Metallurgy intelligent safety temperature control and fault early warning system and method
By collecting and processing data during the metallurgy process, building a temperature prediction model and conducting fault analysis, the problems of metallurgical temperature fluctuations and fault warning are solved, and the safety and intelligence of metallurgical production are improved.
Patent Information
- Application Number
- CN202510597477.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-05-09
- Publication Date
- 2025-08-12
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
It is difficult to establish an accurate temperature prediction model during the metallurgical process, resulting in large temperature fluctuations, affecting product quality and production safety. It is difficult for the existing technology to effectively analyze the types of faults and promptly warn, affecting the continuity and safety of production.
Metallurgical process data is collected through the data acquisition module, and the temperature prediction module is used to construct and train the model based on the processed heating and casting process data to obtain the trained temperature prediction model. The fault type analysis coefficient is calculated based on the operation process data through the fault warning module, and the fault warning is performed in combination with the preset alarm threshold.
It realizes precise control of metallurgical temperature and timely early warning of faults, ensures the continuity and safety of metallurgical production, reduces maintenance costs and production risks, and improves the intelligence and automation level of metallurgical production.
Smart Images
Figure CN120469388A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the field of metallurgical intelligent technology, and in particular relates to a metallurgical intelligent safety temperature control and fault early warning system and method. Background Art
[0002] With the development of intelligent metallurgical technology, technologies for intelligent and safe temperature control and fault warning in metallurgy have emerged. However, the metallurgical process faces many challenges, which restrict the level of intelligent production and the controllability of the production process. The transmission process of natural gas heating furnaces is complex, and it is difficult to establish an accurate metallurgical temperature prediction model. It is difficult to grasp the changes in metallurgical temperature in real time, which in turn affects the mechanical properties and internal quality of the metallurgy. In the deep well casting process, the current reliance on manual experience and intermittent temperature measurement leads to large fluctuations in metallurgical temperature and unstable product quality. In addition, the types of faults during the operation of metallurgical processing equipment are diverse and complex. Existing technologies cannot effectively analyze the fault types and issue timely warnings, which affects the continuity and safety of production. Summary of the Invention
[0003] Based on this, it is necessary to provide a metallurgical intelligent safety temperature control and fault warning system and method that can effectively control the metallurgical temperature and detect potential equipment failure risks in advance to address the above technical problems.
[0004] In a first aspect, the present application provides an intelligent safety temperature control and fault warning system for metallurgy, the system comprising:
[0005] The data acquisition module is used to obtain the heating process data of the natural gas heating furnace in the metallurgical process, the deep well casting process data and the operation process data of the processing equipment.
[0006] The temperature prediction module is used to build and train a model based on the processed heating process data and casting process data to obtain a trained temperature prediction model; the trained temperature prediction model is used to calculate using a formula based on the processed heating process data and casting process data to obtain a temperature prediction value.
[0007] The fault warning module is used to calculate the fault type analysis coefficient of the processing equipment during the operation period based on the operation process data; it is also used to judge the fault type analysis coefficient and the temperature prediction value based on the preset alarm threshold to obtain the fault warning result.
[0008] In one embodiment, model construction and training are performed based on the processed heating process data and casting process data to obtain a trained temperature prediction model, including:
[0009] The processed heating process data and casting process data are randomly divided into training sets and test sets; the training set accounts for 80% and the test set accounts for 20%.
[0010] Recursive feature elimination and logistic regression algorithms are used to screen the data in the training set and obtain the corresponding key feature factors.
[0011] The recurrent neural network algorithm is used to construct a model of key characteristic factors to obtain a pre-trained temperature prediction model.
[0012] The loss function based on the maximum mean difference measures the difference in data distribution between the training set and the test set, and obtains the loss function corresponding to the pre-trained temperature prediction model.
[0013] The back propagation algorithm is used to continuously adjust the parameters to minimize the total loss function, and a trained temperature prediction model for the predicted metallurgical temperature is obtained.
[0014] In one embodiment, a loss function based on the maximum mean difference measures the difference in data distribution between the training set and the test set, and obtains a loss function corresponding to the pre-trained temperature prediction model, including:
[0015] The loss function based on the maximum mean difference is used to measure the difference in data distribution between the training set and the test set to obtain the data difference value.
[0016] The loss function formula of the maximum mean difference is expressed as follows:
[0017]
[0018] Among them, MMD 2 (D s , D t ) represents the maximum mean difference between the data in the training set and the test set, φ(*) represents the function that maps the data to the reproducing kernel Hilbert space, D s represents the training set, D t represents the test set, n and m represent the number of data samples in the training set and test set respectively.
[0019] The total loss function is calculated based on the data difference value combined with the cross entropy loss function of the test set data.
[0020] L=α*MMD 2 (D s , D t )+(1-α)L ce
[0021] Among them, L ce Represents the cross entropy loss function, α represents the balance parameter, and its value range is [0, 1].
[0022] In one embodiment, obtaining a trained temperature prediction model further includes:
[0023] The following formula is used to construct the fitness objective function based on the key characteristic factors in the temperature prediction model using a linear weighted multi-objective optimization method.
[0024]
[0025] Among them, Fitness represents the fitness value; α represents the proportional weight; normal(·) represents the normalization function, and CS represents the key characteristic factor.
[0026] Based on the fitness objective function, the position of the particle is initialized in the feasible solution space to obtain the particle position.
[0027] The particle position of each particle is marked as its corresponding historical optimal position and calculated using the global optimal formula to obtain the global optimal position corresponding to the position of the particle with the minimum fitness value in the particle swarm.
