame temperature measurement model provided by the present invention involves preprocessing flame radiation images of high-temperature flames, constructing flame radiation images and temperatures corresponding to the flame radiation images as a training set, and training a neural network model using the training set to obtain a pulverized coal flame temperature measurement model, which realizes real-time measurement of pulverized coal flame temperature images, significantly reducing real-time measurement time, ensuring real-time measurement of flame temperature images, and achieving high measurement accuracy, which contributes to understanding the temperature distribution pattern of combustion flames, and exploring physical action processes and chemical reaction mechanisms to evaluate combustion efficiency and provide valuable data support for burner evaluation and improvement.

[0025] According to an embodiment of the present invention, an embodiment of a method for training a pulverized coal flame thermometer model is provided. It should be noted that the steps shown in the flowcharts of the figures may be performed by a computer system as a set of computer-executable instructions, and that although a logical order is shown in the flowcharts, in some cases the steps may be performed in a different order than that shown or described herein.

[0026] This embodiment provides a training method for a pulverized coal flame temperature measurement model that can be used in the above-mentioned computer device. As shown in Figure 1, which is a flowchart of the training method for a pulverized coal flame temperature measurement model according to the embodiment of the present invention, the flow includes the following steps S101 to S104.

[0027] In step S101, a sample flame radiation image is acquired.

[0028] In one example, a flame radiation image for a high temperature sample can be captured by a CCD camera.

[0029] In step S102, the sample flame radiation image is pre-processed to obtain the true temperature corresponding to the sample flame radiation image, and the sample flame radiation image and the temperature corresponding to the sample flame radiation image are used as samples to construct a sample set.

[0030] In an embodiment of the present invention, the sample set includes a training set, which includes training flame radiation images and true training temperatures corresponding to the training flame radiation images.

[0031] In one example, the captured flame radiation image is converted into a three-dimensional temperature distribution to obtain the true temperature corresponding to the sample flame radiation image. The flame radiation image and the corresponding true temperature distribution are combined into one sample. The big data sample is constructed as a sample set by changing the temperature function T(x, y, z). The sample set is divided into a training set, a validation set, and a test set in a ratio of 3:1:1, where x, y, z are the three-dimensional coordinates obtained by dividing the grid according to the three-dimensional distribution of the flame radiation image. By constructing the three-dimensional coordinates, the specific position of the grid can be determined, and the temperature distribution function can be calculated from the temperature corresponding to the grid, and a sample set can be further constructed.

[0032] In step S103, the training flame radiation image is input to a pre-constructed initial pulverized coal flame temperature measurement model, and a first temperature corresponding to the training flame radiation image is output.

[0033] In one example, the pulverized coal flame temperature measurement model builds a CNN model framework based on the main framework of AlexNet. As shown in Figure 2, which is a schematic diagram of the configuration of a pulverized coal flame temperature measurement model according to an embodiment of the present invention, the pulverized coal flame temperature measurement model network has a total of 11 layers, including an input layer, a hidden layer, and an output layer, of which the hidden layer is composed of five convolutional layers, three pooling layers, and one fully connected layer. Table 1 is a list of parameters for the CNN network model, and the pulverized coal flame temperature measurement model parameters are as shown in Table 1.

[0034] TIFF2025165849000007.tif203170

[0035] The convolutional layer is the core module of a convolutional neural network, whose main function is to extract features from input data. The first convolutional layer typically extracts low-level features such as edges, lines, and corners, while the deeper layers iteratively extract more complex features from the low-level features. The pooling layer primarily uses a single global feature value instead of the elemental feature distribution within a specific region, further reducing the amount of data and making the extracted image features more clear, thereby improving the robustness of the network, reducing overfitting, and maintaining linear time invariance. The fully connected layer performs a one-dimensional unfolding of the image features extracted through convolution and pooling, and maps and compares them with the output layer data.

[0036] In step S104, the error between the first temperature and the true training temperature is calculated, and the parameters of the initial pulverized coal flame temperature measurement model are updated by the error until the error converges, thereby obtaining a target pulverized coal flame temperature measurement model.

