A multi-modal adaptive and predictive control system based on pit furnace ultra-long carburizing process
By using a multimodal adaptive and predictive control system, the strong carburizing stage is subdivided and a multi-stage control strategy is generated. This solves the problems of uneven carburized layer quality and low control accuracy in the carburizing process of a pit furnace, and achieves stable regulation of carbon potential and temperature, thereby improving the process stability and accuracy of the carburizing process.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- JIANGSU FENGDONG THERMAL TECH CO LTD
- Filing Date
- 2025-11-25
- Publication Date
- 2026-06-05
Smart Images
Figure CN121454946B_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of pit furnaces, and in particular to a multimodal adaptive and predictive control system for ultra-long carburizing processes in pit furnaces. Background Technology
[0002] Current technologies only divide carburizing into two coarse-grained stages: intense carburizing and diffusion. They do not consider the need for further subdivision of the stages due to the dynamic changes in the duration of the intense carburizing stage. The intense carburizing stage requires a rapid attainment and stabilization of high carbon potential, while the diffusion stage requires a steady decline in carbon potential. A single control mode cannot meet both requirements simultaneously. Furthermore, a single control mode cannot adapt to the carbon potential requirements of different stages.
[0003] Solving this technical problem is a technical challenge that needs to be overcome by those skilled in the art. Summary of the Invention
[0004] This application provides a multimodal adaptive and predictive control system for an ultra-long carburizing process in a pit furnace, which at least partially solves the above-mentioned technical problems.
[0005] To achieve the above objectives, according to a first aspect of this application, a multimodal adaptive and predictive control system based on an ultra-long carburizing process in a pit furnace is provided, comprising:
[0006] Parameter detection module; the parameter detection module includes a thermocouple and an oxygen probe; the thermocouple is used to collect temperature data inside the pit furnace in real time; the oxygen probe is used to collect carbon potential data inside the pit furnace in real time.
[0007] PLC control unit; the PLC control unit communicates with the parameter detection module and with the enrichment gas solenoid valve and air solenoid valve of the pit furnace;
[0008] The data processing unit calculates the estimated duration of intensive carburizing based on real-time collected temperature and carbon potential data. If the estimated duration exceeds the carburizing time threshold, the intensive carburizing stage is subdivided into early, middle, and late stages. A control strategy is generated for each of the early, middle, late, and diffusion stages. A transition sub-stage is set between adjacent carburizing process stages, and a weighted fusion method is used within the transition sub-stage to achieve smooth switching of the control strategy. If the estimated duration does not exceed the carburizing time threshold, the intensive carburizing stage is not subdivided.
[0009] The PLC control unit determines the opening adjustment parameters of the enrichment gas solenoid valve and the air solenoid valve based on the control strategy sent by the data processing unit.
[0010] By adopting the above technical solution, the furnace temperature and carbon potential data are acquired in real time through the parameter detection module. The data processing unit determines whether to subdivide the strong carburizing stage based on the estimated duration of the strong carburizing and generates appropriate control strategies for each stage of the strong carburizing and diffusion stages. The transition sub-stage between adjacent stages avoids parameter abrupt changes through the weighted fusion of control strategies. Then, the PLC control unit executes the opening adjustment of the solenoid valve to achieve stable control of carbon potential and temperature during the ultra-long carburizing process, ensuring the uniformity of the carburized layer quality. Attached Figure Description
[0011] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0012] Figure 1 This is a schematic diagram of a multimodal adaptive and predictive control system based on an ultra-long carburizing process in a pit furnace, provided in an exemplary embodiment of this application. Detailed Implementation
[0013] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the protection scope of this application.
[0014] This application provides a multimodal adaptive and predictive control system for an ultra-long carburizing process in a pit furnace. Please refer to [link to relevant documentation]. Figure 1 The multimodal adaptive and predictive control system based on an ultra-long carburizing process in a pit furnace provided in this application includes:
[0015] Parameter detection module; the parameter detection module includes thermocouples and oxygen probes; the thermocouples are used to collect temperature data inside the pit furnace in real time; the oxygen probes are used to collect carbon potential data inside the pit furnace in real time.
[0016] PLC control unit; The PLC control unit communicates with the parameter detection module and with the enrichment gas solenoid valve and air solenoid valve of the pit furnace;
[0017] The data processing unit calculates the estimated duration of intensive carburizing based on real-time collected temperature and carbon potential data. If the estimated duration exceeds the carburizing time threshold, the intensive carburizing stage is subdivided into early, middle, and late stages. A control strategy is generated for each of these stages. A transition sub-stage is set between adjacent carburizing process stages, and a weighted fusion method is used within the transition sub-stage to achieve smooth switching of the control strategy. If the estimated duration does not exceed the carburizing time threshold, the intensive carburizing stage is not subdivided.