[0028] The global optimal formula is expressed as follows:
[0029]
[0030] Among them, best t represents the global optimal position of the particle swarm, Indicates the historical optimal position corresponding to the current particle position of each particle, It represents the fitness value corresponding to the historical optimal position of the i-th particle at the t-th iteration, N represents the size of the particle swarm, and t represents the number of iterations.
[0031] Based on the preset loop termination condition, the dual-strategy search particle swarm algorithm is used to iteratively calculate the global optimal position to obtain the optimal parameters corresponding to the particle position; the loop termination condition is that the fitness value of the particle is less than the preset temperature prediction error threshold or the number of particle swarm iterations reaches the maximum number of iterations.
[0032] The temperature prediction model is updated based on the optimal parameters to obtain a trained temperature prediction model.
[0033] In one embodiment, the temperature prediction model obtains the temperature prediction value by:
[0034] The key characteristic factors are input into the trained temperature prediction model to obtain the basic value of metallurgical temperature prediction.
[0035] The SHAP value is calculated based on each key characteristic factor to obtain the SHAP value corresponding to each key characteristic factor.
[0036] The SHAP value is calculated as follows:
[0037]
[0038] in, represents the SHAP value of key feature factor i, S represents the subset of key feature factors, f(S) represents the predicted basic value of the model when only the key feature factors in subset S are used, and f(S∪{i}) represents the predicted basic value of the model after adding the key feature factors to subset S.
[0039] The metallurgical temperature prediction value is obtained by calculation based on the prediction base value and the SHAP value corresponding to the key characteristic factors.
[0040] T pr =T ba +Σ j SHAPvalue j
[0041] Among them, T pr Represents the predicted metallurgical temperature, T ba Represents the basic value of metallurgical temperature prediction, j represents the key characteristic factor, SHAPvalue j Indicates the SHAP value corresponding to the key feature factor j.
[0042] In one embodiment, after obtaining the SHAP value corresponding to each key feature factor, the method further includes:
[0043] Based on the data in the test set, the SHAP values corresponding to each key characteristic factor were interpreted to obtain the characteristic influence of each key characteristic factor on the temperature prediction model.
[0044] Use the following formula to calculate the feature influence:
[0045]
[0046] Among them, Q j It represents the characteristic influence of the jth key characteristic factor on the metallurgical temperature prediction basic value output by the temperature prediction model. Assume that X i,j represents the jth feature of the i-th test set sample, Represents X i,j SHAP value, n represents the total number of test set samples.
[0047] The temperature prediction model is evaluated based on the feature influence using K-fold cross validation to obtain the final temperature prediction model; the final temperature prediction model is the temperature prediction model with the best performance evaluation result.
[0048] In one embodiment, the fault type analysis coefficient of the processing equipment during the operation period is obtained by calculating the formula based on the operation process data, including:
[0049] The key features of the acquired operation process data are extracted to obtain the operation characteristic parameters.
[0050] Use the following formula to calculate the operating characteristic parameters and obtain the fault type analysis coefficient of the processing equipment during the operating period:
[0051]
[0052] Among them, E i Indicates the failure type analysis coefficient of the processing equipment during the operation period, PL i Indicates the frequency difference between the instantaneous and slow occurrence of processing equipment failures during the operation period, CS i Indicates the percentage of times that processing equipment continues to operate after operating adjustments after a fault occurs, GH i It represents the probability of replacing equipment parts after a processing equipment failure occurs during the operating period; ρ represents the error correction factor, ω1, ω2 and ω3 all represent preset proportional coefficients, and ω1>ω2>ω3>1.
[0053] In a second aspect, the present application also provides a method for intelligent safety temperature control and fault warning in metallurgy, the method comprising:
[0054] Acquire natural gas heating furnace heating process data, deep well casting process data, and processing equipment operation process data in the metallurgical process.
[0055] The model is constructed and trained based on the processed heating process data and casting process data to obtain a trained temperature prediction model; the trained temperature prediction model is used to calculate based on the processed heating process data and casting process data using a formula to obtain a temperature prediction value.
[0056] The fault type analysis coefficient of the processing equipment during the operation period is calculated based on the operation process data; the fault type analysis coefficient and the temperature prediction value are judged based on the preset alarm threshold to obtain the fault warning result.
[0057] In a third aspect, the present application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor implements the above system and method when executing the computer program.
[0058] In a fourth aspect, the present application also provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, the above system and method are implemented.
[0059] The above-mentioned intelligent safety temperature control and fault warning system and method for metallurgy, the data acquisition module collects natural gas heating furnace heating process data, deep well casting process data, and processing equipment operation process data in the metallurgical process. The temperature prediction module constructs and trains a model based on the processed heating process data and casting process data to obtain a trained temperature prediction model. The model calculates the temperature prediction value based on the processed heating and casting process data using a specific formula. The fault warning module calculates the fault type analysis coefficient of the processing equipment during the operation period based on the operation process data, and at the same time, based on the preset alarm threshold, judges the fault type analysis coefficient and the temperature prediction value to obtain a fault warning result. The above steps ensure the quality of metallurgical products and reasonable energy consumption, can promptly detect potential faults in the operation of processing equipment and issue early warnings of temperature anomalies, effectively ensure the continuity and safety of metallurgical production, reduce maintenance costs and production risks, and improve the intelligence and automation level of metallurgical production. BRIEF DESCRIPTION OF THE DRAWINGS
[0060] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the following briefly introduces the drawings required for use in the embodiments or related technical descriptions. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.
