Sintering furnace temperature control method and system based on BP neural network prediction model

Abstract

This invention provides a sintering furnace temperature control method and system based on a BP neural network prediction model. The method includes: constructing an initial model based on an Adaboost BP neural network; training a prediction model using experimental data and the initial model; the experimental data includes the sintering furnace temperature and power of the target sintering; and predicting and adjusting the required heating power of the sintering furnace in real time based on the prediction model and the sintering temperature curve of the target sintering. This invention's sintering furnace temperature control method and system, based on a BP neural network prediction model, uses experimental data to train an Adaboost-based BP neural network to obtain a prediction model, predicts and adjusts the required heating power of the sintering furnace in real time based on the sintering temperature curve of the target sintering, resulting in a more timely and linear heating process with stronger anti-interference capabilities. Furthermore, it improves the stability of sintered product output and reduces the product defect rate.

Description

Technical Field

[0001] This invention relates to the field of intelligent control technology based on neural networks, and in particular to a sintering furnace temperature control method and system based on a BP neural network prediction model. Background Technology

[0002] Sintering is the process of heating metal powder or powder compacts to below their melting point, causing the particles to bond together to obtain a material with the required strength. A sintering furnace is the device that performs this process. During the sintering of powder compacts, the heating and cooling rates need to be precisely controlled to allow the powder to undergo dewaxing, reduction, alloying, and microstructural transformation. Therefore, temperature control during the sintering process is extremely important.

[0003] However, the traditional sintering furnace heating process is characterized by time delay, nonlinearity and strong disturbance, which leads to unstable output of sintered products and a high defect rate.

[0004] In view of this, there is an urgent need for a sintering furnace temperature control method and system based on a BP neural network prediction model, in order to at least solve the above-mentioned shortcomings. Summary of the Invention

[0005] One of the objectives of this invention is to provide a sintering furnace temperature control method and system based on a BP neural network prediction model. The method uses experimental data to train an Adaboost-based BP neural network to obtain a prediction model. Based on the sintering temperature curve of the desired sintering target, the method predicts the heating power required by the sintering furnace and adjusts it in real time. The heating process of the sintering furnace is more timely and linear, and the anti-interference ability is stronger. Furthermore, it improves the output stability of sintered products and reduces the product defect rate.

[0006] The sintering furnace temperature control method based on a BP neural network prediction model provided in this embodiment of the invention includes:

[0007] Step 1: Construct an initial model based on an Adaboost-based BP neural network, and train a prediction model using experimental data and the initial model; the experimental data includes: the sintering furnace temperature and power of the experimental sintering target;

[0008] Step 2: Based on the prediction model and the sintering temperature curve of the desired sintering target, predict the heating power required by the sintering furnace and adjust it in real time.

[0009] Preferably, the method for measuring the temperature of the sintering furnace includes:

[0010] K-type thermocouples were used to measure the temperature distribution at various points in the space inside the sintering furnace over time.

[0011] The sintering furnace temperature control method based on a BP neural network prediction model provided in this embodiment of the invention further includes:

[0012] Before predicting the heating power required by the sintering furnace based on the prediction model and the sintering temperature curve of the desired sintering target and adjusting it in real time, it is necessary to determine whether the temperature distribution of the sintering furnace is uniform.

[0013] If the temperature distribution in the sintering furnace is uneven, a pre-constructed physical model of the sintering furnace is obtained based on the energy conservation, mass conservation, momentum conservation equations and the structural size parameters of the sintering furnace.

[0014] Based on the power of the experimental sintering target and the pre-constructed physical model of the sintering furnace, the simulation results are determined;

[0015] By comparing and analyzing the simulation results and the sintering furnace temperature of the experimental sintering target, and correcting the pre-built sintering furnace physical model, the first sintering furnace physical model is obtained.

[0016] Based on the prediction model, the physical model of the first sintering furnace, and the sintering temperature curve of the desired sintering target, the heating power required by the sintering furnace is predicted and adjusted in real time.