[0037] 3, which is a flowchart of a pulverized coal flame temperature measurement model training according to an embodiment of the present invention, the pulverized coal flame temperature measurement model training may include determining a maximum number of training epochs I, a number of patience epochs p, and a learning rate, randomly initializing the initial pulverized coal flame temperature measurement model, inputting a training set into the initial pulverized coal flame temperature measurement model, outputting a first temperature, comparing the first temperature with the true training temperature to calculate an error, and substituting the error into the model to update the model parameters, which is one iteration of the training process. When the iterations are repeated until the error no longer decreases or reaches a convergence standard, the iterations are stopped and the model training is completed.

[0038] In one implementation scenario, as shown in FIG. 4, which is a flowchart of an algorithm for real-time measurement of pulverized coal flame temperature images according to an embodiment of the present invention, the algorithm for real-time measurement of pulverized coal flame temperature images includes the steps of acquiring a dataset of flame radiation images and temperature distribution; dividing the dataset into a training set, a test set, and a validation set; constructing a CNN network model, which is a model for real-time measurement of pulverized coal flame temperature images; a CNN model training step, which includes inputting the training set into the network, comparing the output data with the actual temperature to obtain an error, and substituting the error into the model to update the model parameters, completing one iteration. When the error no longer decreases or reaches a convergence standard, the iteration is stopped and model training is completed; and a CNN model testing step, which includes inputting the test set into the network, comparing the output data with the actual temperature to obtain an error, and if the error is large, re-adjusting the model parameters and training again; and when the error is within a reasonable range, completing model construction.

[0039] In the training method for a pulverized coal flame temperature measurement model provided in this embodiment, flame radiation images of high-temperature flames are preprocessed, and the flame radiation images and the temperatures corresponding to the flame radiation images are constructed as a training set. A neural network model is trained using the training set to obtain a pulverized coal flame temperature measurement model, thereby realizing real-time measurement of pulverized coal flame temperature images, which can significantly reduce the real-time measurement time and ensure real-time measurement of flame temperature images with high measurement accuracy. This contributes to understanding the temperature distribution pattern of combustion flames, and by exploring the physical action process and chemical reaction mechanism, it can evaluate combustion efficiency and provide effective data support for burner evaluation and improvement.

[0040] This embodiment provides a training method for a pulverized coal flame temperature measurement model that can be used in the above-mentioned computer equipment, etc. As shown in Fig. 5, which is a flowchart of another training method for a pulverized coal flame temperature measurement model according to an embodiment of the present invention, the flow includes the following steps S501 to S507.

[0041] In step S501, a sample flame radiation image is acquired. For details, refer to step S101 in the embodiment shown in Figure 1, and detailed description thereof will be omitted here.

[0042] In step S502, the sample flame radiation image is pre-processed to obtain the true temperature corresponding to the sample flame radiation image, and the sample flame radiation image and the temperature corresponding to the sample flame radiation image are taken as samples to construct a sample set.

[0043] Specifically, the above step S502 includes step S5021.

[0044] In step S5021, the furnace radiative transfer equation is solved for the sample flame radiation image to obtain the true temperature corresponding to the sample flame radiation image.

[0045] In one example, the step is detailed as follows.

[0046] In this manner, by solving the furnace radiative transfer equation and determining the true temperatures corresponding to the flame radiation images, subsequent images and the true temperatures corresponding to the images can be used to facilitate training of a pulverized coal flame temperature measurement model.

[0047] In step S503, the training flame radiation image is input into a pre-constructed initial pulverized coal flame temperature measurement model, and a first temperature corresponding to the training flame radiation image is output. For details, see step S103 in the embodiment shown in Figure 1, and detailed description will be omitted here.

[0048] In step S504, the error between the first temperature and the true training temperature is calculated, and the parameters of the initial pulverized coal flame temperature measurement model are updated by the error until the error converges, thereby obtaining a target pulverized coal flame temperature measurement model.

[0049] In step S5041, the formula Calculate the error between the first temperature and the temperature using TIFF2025165849000008.tif11170.

[0050] In step S5042, the formula Using TIFF2025165849000009.tif14170, the parameters of the initial pulverized coal flame temperature measurement model are updated until the error converges, and the target pulverized coal flame temperature measurement model is obtained.

[0051] In one example, a convolutional neural network is used as the core of the pulverized coal flame temperature measurement model, and the purpose of training the convolutional neural network is to reduce the loss function, and the smaller the loss function, the closer the predicted value is to the true value. In the iteration of the pulverized coal flame temperature measurement model, the mean square error is used as the loss function L. Using the following calculation formula, TIFF2025165849000010.tif11170

[0052] Calculate the error between the first temperature and the temperature, where N is the total number of grids in the training flame radiometric image, and T m is the first temperature, T r is the true training temperature and L is the error.