[0061] Figure 1 A structural block diagram of a metallurgical intelligent safety temperature control and fault warning system provided by an embodiment of the present invention;
[0062] Figure 2 A flowchart of a temperature prediction model obtained by constructing and training a model based on processed heating process data and casting process data provided in an embodiment of the present invention;
[0063] Figure 3 A flow chart of steps for obtaining a trained temperature prediction model provided in an embodiment of the present invention;
[0064] Figure 4 A flow chart of obtaining a temperature prediction value using a temperature prediction model provided in an embodiment of the present invention;
[0065] Figure 5 A flow chart of a method for intelligent safety temperature control and fault warning in metallurgy provided by an embodiment of the present invention. DETAILED DESCRIPTION
[0066] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.
[0067] First, the implementation environment of the embodiment of the present application is described. By way of example, the implementation environment includes a data monitoring device, a data processing system, and an alarm display device.
[0068] In this intelligent, safe temperature control and fault warning system and method for metallurgy, data monitoring equipment accurately transmits real-time data such as temperature, pressure, and flow rate to a data processing system via wired or wireless transmission. Upon receiving the data, the data processing system applies complex algorithms and models for analysis and calculation. Upon identifying an anomaly, it quickly generates a command and transmits the alarm information to an alarm display device via a communication link. Upon receiving the command, the alarm display device intuitively displays the warning information through audio, visual, and pop-up windows, prompting operators to take timely countermeasures.
[0069] Data monitoring equipment includes a variety of sensor types, such as thermocouples and RTDs, which accurately measure temperatures within the furnace and in key areas. These sensors can sensitively detect subtle temperature changes and convert them into electrical signals. Pressure sensors monitor furnace pressure in real time to prevent temperature fluctuations or equipment failures caused by abnormal pressure. Flow sensors monitor the flow of cooling media, fuel, and other components, ensuring stable supply and indirectly maintaining constant temperature and normal equipment operation. Sensors are distributed throughout the metallurgical equipment to continuously collect data.
[0070] Data processing systems typically consist of powerful industrial computers and efficient algorithmic software. These computers receive massive amounts of real-time data transmitted from data monitoring equipment via wired or wireless networks. Complex algorithms and models pre-configured within the system then perform in-depth analysis, comparison, and computation on this data. By mining this data, they can accurately identify normal and abnormal operating conditions. If parameters like temperature and pressure deviate from preset safety ranges, or if early signs of equipment failure are detected, appropriate instructions are rapidly generated, providing a basis for subsequent alarms and equipment control.
[0071] The alarm display device is the operator's direct window into the intelligent metallurgical safety temperature control and fault warning system and method. When the data processing system detects an abnormality and issues an alarm, the alarm display presents the warning information in a variety of intuitive ways. The audible and visual alarm instantly emits a loud siren and flashes lights, quickly attracting the operator's attention. The industrial display screen uses eye-catching graphics, curves, and text to detail the abnormality's location, type, and current values of key parameters. This information allows operators to quickly and accurately understand the on-site situation and take effective countermeasures to prevent the incident from escalating, thereby ensuring the safe and stable operation of metallurgical production.
[0072] In combination with the above implementation environment, the application scenarios of the embodiments of the present application are explained.
[0073] The embodiment of the present application provides a system and method for intelligent safety temperature control and fault warning in metallurgy. With the help of various data monitoring devices throughout the metallurgical equipment, key parameter data such as temperature and pressure are collected in real time, and quickly transmitted to the data processing system through a wired or wireless network. The system uses advanced algorithms to deeply analyze and process data and accurately identify abnormal working conditions. Once a problem is found, an instruction is immediately sent to the alarm display device to remind the operator in the form of sound and light alarms, screen displays, etc. At the same time, the emergency plan is automatically started and the relevant equipment is adjusted to achieve safe temperature control and timely warning of faults in the metallurgical process. For example, the system and method for intelligent safety temperature control and fault warning in metallurgy provided in the embodiment of the present application can be applied to at least one scenario including but not limited to the following scenarios.
[0074] First, this intelligent, safe temperature control and fault warning system and method for metallurgy has been applied to blast furnace ironmaking. During the blast furnace ironmaking process, thermocouples installed at various heights and in key locations within the furnace capture temperature changes in real time, while pressure sensors continuously monitor pressure. Data collected by the equipment is rapidly transmitted to a data processing system via Industrial Ethernet. The data processing system uses complex algorithms to deeply analyze the changing trends and correlations of temperature, pressure, and other data. When an abnormal temperature rise or pressure fluctuation outside a safe range is detected within the furnace, an alarm is quickly generated. The alarm display device responds immediately, with a sharp audible and visual alarm sounding and a flashing red light. The industrial display screen displays the specific location of the anomaly, the current temperature and pressure values, and the normal range in a striking pop-up window. Operators can use this information to promptly adjust operations such as blasting and charging to ensure stable blast furnace operation and avoid equipment failures and production accidents caused by temperature and pressure anomalies.
[0075] Second, this intelligent, safe temperature control and fault warning system and method for metallurgy has been applied to converter steelmaking. During converter steelmaking, immersed thermocouples accurately measure the temperature of the molten steel, and flow sensors closely monitor the flow of oxygen, coolant, and other media. The collected data is rapidly transmitted to the data processing system via a wireless transmission module. Based on a steelmaking process model, the data processing system performs computational analysis on large amounts of real-time data to create a dynamic simulation of the steelmaking process. If conditions such as temperature deviations and flow anomalies that could impact molten steel quality or equipment safety are detected, an alarm is immediately sent to the alarm display device. This device uses rapid alarms and flashing lights to attract the operator's attention, ensuring efficient and safe converter steelmaking and improving molten steel quality.