[0017] Preferably, the sintering furnace temperature is compared and analyzed with the simulation results and the experimental sintering target, and the pre-built sintering furnace physical model is corrected to obtain the first sintering furnace physical model, including:

[0018] A pre-built physical model of a sintering furnace is loaded into the virtual furnace body. Based on the comparison and analysis results of the simulation results and the sintering furnace temperature of the experimental sintering target, the differential temperature points in the virtual furnace body are determined.

[0019] The system displays known influencing factors of temperature difference points at preset viewing locations.

[0020] The target knowledge structure is indexed based on known influencing factors; the target knowledge structure includes the first unknown influencing factor.

[0021] Extract knowledge structure information based on the target knowledge structure, input the knowledge structure information into the error analysis model trained on the structured storage knowledge based on the target knowledge structure, and obtain the error analysis results.

[0022] Based on the semantic description of the error analysis results, the first unknown influencing factor is described in the virtual furnace body, and the engine-driven data of the first unknown influencing factor described by the index is indexed.

[0023] The pre-built sintering furnace physical model was revised based on engine-driven data.

[0024] Preferably, the target knowledge structure is indexed based on known influencing factors, including:

[0025] Based on the first knowledge structure in the big data index of known influencing factors, determine the second knowledge structure viewed by the viewer;

[0026] Based on the second unknown influencing factor in the second knowledge structure, a virtual furnace body is located for viewing by the observer, and the first unknown influencing factor selected by the observer is determined.

[0027] Preferably, based on the prediction model, the physical model of the first sintering furnace, and the sintering temperature curve of the desired sintering target, the heating power required by the sintering furnace is predicted and adjusted in real time, including:

[0028] Based on the prediction model and the physical model of the first sintering furnace, the target correspondence is determined as follows: the correspondence between the temperature distribution of the sintering rack and the predicted power.

[0029] Calculate the deviation between each target temperature in the sintering rack temperature distribution and the required temperature at the current moment;

[0030] The predicted power corresponding to the sintering rack temperature distribution with the smallest deviation is used as the target power for real-time adjustment.

[0031] The sintering furnace temperature control method based on a BP neural network prediction model provided in this embodiment of the invention further includes:

[0032] The target ratio is calculated based on the minimum deviation and the required temperature corresponding to each sintering moment. If the target ratio is greater than the preset ratio threshold and the sintering time period is greater than the preset time period length threshold, the interval distribution of the target sintering moment in the sintering time interval is determined.

[0033] Based on the characteristics of the interval distribution, a deviation correction strategy is determined.

[0034] Preferably, based on the interval distribution characteristics, a deviation correction strategy is determined, including:

[0035] If the interval distribution characteristics meet the first standard distribution characteristics, the heating device operating status database is read according to the interval distribution characteristics to determine the heating device operating status characteristics.

[0036] Based on the characteristics of the heating device's working status and the database of abnormal working status characteristics, identify abnormal heating devices and perform abnormal maintenance.

[0037] Preferably, determining the deviation correction strategy based on the interval distribution characteristics further includes:

[0038] If the interval distribution characteristics conform to the second standard distribution characteristics, determine the design of the arrangement positions of multiple sintering racks in the sintering furnace.

[0039] The physical model of the first sintering furnace was updated based on the arrangement of the sintering racks to obtain the physical model of the second sintering furnace;

[0040] The first temperature field lines at different simulated powers were simulated based on the physical model of the second sintering furnace.

[0041] Based on the sintering temperature curve of the desired sintering target, determine the second temperature field line that the design position passes through at different sintering times;

[0042] Identify and associate the third temperature field line that is most similar to the second temperature field line in the first temperature field line;

[0043] The evaluation value of the sintering rack layout design is determined based on the temperature field line difference between each associated second and third temperature field line.

[0044] The sintering rack layout position designed for the position with the highest evaluation value is used as the new sintering rack layout position.

[0045] The sintering furnace temperature control system based on a BP neural network prediction model provided in this embodiment of the invention includes:

[0046] The training module is used to build an initial model of a BP neural network based on Adaboost, and to train a prediction model using experimental data and the initial model; the experimental data includes: the sintering furnace temperature and power of the experimental sintering target;

[0047] The control module is used to predict the heating power required by the sintering furnace based on the prediction model and the sintering temperature curve of the desired sintering target, and to adjust it in real time.