[0053] Activation functions exist between the outputs of the upper nodes and the inputs of the lower nodes in a convolutional neural network, and their main role is to provide the network with nonlinear modeling capabilities. The outputs of each convolutional layer are activated by a leaky Relu function, which realizes this nonlinearity. To prevent overfitting and improve the model's generalization ability and robustness, a drop-out layer is added to randomly remove some neurons and convolutional kernel parameters during each training epoch, thereby reducing coadaptation between neurons. A BN layer is added after each convolutional layer to standardize the data, which helps improve the noise resistance of the convolutional neural network algorithm.

[0054] Specifically, the pulverized coal flame temperature measurement model is trained using gradient descent, which is an algorithm that updates parameters using backpropagation. The loss function is minimized during the training process, and the weight parameters are iteratively updated by finding the minimum value of the loss function in the convolutional network model. The formula for gradient descent is as follows: TIFF2025165849000011.tif14170

[0055] During the ceremony, TIFF2025165849000012.tif5170 is the weight of the previous iteration, TIFF2025165849000013.tif8170 is the updated weight, TIFF2025165849000014.tif3170 is the learning rate, L is the error, k, j represent the kth layer network and jth neurons of the initial pulverized coal flame temperature measurement model, h is the activation function, and w and b are the weight and offset of the initial pulverized coal flame temperature measurement model.

[0056] In the weight update method, the gradient descent method can obtain the relationship between the optimal solution and the learning rate. If the learning rate is too large, the optimal solution may be skipped. If the learning rate is too small, it may take a long time to converge. Therefore, the optimization algorithm is mainly based on the weight variable items in the above formula from different aspects. TIFF2025165849000015.tif5170 is processed, and a function that changes with the loss function is added to the learning rate. The Adam algorithm uses first- and second-moment estimation to calculate gradients and designs independent adaptive learning rates for different parameters. The Adam algorithm has the smallest average relative error, a relatively short training time, and the best effect. Therefore, the Adam algorithm is selected as the network parameter optimization algorithm.

[0057] In the training process of the pulverized coal flame temperature measurement model, inputting training samples into the network in one batch requires high hardware computing performance, and because it is a global loss function, the reconstruction effect of extreme value data of some samples is likely to deteriorate. Based on this, the batch gradient descent method is selected, and all samples are divided into limited batches and input into the pulverized coal flame temperature measurement model in order to complete one epoch of training. In the above method, the selection of the batch size is very important, as it has a large impact on training, and generally TIFF2025165849000016.tif4170. In the present invention, the batch setting is not limited. Therefore, all samples are divided into limited batches using batch gradient descent, and image data is sequentially input to the input layer. The output layer obtains output data through the operation of the convolutional neural network model. The output data is compared with the three-dimensional temperature of the flame to obtain an error, and the error is substituted into the model to update the model parameters. This is one iteration of the training process. When the error no longer decreases or reaches a convergence standard, the iteration is stopped and the model training is completed.

[0058] In an embodiment of the present invention, the sample set further includes a test set, which includes test flame radiographic images and true test temperatures corresponding to the test flame radiographic images.

[0059] In step S505, the test flame radiation image is input into a target pulverized coal flame temperature measurement model to obtain a second temperature corresponding to the test flame radiation image.

[0060] In step S506, a second error between the second temperature and the true test temperature is calculated.

[0061] In step S507, if the second error is greater than or equal to the preset threshold, the parameters of the target pulverized coal flame temperature measurement model are updated according to the second error, and the next iteration training is performed, and the iteration is stopped until the second error between the second temperature and the true test temperature becomes less than the preset threshold.

[0062] In one example, the test set is input to the model, a second temperature is output, and compared with the true temperature in the test set to obtain a second error. If the second error is large, the model parameters are re-adjusted and re-trained. When the second error is within a reasonable range, the model construction is complete.

[0063] In this model, the test set image data is input into the model, and the second temperature is output and compared with the true temperature to obtain the error. If the error is large, the model parameters are re-adjusted and re-trained. When the error is within a reasonable range, the model construction is complete, thereby further improving the measurement accuracy of the model for real-time flame temperature image measurement.