[0076] Third, the intelligent, safe temperature control and fault warning system and method for metallurgy is applied in continuous casting scenarios. During the continuous casting process, data monitoring equipment conducts comprehensive monitoring of the cooling process of the ingot. Temperature sensors arranged on the surface of the ingot and around the crystallizer continuously collect temperature data to ensure uniform cooling of the ingot, and liquid level sensors control the molten steel level in real time. The data is transmitted to the data processing system via a wired network. The system uses professional algorithms to integrate and analyze the data and establish a real-time ingot status model. When problems such as uneven ingot temperature and abnormal liquid level fluctuations that may lead to ingot defects or equipment failure are detected, the data processing system immediately issues an early warning to the alarm display device. The alarm display device alerts the operator through sound and light alarms, and the industrial display clearly displays the real-time status of the ingot and abnormal data, helping the operator to promptly adjust parameters such as the cooling water volume and billet drawing speed to ensure a smooth continuous casting process, improve ingot quality, and reduce equipment failure rates.
[0077] In one embodiment, Figure 1 As shown, the present application provides a metallurgical intelligent safety temperature control and fault warning system, which may include:
[0078] The data acquisition module 101 is used to acquire the heating process data of the natural gas heating furnace, the deep well casting process data and the operation process data of the processing equipment in the metallurgical process.
[0079] Specifically, the heating process data include at least one of metallurgical weight, heating time, natural gas flow, actual temperature in the furnace, metallurgical target temperature, heating furnace power, natural gas calorific value and metallurgical initial temperature; the casting process data include at least one of metallurgical liquid level, casting speed, cooling medium flow, cooling medium temperature, deep well depth, casting yield, type of additives and addition amount; the operation process data include at least one of the difference in frequency between instantaneous and slow occurrence of operation faults, the proportion of times the operation is continued through operation adjustment after the operation fault occurs, the probability of replacing equipment parts after the operation fault occurs, the deviation value of the operation operation execution time point, the ratio of historical failure rate to operation error rate, the deviation value between the real-time environmental parameters and the critical value of the qualified execution environment parameters, and the frequency of reciprocating changes in the floating trend of the real-time environmental parameter values.
[0080] The temperature prediction module 102 is used to construct and train a model based on the processed heating process data and casting process data to obtain a trained temperature prediction model; the trained temperature prediction model is used to calculate using a formula based on the processed heating process data and casting process data to obtain a temperature prediction value.
[0081] Specifically, first, a series of pre-processing methods such as data cleaning, feature extraction, and normalization are used to collect data on the heating process of natural gas heating furnaces and deep well casting processes to remove noise, fill missing values, and screen out key features closely related to temperature changes, laying a solid foundation for subsequent model training. Subsequently, appropriate machine learning algorithms, such as recurrent neural networks and long short-term memory networks, are selected and combined with the pre-processed data to build a model. Through a large number of iterative training, the model parameters are continuously adjusted so that the model can accurately capture the complex patterns and temperature change laws in the data, thereby obtaining a trained temperature prediction model. In the actual application stage, this trained temperature prediction model will again receive processed heating and casting process data. According to specific calculation formulas, such as linear or nonlinear combination formulas that comprehensively consider multiple key characteristic factors, the input data will be deeply analyzed and calculated, and finally an accurate temperature prediction value will be output, providing a reliable basis for temperature control in the metallurgical production process.
[0082] The fault warning module 103 is used to calculate the fault type analysis coefficient of the processing equipment during the operation period based on the operation process data; it is also used to judge the fault type analysis coefficient and the temperature prediction value based on the preset alarm threshold to obtain the fault warning result.
[0083] First, the module applies advanced data mining and analysis techniques, such as principal component analysis and support vector machine algorithms, to the equipment's operational data. Key features are extracted from multi-dimensional operational data, including vibration, temperature, pressure, and current. A specific calculation formula then comprehensively considers factors such as the frequency difference between instantaneous and gradual equipment failures, the percentage of equipment that can resume operation after a failure, and the probability of component replacement after a failure. Combined with an error correction factor and a preset proportional coefficient, the module accurately calculates the failure type analysis coefficient for the equipment during the operating period. The module also compares this coefficient with the temperature prediction value output by the temperature prediction module and the preset alarm threshold. If either the failure type analysis coefficient or the temperature prediction value exceeds the corresponding threshold, the system identifies a potential failure risk and quickly generates a fault warning, effectively preventing production downtime and economic losses caused by sudden equipment failures and ensuring the safe and efficient operation of metallurgical production processes.
[0084] The above-mentioned intelligent safety temperature control and fault warning system for metallurgy is used. The data acquisition module collects the heating process data of the natural gas heating furnace, the deep well casting process data, and the operation process data of the processing equipment in the metallurgical process. The temperature prediction module constructs and trains the model based on the processed heating process data and the casting process data to obtain a trained temperature prediction model. The model calculates the temperature prediction value based on the processed heating and casting process data using a specific formula. The fault warning module calculates the fault type analysis coefficient of the processing equipment during the operation period based on the operation process data. At the same time, based on the preset alarm threshold, it judges the fault type analysis coefficient and the temperature prediction value to obtain the fault warning result. The above steps ensure the quality of metallurgical products and reasonable energy consumption, can timely detect potential faults in the operation of processing equipment and issue early warnings of temperature anomalies, effectively ensure the continuity and safety of metallurgical production, reduce maintenance costs and production risks, and improve the intelligence and automation level of metallurgical production.
[0085] In one embodiment, Figure 2 As shown, model construction and training are performed based on the processed heating process data and casting process data to obtain a trained temperature prediction model, which may include the following steps:
[0086] In step S201 , the processed heating process data and casting process data are randomly divided to obtain a training set and a test set; the training set accounts for 80% and the test set accounts for 20%.
[0087] Step S202: Using recursive feature elimination and logistic regression algorithms, the data in the training set are subjected to feature screening to obtain corresponding key feature factors.