[0018] The PLC control unit determines the opening adjustment parameters of the enrichment gas solenoid valve and the air solenoid valve based on the control strategy sent by the data processing unit.
[0019] Specifically, the parameter detection module includes a thermocouple and an oxygen probe; the thermocouple is used to collect temperature data in the pit furnace in real time; the oxygen probe is used to collect carbon potential data in the pit furnace in real time. The carbon potential data refers to the activity level of carbon in the carburizing atmosphere in the furnace, and its value directly affects the carbon concentration on the workpiece surface and in the carburized layer.
[0020] The PLC control unit maintains real-time communication with the parameter detection module, receiving temperature and carbon potential data transmitted by the module and forwarding them to the data processing unit. The PLC control unit also communicates with the enrichment gas solenoid valve and air solenoid valve of the pit furnace. It can receive control strategies sent by the data processing unit and determine the opening adjustment parameters of the enrichment gas and air solenoid valves based on these strategies. These opening adjustment parameters are specific values used to control the degree of solenoid valve opening, determining the amount of enrichment gas and air introduced into the furnace, thereby regulating the carbon potential and temperature within the furnace.
[0021] The data processing unit first uses the temperature and carbon potential data collected in real time by the parameter detection module, combined with the carbon diffusion coefficient calculation formula, to obtain the estimated duration of the intensive percolation process. The estimated duration of the intensive percolation process refers to the time required to complete the intensive percolation stage based on the current furnace temperature and carbon potential state, and is used to determine whether the intensive percolation stage needs to be subdivided. When the estimated duration of the intensive percolation process exceeds the carburization time threshold, it indicates that the intensive percolation process is relatively long, and a single control strategy is difficult to adapt to the entire process. At this time, the data processing unit will subdivide the intensive percolation stage into the early stage, the middle stage, and the late stage. If the estimated duration of the intensive percolation process does not exceed the carburization time threshold, the intensive percolation stage is not subdivided, and a single control strategy can meet the control requirements.
[0022] The data processing unit generates a control strategy for each of the early, middle, and late stages of strong carburizing and the diffusion stage. To avoid sudden changes in furnace parameters caused by directly switching control strategies between adjacent carburizing process stages, the data processing unit sets a transition sub-stage between adjacent carburizing process stages. The transition sub-stage refers to the transition period between adjacent process stages. Within the transition sub-stage, the system uses a weighted fusion method to achieve a smooth switching of control strategies, so that the enrichment gas, air flow rate, and temperature adjustment in the furnace transition smoothly and avoid sudden parameter changes.
[0023] Traditional methods do not subdivide the strong permeation stage according to its expected duration. When the expected duration of strong permeation is long, a single control strategy cannot adapt to the changes in carbon diffusion characteristics during the process. This application subdivides the strong permeation stage based on the comparison between the expected duration of strong permeation and the carburization time threshold, and generates adaptive control strategies for each stage. This makes the control strategies more compatible with the process characteristics of each stage, effectively solving the problems of low control accuracy and unstable permeation layer quality during long-term strong permeation.
[0024] In some embodiments, the data processing unit obtains the estimated duration of strong infiltration based on real-time acquired temperature data and carbon potential data, including:
[0025] The carbon diffusion coefficient during the strong infiltration stage was calculated based on the acquired temperature and carbon potential data.
[0026] Based on the carbon diffusion coefficient and the preset carbon concentration gradient, the expected duration of the strong infiltration stage is calculated and denoted as the expected duration of strong infiltration.
[0027] Specifically, the carbon diffusion coefficient refers to the rate parameter of carbon atom diffusion in the carburizing atmosphere of the furnace. Its value directly reflects the speed at which carbon atoms penetrate into the workpiece, and in actual calculations, it is often calculated based on the Arrhenius equation. After obtaining the carbon diffusion coefficient, the data processing unit combines it with a preset carbon concentration gradient and calculates the time required to complete the strong carburizing stage using carbon diffusion equations, such as Fick's second law. This time is the estimated duration of the strong carburizing stage.
[0028] In some embodiments, the strong infiltration stage is further subdivided into a pre-strong infiltration stage, a mid-strong infiltration stage, and a post-strong infiltration stage, specifically including:
[0029] Real-time acquisition of carbon potential data; construction of a carbon potential change sequence within a sliding time window.
[0030] The carbon potential growth rate and carbon potential growth acceleration are obtained by calculating the first and second derivatives based on the carbon potential change sequence.