[0064] The training method for a pulverized coal flame temperature measurement model provided in this embodiment solves the in-furnace radiative transfer equation to determine the true temperature corresponding to the flame radiation image, thereby facilitating the training of the pulverized coal flame temperature measurement model using subsequent images and the true temperature corresponding to the images. By training an initial pulverized coal flame temperature measurement model, the initial pulverized coal flame temperature measurement model can find the input-output relationship from a priori knowledge, achieving a direct solution without the need to actually model the flame physicochemical rules of the combustion process. This has the advantages of very strong large-scale data processing capabilities, high efficiency, high robustness, and high adaptability under different conditions. Test set image data is input into the model, and a second temperature is output, which is compared with the true temperature to obtain an error. If the error is large, the model parameters are re-adjusted and re-trained. When the error is within a reasonable range, the model construction is completed, thereby further improving the measurement accuracy of the model for real-time flame temperature image measurement.

[0065] This embodiment provides a pulverized coal flame temperature measurement method that can be used in the above computer equipment. As shown in Figure 6, which is a flowchart of a pulverized coal flame temperature measurement method according to an embodiment of the present invention, the flow includes the following steps S601 and S602.

[0066] In step S601, a flame radiation image of the measurement target is acquired.

[0067] In step S602, the flame radiation image of the measurement target is input to a pulverized coal flame temperature measurement model, and a temperature distribution image corresponding to the flame radiation image of the measurement target is output.

[0068] In the embodiment of the present invention, the pulverized coal flame temperature measurement model is obtained by training using the above-mentioned training method for a pulverized coal flame temperature measurement model.

[0069] The pulverized coal flame temperature measurement method provided in this embodiment uses a trained pulverized coal flame temperature measurement model to realize real-time measurement of pulverized coal flame temperature images, which can significantly reduce the real-time measurement time while ensuring real-time measurement of flame temperature images and has high measurement accuracy, contributing to understanding the temperature distribution pattern of combustion flames, and exploring the physical action process and chemical reaction mechanism to evaluate combustion efficiency and even evaluate and improve burners.

[0070] This embodiment also provides a pulverized coal flame temperature measurement model training device. This device is used to realize the above-mentioned embodiments and preferred embodiments, and the description thereof will be omitted. As used below, the term "module" can realize a combination of software and / or hardware for a predetermined function. The device described in the following embodiment is preferably realized by software, but it is also possible and conceivable to realize it by hardware or a combination of software and hardware.

[0071] This embodiment provides a pulverized coal flame temperature measurement model training device, which includes a first image acquisition module 701, a data pre-processing module 702, a first temperature output module 703, and a model training module 704, as shown in FIG.

[0072] The first image acquisition module 701 is used to acquire a sample flame radiation image, for details see step S101 of the embodiment shown in Figure 1, and detailed description will be omitted here.

[0073] The data preprocessing module 702 preprocesses the sample flame radiation image to obtain the true temperature corresponding to the sample flame radiation image, and uses the sample flame radiation image and the temperature corresponding to the sample flame radiation image as samples to construct a sample set, where the sample set includes a training set, which includes the training flame radiation image and the true training temperature corresponding to the training flame radiation image. For details, see step S102 in the embodiment shown in Figure 1, and detailed description will be omitted here.

[0074] The first temperature output module 703 is used to input the training flame radiation image into the pre-constructed initial pulverized coal flame temperature measurement model, and output the first temperature corresponding to the training flame radiation image. For details, refer to step S103 in the embodiment shown in Figure 1, and detailed description will be omitted here.

[0075] The model training module 704 calculates the error between the first temperature and the true training temperature, and updates the parameters of the initial pulverized coal flame temperature measurement model according to the error until the error converges, and is used to obtain a target pulverized coal flame temperature measurement model, which is used to measure the pulverized coal flame temperature. For details, see step S104 in the embodiment shown in Figure 1, and detailed description will be omitted here.

[0076] In some alternative embodiments, the data pre-processing module 702: An in-furnace radiative transfer equation unit is included for solving the in-furnace radiative transfer equation for the sample flame radiation image to obtain the true temperature corresponding to the sample flame radiation image.

[0077] In some alternative embodiments, the model training module 704 includes an error calculation unit and a first model parameter update unit.