[0048] The beneficial effects of this invention are as follows:

[0049] This invention utilizes experimental data to train a BP neural network based on Adaboost to obtain a prediction model. Based on the sintering temperature curve of the desired sintering target, it predicts the heating power required by the sintering furnace and adjusts it in real time. The heating process of the sintering furnace is more timely and linear, and its anti-interference ability is stronger. Furthermore, it improves the output stability of sintered products and reduces the product defect rate.

[0050] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in this application.

[0051] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0052] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0053] Figure 1 This is a schematic diagram of a sintering furnace temperature control method based on a BP neural network prediction model in an embodiment of the present invention.

[0054] Figure 2 This is a schematic diagram of a sintering furnace temperature control system based on a BP neural network prediction model in an embodiment of the present invention. Detailed Implementation

[0055] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0056] This invention provides a sintering furnace temperature control method based on a BP neural network prediction model, such as... Figure 1 As shown, it includes:

[0057] Step 1: Construct an initial model based on an Adaboost-based BP neural network, and train a prediction model using experimental data and the initial model; the experimental data includes: the sintering furnace temperature and power of the experimental sintering target;

[0058] The methods for measuring the temperature of the sintering furnace include:

[0059] K-type thermocouples were used to measure the temperature distribution at various points in the space inside the sintering furnace over time.

[0060] Step 2: Based on the prediction model and the sintering temperature curve of the desired sintering target, predict the heating power required by the sintering furnace and adjust it in real time.

[0061] In this embodiment, the test data are historical measurement data of powder (e.g., polytetrafluoroethylene) sintering, including the temperature at various points in the sintering furnace space measured by K-type thermocouples at different power levels during the test.

[0062] In this embodiment, when training the initial model using experimental data, the sintering furnace temperature of the experimental sintering target is used as the input of the initial model, and the power of the experimental sintering target is used as the output of the initial model. When the initial model is trained to convergence, it is used as the prediction model.

[0063] In this embodiment, the desired sintering target is the actual sintering target that needs to be sintered, and the sintering temperature curve of the desired sintering target is the ideal temperature change curve for its entire sintering stage. This curve is determined according to the sintering requirements of the desired sintering target.

[0064] In this embodiment, predicting the heating power required by the sintering furnace based on the prediction model and the sintering temperature curve of the desired sintering target and adjusting it in real time means: reading the ideal sintering temperature in the sintering temperature curve according to the current sintering stage, inputting the ideal sintering temperature into the prediction model, and using the power output by the prediction model as the heating power to be adjusted in real time.

[0065] The working principle and beneficial effects of the above technical solution are as follows:

[0066] This invention utilizes experimental data to train a BP neural network based on Adaboost to obtain a prediction model. Based on the sintering temperature curve of the desired sintering target, it predicts the heating power required by the sintering furnace and adjusts it in real time. The heating process of the sintering furnace is more timely and linear, and its anti-interference ability is stronger. Furthermore, it improves the output stability of sintered products and reduces the product defect rate.

[0067] This invention provides a sintering furnace temperature control method based on a BP neural network prediction model, and further includes:

[0068] Before predicting the heating power required by the sintering furnace based on the prediction model and the sintering temperature curve of the desired sintering target and adjusting it in real time, it is necessary to determine whether the temperature distribution of the sintering furnace is uniform.

[0069] Among them, the uniformity of temperature distribution in the sintering furnace is determined by comparing the temperatures of multiple spatial points under the same power. If the temperature difference between two spatial points is greater than a set value, such as 15℃, the temperature distribution in the sintering furnace is determined to be uneven.

[0070] If the temperature distribution in the sintering furnace is uneven, a pre-constructed physical model of the sintering furnace is obtained based on the energy conservation, mass conservation, momentum conservation equations and the structural size parameters of the sintering furnace.

[0071] Among them, the pre-constructed physical model of the sintering furnace is: a numerical simulation model of the sintering furnace, which simulates the temperature field inside the furnace and predicts the temperature response under different power levels;

[0072] Based on the power of the experimental sintering target and the pre-constructed physical model of the sintering furnace, the simulation results are determined;

[0073] The experimental sintering target is the sintering target in the experimental data; the simulation result is the temperature distribution of the power corresponding to the experimental sintering target, output by the pre-constructed sintering furnace physical model.