[0031] Determine the current carburizing dynamic stage based on the carbon potential growth rate and carbon potential acceleration:
[0032] If the carbon potential growth rate continues to rise and the carbon potential growth acceleration is greater than the first preset threshold, it is determined to be the early stage of strong infiltration, and the first control strategy is executed.
[0033] If the fluctuation of the carbon potential growth rate is less than the preset fluctuation value, it is determined to be the mid-stage of strong infiltration, and the second control strategy is executed.
[0034] If the carbon potential growth rate decreases and the carbon potential growth acceleration is negative and the absolute value exceeds the third preset threshold, it is determined to be the late stage of strong infiltration, and the third control strategy is executed.
[0035] Specifically, the carbon potential change sequence refers to the data set formed by arranging real-time collected carbon potential data in chronological order within a sliding time window. The carbon potential growth rate refers to the amount of change in carbon potential per unit time calculated based on the carbon potential change sequence, used to quantify the speed of carbon potential change. The carbon potential growth acceleration refers to the amount of change in the carbon potential growth rate per unit time calculated based on the carbon potential growth rate, used to quantify the trend of the carbon potential growth rate. The preset fluctuation value refers to the range of carbon potential growth rate fluctuations pre-set according to the process target of carbon potential stabilization in the mid-stage of strong infiltration; the third preset threshold refers to the critical value of the absolute value of carbon potential growth acceleration pre-set according to the process target of a slow decline in carbon potential in the later stage of strong infiltration. When the acceleration is negative and the absolute value exceeds this value, it indicates that the carbon potential growth rate is rapidly declining, i.e., the carbon potential is beginning to fall.
[0036] Specifically, in traditional methods, if the carbon potential increases slowly during a batch of carburizing due to fluctuations in the enrichment gas supply, continuing to implement a stable carbon potential strategy after entering the mid-stage of strong carburizing at a fixed time will result in the final carbon concentration not meeting the standard. Conversely, if the carbon potential reaches a stable state prematurely, but a rapid carbon increase strategy is still implemented at a fixed time, it will lead to excessively high carbon potential and cause embrittlement of the carburized layer. This application solves the problem of the disconnect between the traditional fixed-time division and the actual dynamic carbon potential, reduces the inaccuracy of carbon potential control caused by misjudgment of stages, improves the uniformity of carbon concentration distribution in the carburized layer, and ensures the process stability of ultra-long carburizing processes.
[0037] In some embodiments, the generation of a first control strategy in the early stage of strong infiltration specifically includes:
[0038] Obtain current temperature and carbon potential data;
[0039] Calculate the first carbon potential deviation between the current carbon potential data and the first target carbon potential, and the first temperature deviation between the current temperature data and the first target temperature;
[0040] A fuzzy PID controller is constructed, taking the first carbon potential deviation, the first temperature deviation, the time change rate of the first carbon potential deviation, and the time change rate of the first temperature deviation as inputs, and generating the initial control quantity through preset fuzzy rules.
[0041] The lag time of carbon potential transfer is calculated based on the furnace depth and airflow velocity; the first control strategy is obtained by anticipating the lag time and performing advance compensation on the initial control quantity; the first control strategy includes the opening degree of the enrichment gas valve, the opening degree of the air valve, and the adjustment time.
[0042] Specifically, the data processing unit acquires the current temperature and carbon potential data during the early stage of strong infiltration through the parameter detection module, and compares them with the preset first target temperature and first target carbon potential to calculate the first temperature deviation and the first carbon potential deviation. Based on historical deviation data within the sliding time window, it calculates the time change rate of the first carbon potential deviation and the time change rate of the first temperature deviation. The data processing unit constructs a fuzzy PID controller, taking four parameters—the first carbon potential deviation, the first temperature deviation, and their corresponding time change rates—as input. The controller performs fuzzy inference on the input parameters through preset fuzzy rules, transforming them into preliminary control quantities such as the adjustment amount of the enrichment gas valve opening and the adjustment amount of the air valve opening. Compared with traditional PID, fuzzy PID can better adapt to the nonlinear changes in carbon potential and temperature during the early stage of strong infiltration. Based on the furnace depth and airflow velocity of the pit furnace, the data processing unit can calculate the lag time of carbon potential transfer using the formula: lag time = furnace depth / airflow velocity × correction coefficient. Based on this lag time, it performs advance compensation on the preliminary control quantities; for example, if the lag time is 3 seconds, the execution time of the preliminary control quantities is advanced by 3 seconds.