[0088] Step S203: construct a model for key characteristic factors using a recurrent neural network algorithm to obtain a pre-trained temperature prediction model.
[0089] Step S204 , measuring the difference in data distribution between the training set and the test set based on the loss function of the maximum mean difference, and obtaining the loss function corresponding to the pre-trained temperature prediction model.
[0090] Step S205 , using a back propagation algorithm to continuously adjust parameters so as to minimize the total loss function, thereby obtaining a trained temperature prediction model for predicting the metallurgical temperature.
[0091] Specifically, the processed heating process data and casting process data are first randomly divided into a training set and a test set. Recursive feature elimination and logistic regression algorithms are then used to screen the data in the training set. This algorithm selects key characteristic factors closely related to temperature prediction from a large number of data features. A recurrent neural network algorithm is then used to construct a model based on the screened key characteristic factors, initially obtaining a pre-trained temperature prediction model. The difference in data distribution between the training set and the test set is then measured using the maximum mean difference loss function, thereby deriving the loss function corresponding to the pre-trained temperature prediction model. Finally, the backpropagation algorithm is used to continuously adjust the model parameters to minimize the overall loss function, thereby obtaining a trained temperature prediction model that can accurately predict metallurgical temperatures.
[0092] This embodiment uses recursive feature elimination and logistic regression algorithms to screen the data features of the training set, effectively remove redundant information, extract key characteristic factors closely related to temperature prediction, significantly improve model training efficiency and prediction accuracy, and reduce the risk of overfitting. With the help of the recurrent neural network algorithm, a model is constructed based on key characteristic factors, giving full play to its advantages in processing sequence data, deeply mining the time series laws in the data, and accurately capturing the temperature change trend. The loss function based on the maximum mean difference measures the data distribution difference between the training set and the test set, which helps to optimize the performance of the model under different data distributions and improve the adaptability of the model. Finally, the back propagation algorithm is used to continuously adjust the parameters to minimize the total loss function, so that the model parameters tend to the optimal solution, thereby obtaining a highly accurate trained temperature prediction model. This model can provide a reliable basis for precise temperature control in the metallurgical production process, effectively guarantee product quality, achieve energy consumption optimization, greatly improve the intelligence and automation level of metallurgical production, and reduce production costs and potential risks.
[0093] In one embodiment, measuring the difference in data distribution between the training set and the test set based on a maximum mean difference loss function to obtain a loss function corresponding to a pre-trained temperature prediction model may include the following steps:
[0094] Step S301: Use a loss function based on the maximum mean difference to measure the difference in data distribution between the training set and the test set to obtain a data difference value.
[0095] The loss function formula of the maximum mean difference is expressed as follows:
[0096]
[0097] Among them, MMD 2 (D s , D t ) represents the maximum mean difference between the data in the training set and the test set, φ(*) represents the function that maps the data to the reproducing kernel Hilbert space, D s represents the training set, D t represents the test set, n and m represent the number of data samples in the training set and test set respectively.
[0098] Step S302 , calculating the cross entropy loss function of the data in the test set based on the data difference value to obtain a total loss function.
[0099] L=α*MMD 2 (D s , D t )+(1-α)L ce
[0100] Among them, L ce Represents the cross entropy loss function, α represents the balance parameter, and its value range is [0, 1].
[0101] Specifically, we first use a loss function based on the maximum mean difference to measure the difference in data distribution between the training set and the test set. Then, we perform a comprehensive calculation based on the obtained data difference value and the cross-entropy loss function of the test set data to derive the total loss function.
[0102] This embodiment uses the maximum mean difference loss function to measure data distribution differences, which can clearly quantify the similarity between the data distribution of the training set and the test set, and provide a key basis for the adaptive adjustment of the model between different data subsets. The total loss function is calculated in combination with the cross entropy loss function, which takes into account both the consistency of the data distribution and the accuracy of the model in classifying or predicting the test set data. This enables the model to better balance the relationship between data distribution differences and prediction accuracy during training, thereby optimizing model performance, improving the accuracy of the model's prediction of metallurgical temperature, and providing more reliable model support for temperature control in metallurgical production processes.
[0103] In one embodiment, Figure 3 As shown, obtaining a trained temperature prediction model may further include the following steps:
[0104] Step S401 : Using the following formula, a fitness objective function is constructed based on key characteristic factors in the temperature prediction model using a linear weighted multi-objective optimization method.
[0105]
[0106] Among them, Fitness represents the fitness value; α represents the proportional weight; normal(·) represents the normalization function, and CS represents the key characteristic factor.
[0107] Step S402: Initialize the position of the particle in the feasible solution space based on the fitness objective function to obtain the particle position.
[0108] Step S403 , marking the particle position of each particle as its corresponding historical optimal position and calculating using the global optimal formula to obtain the global optimal position corresponding to the position of the particle with the minimum fitness value in the particle swarm.
[0109] The global optimal formula is expressed as follows:
[0110]
[0111] Among them, best t represents the global optimal position of the particle swarm, Indicates the historical optimal position corresponding to the current particle position of each particle, It represents the fitness value corresponding to the historical optimal position of the i-th particle at the t-th iteration, N represents the size of the particle swarm, and t represents the number of iterations.
[0112] In step S404, the dual-strategy search particle swarm algorithm is used to iteratively calculate the global optimal position based on the preset loop termination condition to obtain the optimal parameters corresponding to the particle position; the loop termination condition is that the fitness value of the particle is less than the preset temperature prediction error threshold or the number of particle swarm iterations reaches the maximum number of iterations.
[0113] Step S405: Update the temperature prediction model based on the optimal parameters to obtain a trained temperature prediction model.