[0074] By comparing and analyzing the simulation results and the sintering furnace temperature of the experimental sintering target, and correcting the pre-built sintering furnace physical model, the first sintering furnace physical model is obtained.

[0075] The first sintering furnace physical model is a physical model that can truly reflect the temperature distribution inside the sintering furnace, obtained by adjusting the pre-constructed sintering furnace physical model based on the simulation results and the analysis results of the sintering furnace temperature of the experimental sintering target. During the adjustment, the historical adjustment records of the engineers are studied and corresponding adjustments are made. The adjustment records include adjustment strategies under different simulation results and actual results comparison scenarios.

[0076] Based on the prediction model, the physical model of the first sintering furnace, and the sintering temperature curve of the desired sintering target, the heating power required by the sintering furnace is predicted and adjusted in real time.

[0077] The process of predicting and adjusting the heating power required by the sintering furnace based on the prediction model, the physical model of the first sintering furnace, and the sintering temperature curve of the desired sintering target refers to: reading the ideal sintering temperature from the sintering temperature curve at the current sintering stage, and considering the influencing factor of uneven temperature distribution inside the furnace, and finally using the prediction model to predict the power that can enable the desired sintering target to reach the ideal sintering temperature at the current sintering stage.

[0078] The working principle and beneficial effects of the above technical solution are as follows:

[0079] Due to differences in the type and distribution of heating devices, the temperature distribution inside the sintering furnace may be uneven during temperature control, which will lead to poor sintering quality. Therefore, before predicting the heating power required by the sintering furnace based on the prediction model and the sintering temperature curve of the desired sintering target and adjusting it in real time, it is necessary to determine whether the temperature distribution of the sintering furnace is uniform.

[0080] Specifically, based on the equations of energy conservation, mass conservation, and momentum conservation, and the structural dimensions of the sintering furnace, a pre-constructed physical model of the sintering furnace was built in ANSYS numerical simulation software for simulation research. The pre-constructed physical model was then compared and analyzed with experimental data to revise the model, resulting in a physical model that accurately reflects the temperature distribution within the sintering furnace (the first sintering furnace physical model). Finally, relevant information on temperature unevenness was extracted from the first sintering furnace physical model. When predicting the heating power, this factor was taken into account. For example, if the ideal temperature during the sintering stage is 360℃ without considering temperature unevenness, the heating power corresponding to 360℃ would be determined based on the prediction model for real-time power adjustment. However, considering the uneven temperature distribution, the actual temperature at sintering rack A at the heating power corresponding to 360℃ is 355℃. Therefore, the heating power needs to be increased to ensure that the temperature at the actual location of the desired sintering target reaches the ideal temperature for the corresponding stage.

[0081] This invention takes into account the uneven temperature distribution inside the sintering furnace. It uses experimental data and a numerical model of the sintering furnace for simulation, compares the simulation results with the actual results, and obtains a physical model (the first sintering furnace physical model) that can truly reflect the temperature distribution inside the sintering furnace. Then, it combines the prediction model, the first sintering furnace physical model and the sintering temperature curve of the target sintering to jointly predict the heating power required by the sintering furnace and adjust it in real time, which further improves the rationality and accuracy of the heating power determination.

[0082] In one embodiment, the simulation results and the sintering furnace temperature of the experimental sintering target are compared and analyzed, and the pre-built sintering furnace physical model is corrected to obtain a first sintering furnace physical model, including:

[0083] A pre-built physical model of a sintering furnace is loaded into the virtual furnace body. Based on the comparison and analysis results of the simulation results and the sintering furnace temperature of the experimental sintering target, the differential temperature points in the virtual furnace body are determined.

[0084] Among them, the virtual furnace body is the digital twin model of the sintering furnace corresponding to the pre-constructed physical model of the sintering furnace; the differential temperature points are the twin points in the virtual furnace body where the simulated temperature and the actual experimental sintering furnace temperature are inconsistent.

[0085] The system displays known influencing factors of temperature difference points at preset viewing locations.