[0043] Existing technologies employing single PID control struggle to adapt to the nonlinear correlation between carbon potential and temperature during the early stages of intense permeation, leading to oscillations in regulation. Furthermore, neglecting the hysteresis characteristics of carbon potential transfer in pit furnaces results in control actions lagging behind carbon potential changes, prolonging the time it takes for the carbon potential to reach the target value. This proposed solution, however, utilizes a fuzzy PID controller. By employing preset fuzzy rules to convert deviations and rates of change into suitable initial control quantities, it effectively addresses the nonlinear conditions during the early stages of intense permeation, reducing regulation oscillations. By calculating the hysteresis duration based on furnace depth and airflow velocity and performing advance compensation, it can correct the control quantity in advance to offset the hysteresis effect, preventing carbon potential overshoot or slow response.
[0044] In some embodiments, a second control strategy is generated during the mid-stage of strong infiltration, specifically including:
[0045] Obtain the second target carbon potential and second target temperature during the mid-stage of strong infiltration; set the allowable range of second carbon potential fluctuation based on the second target carbon potential; set the allowable range of second temperature fluctuation based on the second target temperature;
[0046] Calculate the second carbon potential deviation between the current carbon potential data and the second target carbon potential, and the second temperature deviation between the current temperature data and the second target temperature;
[0047] Identify the nonlinear characteristics inside the pit furnace, including: the sensitivity parameter of carbon potential to enriched gas concentration, the response delay parameter of carbon potential to temperature change, and the coupling strength parameter of temperature and carbon potential.
[0048] Based on the identification results of nonlinear characteristics and the second carbon potential deviation and second temperature deviation, the PID controller parameters and heating element power are adjusted.
[0049] A second control strategy is generated based on the adjusted PID control parameters. The second control strategy includes the real-time opening degree of the enrichment gas valve, the real-time opening degree of the air valve, and the power adjustment parameters of the heating element.
[0050] The process of adjusting the PID controller parameters and heating element power based on the identification results of nonlinear characteristics, the second carbon potential deviation, and the second temperature deviation includes:
[0051] When the deviation of the second carbon potential is within the allowable range of the second carbon potential fluctuation and the sensitivity parameter of the carbon potential to the enrichment gas concentration increases, the proportional coefficient is reduced according to the preset rules; when the deviation of the second carbon potential exceeds the allowable range and the sensitivity parameter of the carbon potential to the enrichment gas concentration decreases, the proportional coefficient is increased according to the preset rules.
[0052] When the second temperature deviation is within the allowable range of the second temperature fluctuation and the response delay parameter of the carbon potential to temperature change increases, the integral coefficient is increased according to the preset rule to enhance the system's ability to compensate for hysteresis characteristics.
[0053] When the second temperature deviation exceeds the allowable range of the second temperature fluctuation and the coupling strength parameter between temperature and carbon potential increases, the differential coefficient is increased according to the preset rule to weaken the overshoot caused by coupling interference.
[0054] When the second temperature deviation exceeds the second temperature fluctuation allowable range, the power of the heating element is adjusted by the PLC control unit to bring the temperature back to the second temperature fluctuation allowable range;
[0055] If the temperature is higher than the second target temperature and the second temperature deviation exceeds the allowable range, the air valve opening will be increased by a preset ratio to suppress excessive carbon potential; if the temperature is lower than the second target temperature and the second temperature deviation exceeds the allowable range, the enrichment gas valve opening will be increased by a preset ratio to compensate for insufficient carbon potential.
[0056] Specifically, this solution solves the problem of poor adaptability of fixed parameters by identifying nonlinear characteristics online, enabling PID parameters to be adjusted according to changes in sensitivity, response delay, and coupling strength; it avoids control inaccuracies caused by lag and coupling by compensating for response delay with integral coefficients and weakening coupling interference with derivative coefficients; and it achieves temperature-carbon decoupling by compensating for valve opening during temperature deviations, reducing the impact of temperature fluctuations on carbon potential.
[0057] In some embodiments, identifying nonlinear characteristics within a pit furnace includes:
[0058] The input dataset includes real-time carbon potential data, second target carbon potential, second carbon potential deviation, real-time temperature data, second target temperature, second temperature deviation, real-time opening data of enriched gas solenoid valve, and real-time opening data of air solenoid valve.
[0059] The input dataset is fed into a pre-trained identification model, which includes an input layer, hidden layers, and an output layer. In the input layer, an input node is set for each of the following parameters: real-time carbon potential, second target carbon potential, second carbon potential deviation, real-time temperature, second target temperature, second temperature deviation, enrichment gas valve opening, and air valve opening. In the hidden layer, a preset number of hidden layer nodes are set, and a nonlinear activation function is used to fit the complex nonlinear relationships within the furnace. The output layer corresponds to the nonlinear characteristic parameters within the furnace to be identified, including the sensitivity parameter of carbon potential to enrichment gas concentration, the response delay parameter of carbon potential to temperature changes, and the coupling strength parameter between temperature and carbon potential. Each nonlinear characteristic parameter corresponds to an output node. A sliding time window is used to select historical input data as training samples. A loss function is defined to quantify the deviation between the identification results output by the identification model and the actual nonlinear characteristics within the furnace.