[0114] Specifically, based on the key characteristic factors of the temperature prediction model, a linearly weighted multi-objective optimization method is used to construct a fitness objective function. Based on this fitness objective function, the particle positions are initialized within the feasible solution space to obtain the particle positions. The particle positions of each particle are then marked as their corresponding historical optimal positions. The global optimal position of the particle swarm is then determined using a global optimal formula. Then, based on the preset loop termination conditions (i.e., the particle fitness value is less than the preset temperature prediction error threshold or the number of particle swarm iterations reaches the maximum number of iterations), a dual-strategy search particle swarm algorithm is used to iteratively calculate the global optimal position, thereby obtaining the optimal parameters corresponding to the particle positions. Finally, the temperature prediction model is updated based on the obtained optimal parameters, ultimately obtaining a trained temperature prediction model.
[0115] The optimal parameters obtained in this example update the temperature prediction model, enabling the model to fully learn the underlying patterns in the data, significantly improving its accuracy and stability. This provides a more accurate and reliable model for metallurgical temperature prediction, helping to improve temperature control during metallurgical production, ensure product quality, reduce production costs, and enhance the market competitiveness of enterprises.
[0116] In one embodiment, Figure 4 As shown, the temperature prediction model can obtain the temperature prediction value in the following ways:
[0117] Step S501: Input each key characteristic factor into the trained temperature prediction model to obtain a basic value for metallurgical temperature prediction.
[0118] Step S502 : SHAP value calculation is performed based on each key feature factor to obtain the SHAP value corresponding to each key feature factor.
[0119] The SHAP value is calculated as follows:
[0120]
[0121] in, represents the SHAP value of key feature factor i, S represents the subset of key feature factors, f(S) represents the predicted basic value of the model when only the key feature factors in subset S are used, and f(S∪{i}) represents the predicted basic value of the model after adding the key feature factors to subset S.
[0122] Step S503 , performing calculations based on the prediction base value and the SHAP values corresponding to the key characteristic factors to obtain the metallurgical temperature prediction value.
[0123] T pr =T ba +Σ j SHAPvalue j
[0124] Among them, T pr Represents the predicted metallurgical temperature, T ba Represents the basic value of metallurgical temperature prediction, j represents the key characteristic factor, SHAPvalue j Indicates the SHAP value corresponding to the key feature factor j.
[0125] First, the previously selected key characteristic factors are input into the trained temperature prediction model. The model is then used to calculate the metallurgical temperature prediction base value. The SHAP value corresponding to each key characteristic factor is then calculated using a formula. Finally, based on the obtained prediction base value and the SHAP value corresponding to the key characteristic factor, the metallurgical temperature prediction value is calculated using a formula.
[0126] This embodiment inputs key characteristic factors into the trained model to obtain the prediction base value, making full use of the rules that the model has learned to provide a basis for subsequent accurate predictions. Calculating the SHAP value can deeply analyze the degree of influence of each key characteristic factor on the model prediction results, making the model's prediction results more interpretable. By combining the prediction base value with the SHAP value to calculate the metallurgical temperature prediction value, the overall prediction ability of the model and the individual contribution of each key characteristic factor are comprehensively considered, greatly improving the accuracy of the metallurgical temperature prediction. This helps to ensure the quality stability of aluminum products, optimize production processes, reduce energy consumption and production costs, improve the intelligence and refinement level of metallurgical production, and enhance the competitiveness of enterprises in the market.
[0127] In one embodiment, after obtaining the SHAP value corresponding to each key characteristic factor, the following steps may be further included:
[0128] Step S601: interpreting the SHAP value corresponding to each key characteristic factor based on the data in the test set to obtain the characteristic influence of each key characteristic factor on the temperature prediction model.
[0129] Use the following formula to calculate the feature influence:
[0130]
[0131] Among them, Q j It represents the characteristic influence of the jth key characteristic factor on the metallurgical temperature prediction basic value output by the temperature prediction model. Assume that X i,j represents the jth feature of the i-th test set sample, Represents X i,j SHAP value, n represents the total number of test set samples.
[0132] Step S602 , performing performance evaluation on the temperature prediction model using K-fold cross validation based on the feature influence to obtain a final temperature prediction model; the final temperature prediction model is the temperature prediction model with the best performance evaluation result.
[0133] Specifically, the SHAP values corresponding to each key characteristic factor were calculated using a formula based on the data in the test set, deriving the characteristic influence of each key characteristic factor on the temperature prediction model. Based on the obtained characteristic influence, a K-fold cross-validation method was then used to conduct a comprehensive performance evaluation of the temperature prediction model. During this process, the dataset was partitioned multiple times and the model was repeatedly trained and validated. By comprehensively considering the performance of the models under different partitions, the model with the best performance evaluation results was selected and determined as the final temperature prediction model.
[0134] This embodiment ensures that the final temperature prediction model selected has good performance on different data subsets, greatly improving the stability, accuracy and generalization ability of the model, providing more reliable guarantees for accurate temperature prediction in metallurgical production processes, helping to improve production efficiency, ensure product quality, reduce production costs, and promote the intelligent development of the metallurgical industry.
[0135] In one embodiment, obtaining a fault type analysis coefficient of a processing device during an operation period by calculating a formula based on the operation process data may include the following steps:
[0136] Step S701 : extract key features from the acquired operation process data to obtain operation feature parameters.