[0086] Among them, the preset viewing location is, for example, the part of the virtual furnace that is not covered near the temperature difference point; the known influencing factors are the influencing factors of the known temperature difference points determined based on the construction principle of the pre-built sintering furnace physical model, which are input by the simulation error attribution personnel;

[0087] The target knowledge structure is indexed based on known influencing factors; the target knowledge structure includes the first unknown influencing factor.

[0088] The target knowledge structure is a knowledge graph structure, and all known influencing factors can be found in the corresponding entities of the target knowledge structure; the first unknown influencing factor is the influencing factor corresponding to the entity that has a structural relationship with the target knowledge structure.

[0089] Extract knowledge structure information based on the target knowledge structure, input the knowledge structure information into the error analysis model trained on the structured storage knowledge based on the target knowledge structure, and obtain the error analysis results.

[0090] Among them, the knowledge structure information is the relevant information of the sintering furnace in reality corresponding to the target knowledge structure, such as: the distribution of sintering racks, the structural size parameters of the sintering furnace, etc.; the structural storage knowledge of the target knowledge structure is the entity relationship corresponding to its graph entity; the error analysis result is the reason for the simulation error analyzed by the error analysis model based on the real information.

[0091] Based on the semantic description of the error analysis results, the first unknown influencing factor is described in the virtual furnace body, and the engine-driven data of the first unknown influencing factor described by the index is indexed.

[0092] In this process, the first unknown influencing factor is described in the virtual furnace body by generating a corresponding virtual part in the virtual furnace body based on the described first unknown influencing factor; the engine-driven data is back-inferred based on the virtual furnace body's preset twin rules and the virtual part generated by the description.

[0093] The pre-built sintering furnace physical model was revised based on engine-driven data;

[0094] Specifically, when correcting the pre-built sintering furnace physical model based on engine-driven data, the pre-built sintering furnace physical model is corrected based on the correlation between the engine-driven data and the pre-built sintering furnace physical model.

[0095] The target knowledge structure, based on known influencing factors, includes:

[0096] Based on the first knowledge structure in the big data index of known influencing factors, determine the second knowledge structure viewed by the viewer;

[0097] The first knowledge structure is a knowledge graph structure based on big data indexing that contains all graph entities corresponding to known influencing factors;

[0098] Based on the second unknown influencing factor in the second knowledge structure, a virtual furnace body is located for viewing by the observer, and the first unknown influencing factor selected by the observer is determined.

[0099] The virtual furnace body is located based on the second unknown influencing factor. This means automatically adjusting the virtual furnace body's perspective and displaying the location of the virtual furnace body that generates the second unknown influencing factor to the viewer. This allows the viewer to verify whether the second unknown influencing factor actually exists. If the influence of the second unknown influencing factor is confirmed, the corresponding second knowledge structure is used as the target knowledge structure, and the corresponding second unknown influencing factor is used as the first unknown influencing factor in the target knowledge structure.

[0100] The working principle and beneficial effects of the above technical solution are as follows:

[0101] This invention performs digital twinning on a pre-constructed physical model of a sintering furnace, determines the differential temperature points in the virtual furnace body, identifies known influencing factors, and simultaneously indexes a target knowledge structure containing known influencing factors. Based on the target knowledge structure, it extracts real-world knowledge structure information and inputs this information into an error analysis model trained on the structured knowledge storage based on the target knowledge structure to obtain error analysis results.

[0102] Based on the semantic description of the error analysis results, the first unknown influencing factor is automatically described in the virtual furnace. Based on the preset twin rules and the corresponding engine-driven data of the virtual part index, the pre-built sintering furnace physical model is corrected according to the correspondence between the engine-driven data and the numerical model, which helps the viewer to correct the pre-built sintering furnace physical model, making it more user-friendly.

[0103] In one embodiment, based on a prediction model, a physical model of the first sintering furnace, and the sintering temperature curve of the desired sintering target, the required heating power of the sintering furnace is predicted and adjusted in real time, including:

[0104] Based on the prediction model and the physical model of the first sintering furnace, the target correspondence is determined as follows: the correspondence between the temperature distribution of the sintering rack and the predicted power.