[0060] The identification model outputs nonlinear characteristic identification results, which include the sensitivity parameter of carbon potential to enriched gas concentration, the response delay parameter of carbon potential to temperature change, and the coupling strength parameter between temperature and carbon potential.
[0061] Specifically, the nonlinear characteristics inside the furnace during the mid-stage of strong infiltration change dynamically with the operating conditions. This application uses a neural network with a nonlinear activation function to fit the complex nonlinear characteristics inside the furnace.
[0062] In some embodiments, the third control strategy for generating strong infiltration in the later stages specifically includes:
[0063] Obtain the third target carbon potential in the later stage of strong infiltration; the third target carbon potential is the initial carbon potential in the diffusion stage.
[0064] Set a carbon potential decrease curve in the later stage of strong infiltration. The carbon potential decrease curve starts from the current carbon potential and ends at the third target carbon potential. The slope of the carbon potential decrease curve corresponds to the preset carbon potential decrease rate.
[0065] Calculate the third carbon potential deviation between the current carbon potential and the target carbon potential at the corresponding moment of the decline curve;
[0066] A predictive control model is constructed. The inputs of the predictive control model include the third carbon potential deviation, real-time temperature data and historical carbon potential change sequence, and the prediction step size of the model is preset based on the in-furnace heat and mass transfer characteristics.
[0067] The predictive control model calculates the carbon potential change trend within a preset time period; using the third target carbon potential as a constraint, it generates a decreasing adjustment parameter for the opening of the enrichment gas valve as an output; when the carbon potential fluctuation amplitude of the carbon potential change trend exceeds the preset third fluctuation threshold, it generates an air valve opening adjustment parameter to suppress the fluctuation.
[0068] A third control strategy is generated based on the output of the predictive control model. The third control strategy includes the real-time decreasing opening of the enrichment gas valve and the real-time adjusting opening of the air valve.
[0069] Specifically, the third target carbon potential refers to the preset final carbon potential target value in the later stage of strong permeation, which is equal to the initial carbon potential in the diffusion stage. The carbon potential decline curve refers to the curve describing the change of carbon potential over time in the later stage of strong permeation. It starts with the current carbon potential at the beginning of the later stage of strong permeation and ends with the third target carbon potential. The shape of the curve is determined by the preset carbon potential decline rate, used to control the path of carbon potential decline and avoid excessively fast or slow carbon reduction. The aforementioned predictive control model is a control model based on future operating condition prediction. By analyzing current deviations, real-time data, and historical trends, it predicts parameter changes over a future period and generates control strategies in advance to offset potential deviations, making it suitable for the later stage of strong permeation. The prediction step size refers to the length of the future prediction period set in the predictive control model, and its value is determined based on the heat and mass transfer characteristics in the furnace, such as the temperature response rate. The decreasing adjustment parameter for the enrichment gas valve opening refers to the parameter output by the predictive control model used to gradually reduce the enrichment gas injection rate. Its value matches the slope of the carbon potential decline curve, ensuring that the enrichment gas reduction rate is consistent with the carbon potential decline rate. The air valve opening adjustment parameter refers to the amount of air valve opening adjustment output by the model when the carbon potential fluctuation exceeds the third fluctuation threshold. This adjustment is used to quickly suppress carbon potential fluctuations and maintain the stability of the descent curve. The real-time decreasing opening of the enrichment gas valve refers to the real-time opening degree of the enrichment gas valve in the third control strategy, which gradually decreases over time according to the decreasing adjustment parameter.
[0070] This application ensures that the rate of carbon potential decline matches the diffusion demand by pre-setting a carbon potential decline curve; the predictive control model adjusts the enrichment gas opening in advance by predicting future carbon potential change trends, thus offsetting the adjustment delay caused by hysteresis; the real-time adjustment parameters of the air valve can quickly suppress carbon potential fluctuations exceeding the third fluctuation threshold, achieving a seamless connection between the late stage of strong infiltration and the diffusion stage.
[0071] In some embodiments, a predictive control model is constructed to calculate the carbon potential change trend within a preset future time period. Using a third target carbon potential as a constraint, a decreasing adjustment parameter for the opening of the enrichment gas valve is generated as the output, including:
[0072] The time-domain features of the third carbon potential deviation are extracted to obtain the deviation amplitude and the cumulative deviation, which are denoted as the third carbon potential deviation features; the real-time temperature data are statistically analyzed to generate the temperature fluctuation coefficient; the historical carbon potential change sequence is trend-fitted to extract the historical carbon potential change rate and the change rate fluctuation rate, which are denoted as the historical carbon potential change features.