[0078] The error calculation unit is TIFF2025165849000017.tif11170 is used to calculate the error between the first temperature and the temperature, where N is the total number of grids in the training flame radiometric image, and T m is the first temperature, T r is the true training temperature and L is the error.

[0079] The first model parameter update unit is TIFF2025165849000018.tif14170, the parameters of the initial pulverized coal flame temperature measurement model are updated until the error converges, and the target pulverized coal flame temperature measurement model is obtained, where: TIFF2025165849000019.tif5170 is the weight of the previous iteration, TIFF2025165849000020.tif8170 is the updated weight, TIFF2025165849000021.tif3170 is the learning rate, L is the error, k, j represent the kth layer network and jth neurons of the initial pulverized coal flame temperature measurement model, h is the activation function, and w and b are the weight and offset of the initial pulverized coal flame temperature measurement model.

[0080] In some optional embodiments, the sample set further includes a test set, the test set includes a test flame radiation image and a true test temperature corresponding to the test flame radiation image, and the pulverized coal flame temperature measurement model training device further includes a second temperature output unit, a second error calculation unit, and a second model parameter update unit.

[0081] The second temperature output unit is used to input the test flame radiographic image into the target pulverized coal flame temperature measurement model to obtain a second temperature corresponding to the test flame radiographic image.

[0082] The second error calculation unit is used to calculate a second error between the second temperature and the true test temperature.

[0083] The second model parameter updating unit is used for updating the parameters of the target pulverized coal flame temperature measurement model according to the second error when the second error is equal to or greater than a preset threshold, and performing the next iteration training, and stopping the iteration until the second error between the second temperature and the true test temperature is less than a preset threshold, the preset threshold being for evaluating the pulverized coal flame temperature measurement model.

[0084] The further functions of each module and unit described above are the same as those of the corresponding embodiment described above, and detailed descriptions thereof will be omitted here.

[0085] The pulverized coal flame temperature measurement model training device in this embodiment is expressed in the form of functional units, where a unit is an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory that executes one or more software or specific programs, and / or other devices that can provide the above functions.

[0086] This embodiment also provides a pulverized coal flame temperature measurement model training device, which is used to realize the above-mentioned embodiments and preferred embodiments, and the description thereof will be omitted. As used below, the term "module" can realize a combination of software and / or hardware for a predetermined function. The device described in the following embodiment is preferably realized by software, but it is also possible and conceivable to realize it by hardware or a combination of software and hardware.

[0087] This embodiment provides a pulverized coal flame temperature measurement device, which includes a second image acquisition module 801 and a temperature measurement module 802, as shown in FIG.

[0088] The second image acquisition module 801 is used to acquire a flame radiation image of the measurement target, for details see step S601 in the embodiment shown in Figure 6, and detailed description will be omitted here.

[0089] The temperature measurement module 802 is used to input a flame radiation image of a measurement target into a pulverized coal flame temperature measurement model and output a temperature distribution image corresponding to the flame radiation image of the measurement target, where the pulverized coal flame temperature measurement model is obtained by training using the pulverized coal flame temperature measurement model training device of the third aspect. For details, see step S602 of the embodiment shown in Figure 6, and detailed description will be omitted here.

[0090] The further functions of each module and unit described above are the same as those of the corresponding embodiment described above, and detailed descriptions thereof will be omitted here.

[0091] The pulverized coal flame temperature measurement device in this embodiment is expressed in the form of a functional unit, where a unit is an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory that executes one or more software or specific programs, and / or other devices that can provide the above functions.

[0092] The embodiment of the present invention further provides a computer device including the pulverized coal flame temperature measurement model training device shown in FIG. 7 and the pulverized coal flame temperature measurement device shown in FIG.

[0093] Referring to FIG. 9, which is a schematic diagram of a computer device provided according to an alternative embodiment of the present invention, the computer device includes one or more processors 10, memory 20, and interfaces for connecting components, including high-speed and low-speed interfaces. The components are communicatively connected to each other via different buses and may be mounted on a common mainboard or in other configurations as needed. The processors can process instructions executed within the computer device, including instructions stored in or on memory for displaying graphical user interface (GUI) information on an external input / output device (e.g., a display device coupled to the interface). In some alternative embodiments, multiple processors and / or multiple buses can be used along with multiple memories as needed. Similarly, multiple computer devices may be connected, each providing a portion of the required operations (e.g., a server array, a set of blade servers, or a multiprocessor system). FIG. 9 illustrates a single processor 10 as an example.