[0105] The relationship between the temperature distribution of the sintering rack and the predicted power is as follows: the relationship between different predicted powers and the temperature distribution at each sintering rack under the predicted power.

[0106] Calculate the deviation between each target temperature in the sintering rack temperature distribution and the required temperature at the current moment;

[0107] The degree of deviation is the mean square error between each target temperature and the required temperature.

[0108] The predicted power corresponding to the sintering rack temperature distribution with the smallest deviation is used as the target power for real-time adjustment.

[0109] The working principle and beneficial effects of the above technical solution are as follows:

[0110] This invention determines the target correspondence, quantifies the degree of deviation between the sintering rack temperature distribution and the required temperature in the target correspondence, and uses the predicted power corresponding to the sintering rack temperature distribution with the smallest deviation as the target power for real-time adjustment, thereby further improving the suitability of the real-time adjustment heating power setting.

[0111] In one embodiment, based on a prediction model, a physical model of the first sintering furnace, and the sintering temperature curve of the desired sintering target, the heating power required by the sintering furnace is predicted and adjusted in real time, and the method further includes:

[0112] The target ratio is calculated based on the minimum deviation and the required temperature corresponding to each sintering moment. If the target ratio is greater than the preset ratio threshold and the sintering time period is greater than the preset time period length threshold, the interval distribution of the target sintering moment in the sintering time interval is determined.

[0113] The target ratio is the result obtained by dividing the square root of the deviation by the required temperature; the preset ratio threshold and the preset time period length threshold are both set manually; the target sintering time is the sintering time when the target ratio is greater than the preset ratio threshold, and the sintering time period length is the total time length of the target sintering time; the interval distribution is the interval correspondence between the target sintering time and the time interval corresponding to the total sintering process.

[0114] Based on the characteristics of the interval distribution, a deviation correction strategy is determined.

[0115] Among them, the interval distribution feature is: the characteristic representation of the interval correspondence between the target sintering time and the time interval corresponding to the total sintering process, such as: the image features extracted after marking the target sintering time on the time axis corresponding to the time interval; the deviation correction strategy is: the scheme for correcting temperature deviation.

[0116] The working principle and beneficial effects of the above technical solution are as follows:

[0117] When the temperature deviation is the same, the larger the base temperature in the sintering furnace (the temperature required at the sintering time), the smaller the impact of the temperature deviation on the final sintering result. However, the temperature deviation and the required temperature at the sintering time are different throughout the sintering process, making it impossible to accurately quantify the impact of the temperature deviation on the sintering process.

[0118] Therefore, this invention introduces a target ratio, which represents the ratio of the minimum temperature deviation at each sintering moment to the required temperature. The larger the target ratio, the greater the impact of the temperature deviation on the sintering process.

[0119] Since temperature control based on the predicted power corresponding to the sintering rack temperature distribution with the minimum deviation at each sintering moment is already the optimal temperature control state for the current sintering furnace, if the sintering time length corresponding to the statistical target ratio is greater than the preset ratio threshold is still greater than the time length threshold, it indicates that temperature deviation correction is required.

[0120] This invention introduces a target ratio quantification method to measure the impact of temperature deviation on the sintering process in achieving optimal temperature control. Based on the duration of this impact throughout the sintering process, it determines whether temperature deviation correction is necessary. Based on the time information of the deviation, it determines the deviation correction strategy, further improving the subsequent temperature control accuracy.

[0121] In one embodiment, determining the deviation correction strategy based on the interval distribution characteristics includes:

[0122] If the interval distribution characteristics meet the first standard distribution characteristics, the heating device operating status database is read according to the interval distribution characteristics to determine the heating device operating status characteristics.

[0123] Among them, the first standard distribution feature is, for example, the target sintering time is marked on the time axis in clusters and clusters; the heating device working status library stores the working status of the heating device that intervenes in heating at different time points, and reads the working status of the heating device that intervenes in heating at the corresponding target sintering time according to the first standard distribution feature, and its working parameters are the working status features of the heating device.

[0124] Based on the characteristics of the heating device's working status and the database of abnormal working status characteristics, identify abnormal heating devices and perform abnormal maintenance.