[0073] The structure of the predictive control model is designed. The predictive control model includes a mechanism layer and a data correction layer. The mechanism layer is based on the in-furnace carbon transfer equation to build a basic predictive model. The input parameters of the mechanism layer include enriched gas concentration, temperature, and carbon diffusion coefficient. The data correction layer is based on the third carbon potential deviation characteristics, temperature fluctuation coefficient, and carbon potential historical change characteristics to train a BP neural network and perform error compensation on the carbon potential prediction value output by the mechanism layer.
[0074] Input the current enrichment gas valve opening, real-time temperature data, third carbon potential deviation characteristics, temperature fluctuation coefficient, and historical carbon potential change characteristics into the predictive control model to obtain the initial carbon potential change curve within a preset future time period.
[0075] Construct an optimization objective function constrained by the third target carbon potential; the objective function has the following constraints: the total deviation between the predicted carbon potential value and the carbon potential decline curve within the future preset time period is minimized, the rate of change of carbon potential does not exceed the preset maximum decline rate, the change in the opening of the enrichment gas valve does not exceed the preset single adjustment upper limit, and the third target carbon potential is used as the terminal constraint.
[0076] The objective function is solved using a rolling optimization algorithm to obtain the sequence of enriched gas valve openings within the future prediction step. The first element of the enriched gas valve opening sequence is taken as the current enriched gas valve opening reduction adjustment parameter.
[0077] Specifically, the predictive control model predicts the carbon potential change trend through the mechanism layer and data fusion layer, and generates enrichment gas valve adjustment parameters under multiple constraints by combining the rolling optimization algorithm. This effectively solves the problems of lag in carbon potential decline, large fluctuations, and inaccurate connection with the diffusion stage in traditional control, and realizes a smooth decline of carbon potential along the preset curve in the later stage of strong carburization, thereby improving the continuity of the carburization process and the accuracy of carbon potential control.
[0078] In some embodiments, a transition sub-stage is provided between adjacent carburizing process stages, and a smooth switching of the control strategy is achieved by weighted fusion of the control strategies of adjacent stages within the transition sub-stage, including:
[0079] Extract the control strategy parameters of two adjacent stages to form a set of control parameters for adjacent stages. The control strategy parameters include the opening degree of the enrichment gas valve and the opening degree of the air valve. If the adjacent stage includes a strong infiltration mid-term, the control strategy parameters also include the heating element power adjustment parameters.
[0080] Calculate the difference between corresponding parameters in the control parameter sets of adjacent stages to form a parameter difference dataset; if the difference of any parameter in the parameter difference dataset exceeds the preset parameter difference threshold, the mean filtering algorithm is used to smooth the parameter difference dataset to obtain a smoothed difference dataset; if the differences do not exceed the limit, the parameter difference dataset is directly used as the smoothed difference dataset.
[0081] Obtain the preset duration of the transition sub-stage; the duration of the transition sub-stage is preset based on the process characteristics of the adjacent stages;
[0082] Based on the smoothed difference dataset and the duration of the transition sub-stage, the initial value of the weighted fusion coefficient is calculated; the control parameter sets of adjacent stages are weighted and calculated using the initial value of the weighted fusion coefficient to obtain the preliminary control parameter set of the transition sub-stage.
[0083] Calculate the deviation between the actual carbon potential corresponding to the carbon potential adjustment parameters in the initial control parameter set and the target carbon potential for the transition stage; if the deviation is greater than the preset carbon potential deviation threshold, optimize the carbon potential adjustment parameters using a linear regression algorithm to obtain the optimized carbon potential adjustment parameter set.
[0084] The optimized carbon potential adjustment parameter set is fused with the preliminary control parameter set according to a preset weight to obtain the transition stage parameter set; the transition stage parameter set is the control parameter set for the transition sub-stage.
[0085] The data processing unit sends the transition phase parameters to the PLC control unit to achieve a smooth switch between control strategies for adjacent phases.
[0086] Specifically, when switching control strategies directly between adjacent stages, the difference in parameters between the two stages can easily lead to drastic fluctuations in carbon potential and temperature within the furnace. This application avoids abrupt parameter changes by using differential smoothing to process the control parameters; the linearly changing weighted fusion coefficient ensures a smooth parameter transition. Ultimately, this improves the process stability of the ultra-long carburizing process.
[0087] It should be noted that the multimodal adaptive and predictive control system based on the ultra-long carburizing process of a pit furnace provided in this embodiment of the invention is used to execute all process steps of the multimodal adaptive and predictive control system based on the ultra-long carburizing process of a pit furnace in the above embodiment. The working principles and beneficial effects of the two are one-to-one, so they will not be described again.