[0094] The processor 10 may be a central processing unit, a network processor, or a combination thereof. Here, the processor 10 may further include a hardware chip. The hardware chip may be an application specific integrated circuit, a programmable logic device, or a combination thereof. The programmable logic device may be a complex programmable logic device, a field programmable gate array, a general-purpose array logic, or any combination thereof.

[0095] Here, the memory 20 stores instructions executable by at least one processor 10, thereby causing the at least one processor 10 to execute and realize the methods shown in the above embodiments.

[0096] The memory 20 may include a program storage area capable of storing an operating system and / or applications required for at least one function, and a data storage area capable of storing data generated by use of the computer device. The memory 20 may also include high-speed random access memory or non-transitory memory such as at least one magnetic disk storage device, flash memory device, or other non-transitory solid-state storage device. In some alternative embodiments, the memory 20 may optionally include memory located remotely from the processor 10, and these remote memories may be connected to the computer device via a network. Examples of such networks include, but are not limited to, the Internet, an internal company network, a local area network, a mobile communication network, and combinations thereof.

[0097] Memory 20 may include volatile memory, such as random access memory, non-volatile memory, such as flash memory, a hard disk or solid state disk, or a combination of the above types of memory.

[0098] The computer device further includes an input device 30 and an output device 40. The processor 10, the memory 20, the input device 30 and the output device 40 can be connected via a bus or other methods, and FIG. 9 illustrates the connection via a bus.

[0099] The input device 30 can receive input numeric or character information and generate key signal inputs related to user settings and function control of the computing device, and can be, for example, a touch panel, a keypad, a mouse, a trackpad, a pointing lever, one or more mouse buttons, a trackball, a joystick, etc. The output device 40 can include a display device, an auxiliary lighting device (e.g., LED), a tactile feedback device (e.g., vibration motor), etc. The display device can include, but is not limited to, a liquid crystal display, a light-emitting diode display, and a plasma display. In some optional embodiments, the display device can be a touch panel.

[0100] The embodiments of the present invention further provide a computer-readable storage medium. The methods according to the above-described embodiments of the present invention can be implemented in hardware, firmware, or by computer code that can be recorded on a storage medium, or that is initially stored on a remote storage medium or a non-transitory machine-readable storage medium downloaded from a network and then stored on a local storage medium, such that the methods described herein can be processed by software, such as that stored on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. Here, the storage medium may be a magnetic disk, optical disk, read-only memory, random-access memory, flash memory, hard disk, solid-state disk, etc., and may further include a combination of the above types of memory. It should be understood that a computer, processor, microprocessor controller, or programmable hardware includes a storage component that can store or receive software or computer code, and that the software or computer code, when accessed and executed by the computer, processor, or hardware, implements the methods illustrated in the above-described embodiments.

[0101] Some aspects of the present invention may be implemented as a computer program product, such as computer program instructions, which, when executed by a computer, can invoke or provide the methods and / or technical solutions of the present invention through the operation of the computer. As those skilled in the art will appreciate, computer program instructions may exist in a computer-readable medium, including, but not limited to, a source file, an executable file, an installation package file, etc. Accordingly, computer program instructions may be executed by a computer in a manner that includes, but is not limited to, the computer directly executing the instructions, compiling the instructions and then executing a corresponding compiled program, reading and executing the instructions, or reading and installing the instructions and then executing a corresponding installed program. The computer-readable medium may be any available computer-readable storage medium or communication medium accessible by a computer.

[0102] Although the present invention will be described with reference to the drawings, those skilled in the art will be able to make various modifications and changes without departing from the spirit and scope of the present invention, and all such modifications and changes are included within the scope defined by the appended claims.

Claims

1. acquiring a sample flame radiographic image; pre-processing the sample flame radiation images to obtain true temperatures corresponding to the sample flame radiation images, and constructing a sample set using the sample flame radiation images and the temperatures corresponding to the sample flame radiation images as samples, the sample set including a training set, the training set including training flame radiation images and true training temperatures corresponding to the training flame radiation images; inputting the training flame radiation image into a pre-constructed initial pulverized coal flame temperature measurement model, and outputting a first temperature corresponding to the training flame radiation image; a step of calculating an error between the first temperature and the true training temperature, and updating parameters of the initial pulverized coal flame temperature measurement model using the error until the error converges, to obtain a target pulverized coal flame temperature measurement model, wherein the target pulverized coal flame temperature measurement model is used to measure the pulverized coal flame temperature.