[0125] The abnormal operating status feature library includes the operating parameters and abnormal maintenance strategies of different heating devices when they are operating abnormally. The operating status features of the heating device are matched with the abnormal operating status features of the corresponding heating device in the library. If the match is found, maintenance is performed according to the corresponding abnormal maintenance strategy.

[0126] The working principle and beneficial effects of the above technical solution are as follows:

[0127] The timing information of the deviation reveals the possible causes of the deviation. For example, if the deviation only occurs within a certain period of time, and a certain heating device is working simultaneously during that period, then there is a high probability that the heating device is malfunctioning, leading to the heating deviation.

[0128] Therefore, when the target sintering time appears in clusters on the time axis, the operating parameters (heating device operating status characteristics) of the heating device involved in the heating are determined based on the corresponding target sintering time and the heating device operating status database. An abnormal operating status feature database is introduced, and the heating device operating status characteristics are matched with the abnormal operating status characteristics in the database to identify abnormal heating devices and perform abnormal maintenance, thus improving the targeted nature of heating device anomaly identification and handling.

[0129] In one embodiment, determining the deviation correction strategy based on the interval distribution characteristics further includes:

[0130] If the interval distribution characteristics conform to the second standard distribution characteristics, determine the design of the arrangement positions of multiple sintering racks in the sintering furnace.

[0131] Among them, the second standard distribution characteristics are: the target sintering time is marked on the time axis and appears discretely on the entire time axis; the sintering rack arrangement design is: the design scheme of different sintering rack arrangement positions in the sintering furnace;

[0132] The physical model of the first sintering furnace was updated based on the arrangement of the sintering racks to obtain the physical model of the second sintering furnace;

[0133] When the physical model of the first sintering furnace is updated based on the arrangement of the sintering racks to obtain the physical model of the second sintering furnace, the thermodynamic and fluid dynamic relationships in the numerical model are re-determined based on the arrangement of the sintering racks and the model is updated, and then the experimental data is used for verification.

[0134] The first temperature field lines at different simulated powers were simulated based on the physical model of the second sintering furnace.

[0135] Among them, the design position is the sintering rack position designed in the sintering rack arrangement position design; the first temperature field line is the temperature field line that the sintering position corresponding to the sintering position under different power is designed in the sintering rack arrangement position.

[0136] Based on the sintering temperature curve of the desired sintering target, determine the second temperature field line that the design position passes through at different sintering times;

[0137] Among them, the second temperature field line is the ideal temperature field line that the design position passes through at different sintering times;

[0138] Identify and associate the third temperature field line that is most similar to the second temperature field line in the first temperature field line;

[0139] The evaluation value of the sintering rack layout design is determined based on the temperature field line difference between each associated second and third temperature field line.

[0140] The temperature field line difference is the goodness of fit between the associated second and third temperature field lines; the evaluation value is the sum of the goodness of fit at each sintering moment.

[0141] The sintering rack layout position designed for the position with the highest evaluation value is used as the new sintering rack layout position.

[0142] The working principle and beneficial effects of the above technical solution are as follows:

[0143] The timing information of the deviation reveals the possible causes of the deviation. For example, if the deviation occurs discretely throughout the sintering process, it should be attributed to internal factors, such as unreasonable arrangement of the sintering racks leading to uneven heating in the sintering furnace, resulting in random and discrete heating deviations.

[0144] Therefore, multiple sintering rack arrangement designs, besides the current arrangement, are obtained. Since changes in the sintering rack arrangement will alter the fluid motion within the sintering furnace, the thermodynamic and hydrodynamic relationships in the numerical model are redefined based on the designed sintering rack positions, and the model is updated. Experimental data is then used for verification.

[0145] A new physical model of the second sintering furnace was used to simulate the first temperature field line under different simulated powers. The first temperature field line characterizes the temperature field line distribution under all heating conditions of the second sintering furnace physical model. In addition, the ideal temperature field line (second temperature field line) traversed by the design position at different sintering times was determined. The third temperature field line, which is most similar to the second temperature field line among the first temperature field lines, was identified. The third temperature field line represents the temperature distribution that best meets the temperature control requirements at the design position. The sum of the fitting goodness of the second and third temperature field lines corresponding to each sintering time was used as the evaluation value. The sintering rack arrangement position corresponding to the highest evaluation value was determined, improving the rationality of the sintering rack arrangement position setting.