[0088] In the description of this application, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more features. In the description of this application, "multiple" means two or more, unless otherwise explicitly specified.
[0089] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.
[0090] The embodiments, implementation methods, and related technical features of this application can be combined and substituted for each other without conflict.
[0091] The above are merely preferred embodiments of this application and are not intended to limit this application in any way. Any simple modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of this application without departing from the scope of the technical solution of this application shall still fall within the scope of the technical solution of this application.
Claims
1. A multimodal adaptive and predictive control system for an ultra-long carburizing process in a pit furnace, characterized in that, include: Parameter detection module; the parameter detection module includes a thermocouple and an oxygen probe; The thermocouple is used to collect temperature data inside the pit furnace in real time. The oxygen probe is used to collect carbon potential data in the pit furnace in real time. PLC control unit; the PLC control unit communicates with the parameter detection module and with the enrichment gas solenoid valve and air solenoid valve of the pit furnace; The data processing unit calculates the estimated duration of intensive carburizing based on real-time collected temperature and carbon potential data. If the estimated duration exceeds the carburizing time threshold, the intensive carburizing stage is subdivided into early, middle, and late stages. A control strategy is generated for each of the early, middle, late, and diffusion stages. A transition sub-stage is set between adjacent carburizing process stages, and a weighted fusion method is used within the transition sub-stage to achieve smooth switching of the control strategy. If the estimated duration does not exceed the carburizing time threshold, the intensive carburizing stage is not subdivided. The PLC control unit determines the opening adjustment parameters of the enrichment gas solenoid valve and the air solenoid valve based on the control strategy sent by the data processing unit. The strong infiltration stage is further divided into the early stage, the middle stage, and the late stage, specifically including: Real-time acquisition of carbon potential data; construction of a carbon potential change sequence within a sliding time window. The carbon potential growth rate and carbon potential growth acceleration are obtained by calculating the first and second derivatives based on the carbon potential change sequence. Determine the current carburizing dynamic stage based on the carbon potential growth rate and carbon potential acceleration: If the carbon potential growth rate continues to rise and the carbon potential growth acceleration is greater than the first preset threshold, it is determined to be the early stage of strong infiltration, and the first control strategy is executed. If the fluctuation of the carbon potential growth rate is less than the preset fluctuation value, it is determined to be the mid-stage of strong infiltration, and the second control strategy is executed. If the carbon potential growth rate decreases and the carbon potential growth acceleration is negative and the absolute value exceeds the third preset threshold, it is determined to be the late stage of strong infiltration, and the third control strategy is executed.
2. The system according to claim 1, characterized in that, The data processing unit obtains the estimated duration of strong infiltration based on real-time collected temperature and carbon potential data, including: The carbon diffusion coefficient during the strong infiltration stage was calculated based on the acquired temperature and carbon potential data. Based on the carbon diffusion coefficient and the preset carbon concentration gradient, the expected duration of the strong infiltration stage is calculated and denoted as the expected duration of strong infiltration.
3. The system according to claim 2, characterized in that, The first control strategy for generating the strong infiltration phase specifically includes: Obtain current temperature and carbon potential data; Calculate the first carbon potential deviation between the current carbon potential data and the first target carbon potential, and the first temperature deviation between the current temperature data and the first target temperature; A fuzzy PID controller is constructed, taking the first carbon potential deviation, the first temperature deviation, the time change rate of the first carbon potential deviation, and the time change rate of the first temperature deviation as inputs, and generating the initial control quantity through preset fuzzy rules. The lag time for carbon potential transfer is calculated based on the furnace depth and airflow velocity; a first control strategy is obtained by performing advance compensation on the preliminary control quantity based on the lag time; the first control strategy includes the opening degree of the enrichment gas valve, the opening degree of the air valve, and the adjustment time.