2. The step of pre-processing the sample flame radiation image to obtain a true temperature corresponding to the sample flame radiation image, as described above, comprises:

2. The method of claim 1, further comprising solving a furnace radiative transfer equation for the sample flame radiographic image to obtain a true temperature corresponding to the sample flame radiographic image.

3. The step of calculating an error between the first temperature and the true training temperature, updating parameters of the initial pulverized coal flame temperature measurement model by the error until the error converges, and obtaining a target pulverized coal flame temperature measurement model, official where N is the total number of grids in the training flame radiation image, and T ｍ is the first temperature, T ｒ is the true training temperature and L is the error; official A step of updating parameters of the initial pulverized coal flame temperature measurement model until the error converges, and obtaining a target pulverized coal flame temperature measurement model, wherein: is the weight of the previous iteration, are the updated weights, is a learning rate, L is the error, k, j represent the j-th neuron in the k-th layer network of the initial pulverized coal flame temperature measurement model, h is an activation function, and w and b are a weight and an offset of the initial pulverized coal flame temperature measurement model.

4. A method wherein the sample set further comprises a test set, the test set comprising test flame radiographic images and true test temperatures corresponding to the test flame radiographic images, inputting the test flame radiation image into the target pulverized coal flame temperature measurement model to obtain a second temperature corresponding to the test flame radiation image; calculating a second error between the second temperature and the true test temperature; 2. The method of claim 1, further comprising: if the second error is equal to or greater than a preset threshold, updating parameters of the target pulverized coal flame temperature measurement model with the second error, performing a next iteration of training, and stopping the iteration until the second error between the second temperature and the true test temperature is less than the preset threshold, wherein the preset threshold is for evaluating the pulverized coal flame temperature measurement model.

5. acquiring a flame radiation image of the object to be measured; a step of inputting the flame radiation image of the measurement target into a pulverized coal flame temperature measurement model and calculating the temperature of the flame radiation image of the measurement target, wherein the pulverized coal flame temperature measurement model is obtained by training using the training method for a pulverized coal flame temperature measurement model according to any one of claims 1 to 4.

6. a first image acquisition module for acquiring a sample flame radiographic image; a data pre-processing module for pre-processing the sample flame radiation images to obtain true temperatures corresponding to the sample flame radiation images, and constructing a sample set using the sample flame radiation images and the temperatures corresponding to the sample flame radiation images as samples, the sample set including a training set, the training set including training flame radiation images and true training temperatures corresponding to the training flame radiation images; a first temperature output module for inputting the training flame radiation image into a pre-constructed initial pulverized coal flame temperature measurement model and outputting a first temperature corresponding to the training flame radiation image; a model training module for calculating an error between the first temperature and the true training temperature, and updating parameters of the initial pulverized coal flame temperature measurement model using the error until the error converges, to obtain a target pulverized coal flame temperature measurement model, wherein the target pulverized coal flame temperature measurement model is used to measure the pulverized coal flame temperature.

7. a second image acquisition module for acquiring a flame radiation image of the object to be measured; a temperature measurement module for inputting the flame radiation image of the measurement target into a pulverized coal flame temperature measurement model and outputting a temperature distribution image corresponding to the flame radiation image of the measurement target, wherein the pulverized coal flame temperature measurement model is obtained by training using the pulverized coal flame temperature measurement model training device described in claim 6.

8. A computer device comprising: a memory and a processor, the memory and the processor being communicatively connected to each other; computer instructions stored in the memory; and the processor executing the computer instructions to perform the method for training a pulverized coal flame temperature measurement model according to any one of claims 1 to 4, or the method for measuring pulverized coal flame temperature according to claim 5.

9. A computer-readable storage medium having stored thereon computer instructions for causing a computer to execute the method for training a pulverized coal flame temperature measurement model according to any one of claims 1 to 4, or the method for measuring pulverized coal flame temperature according to claim 5.

10. A computer program product comprising computer instructions for causing a computer to execute the method for training a pulverized coal flame temperature measurement model according to any one of claims 1 to 4, or the method for measuring pulverized coal flame temperature according to claim 5.

Images