[0137] Step S702: Calculate the operating characteristic parameters using the following formula to obtain the fault type analysis coefficient of the processing equipment during the operating period:
[0138]
[0139] Among them, E i Indicates the failure type analysis coefficient of the processing equipment during the operation period, PL i Indicates the frequency difference between the instantaneous and slow occurrence of processing equipment failures during the operation period, CS i Indicates the percentage of times that processing equipment continues to operate after operating adjustments after a fault occurs, GH i It represents the probability of replacing equipment parts after a processing equipment failure occurs during the operating period; ρ represents the error correction factor, ω1, ω2 and ω3 all represent preset proportional coefficients, and ω1>ω2>ω3>1.
[0140] First, we conduct a detailed analysis of the acquired processing equipment operating data. Using advanced data processing techniques and algorithms, we extract key features and generate operational characteristic parameters. These parameters encompass key information about the equipment's operation and provide a valuable foundation for subsequent analysis. We then use a formula to perform comprehensive calculations on these parameters, ultimately deriving the failure type analysis coefficients for the processing equipment during its operation.
[0141] This embodiment extracts key features from operational process data, accurately filtering out important information related to equipment failures from complex data, removing redundant data, and improving analysis efficiency and accuracy. A specific formula is used to calculate the fault type analysis coefficient, comprehensively considering multiple factors such as the frequency characteristics of equipment failures, the operating status of the equipment after the failure, and the replacement of parts. At the same time, the error correction factor and the preset proportional coefficient are set to make the calculation results more scientific and reasonable. The resulting fault type analysis coefficient can provide a strong basis for equipment fault diagnosis and maintenance, helping staff understand the potential failure risks of the equipment in advance, take preventive measures in a timely manner, and reduce equipment downtime and maintenance costs.
[0142] In one embodiment, Figure 5 As shown, the present application also provides a method for intelligent safety temperature control and fault warning in metallurgy, which may include the following steps:
[0143] Step S801, obtaining the natural gas heating furnace heating process data, deep well casting process data and processing equipment operation process data in the metallurgical process.
[0144] Step S802: construct and train a model based on the processed heating process data and casting process data to obtain a trained temperature prediction model; the trained temperature prediction model is used to calculate using a formula based on the processed heating process data and casting process data to obtain a temperature prediction value.
[0145] Step S803, calculating the fault type analysis coefficient of the processing equipment in the operation period based on the operation process data; judging the fault type analysis coefficient and the temperature prediction value based on the preset alarm threshold to obtain a fault warning result.
[0146] The above-mentioned intelligent, safe temperature control and fault warning method for metallurgy first collects heating process data from natural gas heating furnaces, deep-well casting processes, and operating data from processing equipment. Next, the collected heating and casting process data is meticulously processed. Models are then constructed and trained based on the processed data. Advanced algorithms and techniques are then used to perform multiple rounds of iterative optimization, ultimately yielding a trained temperature prediction model. This model accurately calculates temperature predictions based on the processed heating and casting process data using a specific formula. Finally, based on the acquired operating data, a fault type analysis coefficient for the processing equipment during the operating period is calculated. This coefficient, along with the predicted temperature value, is then compared with a preset alarm threshold to generate a fault warning. These steps ensure the quality of metallurgical products and reasonable energy consumption, enabling timely detection of potential faults in processing equipment and early warning of temperature anomalies. This effectively ensures the continuity and safety of metallurgical production, reduces maintenance costs and production risks, and enhances the level of intelligent and automated metallurgical production.
[0147] It should be understood that, although the various steps in the flowcharts involved in the various embodiments described above are displayed in sequence according to the instructions of the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and these steps can be executed in other orders. Moreover, at least a portion of the steps in the flowcharts involved in the various embodiments described above can include multiple steps or multiple stages, and these steps or stages are not necessarily executed and completed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a portion of steps or stages in other steps.
[0148] In one embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of the aforementioned intelligent metallurgical safety temperature control and fault warning system and method are implemented.
[0149] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.
[0150] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to the partial description of the method embodiments. The device embodiments described above are merely illustrative, wherein the components described as separate parts may or may not be physically separated, and the parts displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the disclosed solution. A person of ordinary skill in the art can understand and implement it without expending creative work.
[0151] The above-described embodiments merely represent several implementation methods of the embodiments of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the patent application. It should be noted that a person skilled in the art may make various modifications and improvements without departing from the concept of the embodiments of the present application, and these modifications and improvements fall within the scope of protection of the embodiments of the present application.
Claims
1. An intelligent safety temperature control and fault warning system for metallurgy, characterized in that: The system comprises: The data acquisition module is used to obtain the heating process data of the natural gas heating furnace in the metallurgical process, the deep well casting process data and the operation process data of the processing equipment; a temperature prediction module, configured to construct and train a model based on the processed heating process data and casting process data to obtain a trained temperature prediction model; the trained temperature prediction model is configured to calculate a temperature prediction value based on the processed heating process data and casting process data using a formula; The fault warning module is used to calculate the fault type analysis coefficient of the processing equipment during the operation period based on the operation process data; it is also used to judge the fault type analysis coefficient and the temperature prediction value based on a preset alarm threshold to obtain a fault warning result.
2. The system according to claim 1, wherein: Model construction and training are performed based on the processed heating process data and casting process data to obtain a trained temperature prediction model, including: The processed heating process data and casting process data are randomly divided into a training set and a test set, wherein the training set accounts for 80% and the test set accounts for 20%; Using recursive feature elimination and logistic regression algorithms to perform feature screening on the data in the training set to obtain corresponding key feature factors; A recurrent neural network algorithm is used to construct a model for the key characteristic factors to obtain a pre-trained temperature prediction model; Measuring the difference in data distribution between the training set and the test set based on a loss function of maximum mean difference to obtain a loss function corresponding to the pre-trained temperature prediction model; The back propagation algorithm is used to continuously adjust the parameters so as to minimize the total loss function, thereby obtaining a trained temperature prediction model for predicting the metallurgical temperature.