[0146] This invention provides a sintering furnace temperature control system based on a BP neural network prediction model, such as... Figure 2 As shown, it includes:

[0147] Training module 1 is used to build an initial model of a BP neural network based on Adaboost, and to train a prediction model using experimental data and the initial model; the experimental data includes: the sintering furnace temperature and power of the experimental sintering target;

[0148] Control module 2 is used to predict the heating power required by the sintering furnace based on the prediction model and the sintering temperature curve of the desired sintering target, and adjust it in real time.

[0149] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A sintering furnace temperature control method based on a BP neural network prediction model, characterized in that, include: Step 1: Construct an initial model based on an Adaboost-based BP neural network, and train a prediction model using experimental data and the initial model; the experimental data includes: the sintering furnace temperature and power of the experimental sintering target; Step 2: Based on the prediction model and the sintering temperature curve of the desired sintering target, predict the heating power required by the sintering furnace and adjust it in real time; Before predicting the heating power required by the sintering furnace based on the prediction model and the sintering temperature curve of the desired sintering target and adjusting it in real time, it is necessary to determine whether the temperature distribution of the sintering furnace is uniform. If the temperature distribution in the sintering furnace is uneven, a pre-constructed physical model of the sintering furnace is obtained based on the energy conservation, mass conservation, momentum conservation equations and the structural size parameters of the sintering furnace. Based on the power of the experimental sintering target and the pre-constructed physical model of the sintering furnace, the simulation results are determined; By comparing and analyzing the simulation results and the sintering furnace temperature of the experimental sintering target, and correcting the pre-constructed sintering furnace physical model, the first sintering furnace physical model is obtained. Based on the prediction model, the physical model of the first sintering furnace, and the sintering temperature curve of the desired sintering target, the heating power required by the sintering furnace is predicted and adjusted in real time. Among them, based on the prediction model, the physical model of the first sintering furnace, and the sintering temperature curve of the desired sintering target, the heating power required by the sintering furnace is predicted and adjusted in real time, including: Based on the prediction model and the physical model of the first sintering furnace, the target correspondence is determined as follows: the correspondence between the temperature distribution of the sintering rack and the predicted power. Calculate the deviation between each target temperature in the sintering rack temperature distribution and the required temperature at the current moment; The predicted power corresponding to the sintering rack temperature distribution with the smallest deviation is used as the target power for real-time adjustment. The target ratio is calculated based on the minimum deviation and the required temperature corresponding to each sintering moment. If the target ratio is greater than the preset ratio threshold and the sintering time period is greater than the preset time period length threshold, the interval distribution of the target sintering moment in the sintering time interval is determined. Based on the characteristics of the interval distribution, a deviation correction strategy is determined.

2. The sintering furnace temperature control method based on a BP neural network prediction model as described in claim 1, characterized in that, Methods for measuring the temperature of a sintering furnace include: K-type thermocouples were used to measure the temperature distribution at various points in the space inside the sintering furnace over time.

3. The sintering furnace temperature control method based on a BP neural network prediction model as described in claim 1, characterized in that, Based on the characteristics of the interval distribution, determine the deviation correction strategy, including: If the interval distribution characteristics meet the first standard distribution characteristics, the heating device operating status database is read according to the interval distribution characteristics to determine the heating device operating status characteristics. Based on the characteristics of the heating device's operating status and the database of abnormal operating status characteristics, identify abnormal heating devices and perform abnormal maintenance.

4. A sintering furnace temperature control system based on a BP neural network prediction model, employing the sintering furnace temperature control method based on a BP neural network prediction model as described in claim 1, characterized in that... include: The training module is used to build an initial model of a BP neural network based on Adaboost, and to train a prediction model using experimental data and the initial model; the experimental data includes: the sintering furnace temperature and power of the experimental sintering target; The control module is used to predict the heating power required by the sintering furnace based on the prediction model and the sintering temperature curve of the desired sintering target, and to adjust it in real time.

Images