3. The system according to claim 2, characterized in that The loss function based on the maximum mean difference measures the difference in data distribution between the training set and the test set to obtain the loss function corresponding to the pre-trained temperature prediction model, including: Measuring the difference in data distribution between the training set and the test set using a loss function based on maximum mean difference to obtain a data difference value; The loss function formula of the maximum mean difference is expressed as follows: Among them, MMD 2 (D s , D t ) represents the maximum mean difference between the data in the training set and the test set, φ(*) represents the function that maps the data to the reproducing kernel Hilbert space, D s represents the training set, D t represents the test set, n and m represent the number of data samples in the training set and test set respectively; Calculate the total loss function based on the data difference value combined with the cross entropy loss function of the data in the test set; L=α*MMD 2 (D s ,D t )+(1-α)L ce Among them, L ce Represents the cross entropy loss function, α represents the balance parameter, and its value range is [0, 1].
4. The system according to claim 1, wherein: The step of obtaining the trained temperature prediction model further includes: The fitness objective function is constructed using the following formula based on the key characteristic factors in the temperature prediction model using a linearly weighted multi-objective optimization method: Where, Fitness represents the fitness value; α represents the proportional weight; normal(·) represents the normalization function, and CS represents the key characteristic factor; Initializing the position of the particle in the feasible solution space based on the fitness objective function to obtain the particle position; The particle position of each particle is marked as its corresponding historical optimal position and calculated using the global optimal formula to obtain the global optimal position corresponding to the position of the particle with the minimum fitness value in the particle swarm; The global optimal formula is expressed as follows: Among them, best t represents the global optimal position of the particle swarm, Indicates the historical optimal position corresponding to the current particle position of each particle, represents the fitness value corresponding to the historical optimal position of the i-th particle at the t-th iteration, N represents the size of the particle swarm, and t represents the number of iterations; Based on a preset loop termination condition, a dual-strategy search particle swarm algorithm is used to iteratively calculate the global optimal position to obtain the optimal parameters corresponding to the particle position; the loop termination condition is that the fitness value of the particle is less than a preset temperature prediction error threshold or the number of particle swarm iterations reaches a maximum number of iterations; The temperature prediction model is updated based on the optimal parameters to obtain a trained temperature prediction model.
5. The system according to claim 1, wherein: The temperature prediction model obtains the temperature prediction value by the following method: Inputting each of the key characteristic factors into the trained temperature prediction model to obtain a basic value for metallurgical temperature prediction; Calculate the SHAP value based on each of the key characteristic factors to obtain the SHAP value corresponding to each of the key characteristic factors; The calculation formula of the SHAP value is as follows: in, represents the SHAP value of key feature factor i, S represents the subset of key feature factors, f(S) represents the prediction base value of the model when only the key feature factors in subset S are used, and f(S∪{i}) represents the prediction base value of the model after adding the key feature factors to subset S; Calculate the predicted temperature value of the metallurgical process according to the predicted basic value and the SHAP value corresponding to the key characteristic factor; T pr =T ba +Σ j SHAPvalue j Among them, T pr Represents the predicted metallurgical temperature, T ba Represents the basic value of metallurgical temperature prediction, j represents the key characteristic factor, SHAPvalue j Indicates the SHAP value corresponding to the key feature factor j.
6. The system according to claim 5, characterized in that After obtaining the SHAP value corresponding to each of the key feature factors, the method further includes: Interpreting the SHAP value corresponding to each key characteristic factor based on the data in the test set to obtain the characteristic influence of each key characteristic factor on the temperature prediction model; The feature influence is calculated using the following formula: Among them, Q j It represents the characteristic influence of the jth key characteristic factor on the metallurgical temperature prediction basic value output by the temperature prediction model. Assume that X i,j represents the jth feature of the i-th test set sample, Represents X i,j SHAP value, n represents the total number of test set samples; The temperature prediction model is subjected to performance evaluation using K-fold cross validation based on the feature influence to obtain a final temperature prediction model; the final temperature prediction model is the temperature prediction model with the best performance evaluation result.
7. The system according to claim 1, wherein: The calculation based on the operation process data using a formula to obtain the fault type analysis coefficient of the processing equipment during the operation period includes: Extracting key features from the acquired operation process data to obtain operation feature parameters; The operating characteristic parameters are calculated using the following formula to obtain the fault type analysis coefficient of the processing equipment during the operating period: Among them, E i Indicates the failure type analysis coefficient of the processing equipment during the operation period, PL i Indicates the frequency difference between the instantaneous and slow occurrence of processing equipment failures during the operation period, CS i Indicates the percentage of times that processing equipment continues to operate after operating adjustments after a fault occurs, GH i It represents the probability of replacing equipment parts after a processing equipment failure occurs during the operating period; ρ represents the error correction factor, ω1, ω2 and ω3 all represent preset proportional coefficients, and ω1>ω2>ω3>1.
8. A method for intelligent safety temperature control and fault warning in metallurgy, characterized in that: The method comprises: Acquire natural gas heating furnace heating process data, deep well casting process data, and processing equipment operation process data in metallurgical processes; Model construction and training are performed based on the processed heating process data and casting process data to obtain a trained temperature prediction model; the trained temperature prediction model is used to calculate using a formula based on the processed heating process data and casting process data to obtain a temperature prediction value; The fault type analysis coefficient of the processing equipment during the operation period is calculated based on the operation process data; the fault type analysis coefficient and the temperature prediction value are judged based on a preset alarm threshold to obtain a fault warning result.
9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the system according to any one of claims 1 to 8 is implemented.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the system according to any one of claims 1 to 8 is implemented.