4. The system according to claim 3, characterized in that, The second control strategy for generating the strong infiltration mid-term specifically includes: Obtain the second target carbon potential and second target temperature during the mid-stage of strong infiltration; set the allowable range of second carbon potential fluctuation based on the second target carbon potential; set the allowable range of second temperature fluctuation based on the second target temperature; Calculate the second carbon potential deviation between the current carbon potential data and the second target carbon potential, and the second temperature deviation between the current temperature data and the second target temperature; Identify the nonlinear characteristics inside the pit furnace, including: the sensitivity parameter of carbon potential to enriched gas concentration, the response delay parameter of carbon potential to temperature change, and the coupling strength parameter of temperature and carbon potential. Based on the identification results of the nonlinear characteristics and the second carbon potential deviation and the second temperature deviation, the PID controller parameters and the heating element power are adjusted. A second control strategy is generated based on the adjusted PID control parameters. The second control strategy includes the real-time opening degree of the enrichment gas valve, the real-time opening degree of the air valve, and the power adjustment parameters of the heating element. The adjustment of PID controller parameters and heating element power based on the identification results of the nonlinear characteristics, the second carbon potential deviation, and the second temperature deviation includes: When the deviation of the second carbon potential is within the allowable range of the second carbon potential fluctuation and the sensitivity parameter of the carbon potential to the enrichment gas concentration increases, the proportional coefficient is reduced according to the preset rules; when the deviation of the second carbon potential exceeds the allowable range and the sensitivity parameter of the carbon potential to the enrichment gas concentration decreases, the proportional coefficient is increased according to the preset rules. When the second temperature deviation is within the allowable range of the second temperature fluctuation and the response delay parameter of the carbon potential to temperature change increases, the integral coefficient is increased according to the preset rule to enhance the system's ability to compensate for hysteresis characteristics. When the second temperature deviation exceeds the allowable range of the second temperature fluctuation and the coupling strength parameter between temperature and carbon potential increases, the differential coefficient is increased according to the preset rule to weaken the overshoot caused by coupling interference. When the second temperature deviation exceeds the second temperature fluctuation allowable range, the power of the heating element is adjusted by the PLC control unit to bring the temperature back to the second temperature fluctuation allowable range; If the temperature is higher than the second target temperature and the second temperature deviation exceeds the allowable range, the air valve opening will be increased by a preset ratio to suppress excessive carbon potential; if the temperature is lower than the second target temperature and the second temperature deviation exceeds the allowable range, the enrichment gas valve opening will be increased by a preset ratio to compensate for insufficient carbon potential.
5. The system according to claim 4, characterized in that, Identifying the nonlinear characteristics within a pit furnace, including: The input dataset includes real-time carbon potential data, second target carbon potential, second carbon potential deviation, real-time temperature data, second target temperature, second temperature deviation, real-time opening data of enriched gas solenoid valve, and real-time opening data of air solenoid valve. The input dataset is fed into a pre-trained identification model, which includes an input layer, a hidden layer, and an output layer. In the input layer, an input node is set for each of the following parameters: real-time carbon potential, second target carbon potential, second carbon potential deviation, real-time temperature, second target temperature, second temperature deviation, enrichment gas valve opening, and air valve opening. In the hidden layer, a preset number of hidden layer nodes are set, and a nonlinear activation function is used to fit the complex nonlinear relationships within the furnace. The output layer corresponds to the nonlinear characteristic parameters within the furnace to be identified. These nonlinear characteristic parameters include the sensitivity parameter of carbon potential to enrichment gas concentration, the response delay parameter of carbon potential to temperature changes, and the coupling strength parameter between temperature and carbon potential. Each nonlinear characteristic parameter corresponds to an output node. A sliding time window is used to select historical input data as training samples. A loss function is defined to quantify the deviation between the identification results output by the identification model and the actual nonlinear characteristics within the furnace. The identification model outputs nonlinear characteristic identification results, which include the sensitivity parameter of carbon potential to enriched gas concentration, the response delay parameter of carbon potential to temperature change, and the coupling strength parameter between temperature and carbon potential.
6. The system according to claim 5, characterized in that, The third control strategy for the generation of the strong infiltration stage specifically includes: Obtain the third target carbon potential in the later stage of strong infiltration; the third target carbon potential is the initial carbon potential in the diffusion stage; A carbon potential decrease curve is set in the later stage of strong infiltration. The carbon potential decrease curve starts from the current carbon potential and ends at the third target carbon potential. The slope of the carbon potential decrease curve corresponds to the preset carbon potential decrease rate. Calculate the third carbon potential deviation between the current carbon potential and the target carbon potential at the corresponding moment of the decline curve; A predictive control model is constructed. The inputs of the predictive control model include the third carbon potential deviation, real-time temperature data and historical carbon potential change sequence, and the prediction step size of the model is preset based on the in-furnace heat and mass transfer characteristics. The predictive control model calculates the carbon potential change trend within a preset time period; using the third target carbon potential as a constraint, it generates a decreasing adjustment parameter for the opening of the enrichment gas valve as an output; when the carbon potential fluctuation amplitude of the carbon potential change trend exceeds the preset third fluctuation threshold, it generates an air valve opening adjustment parameter to suppress the fluctuation. A third control strategy is generated based on the output of the predictive control model. The third control strategy includes the real-time decreasing opening of the enrichment gas valve and the real-time adjusting opening of the air valve.
7. The system according to claim 6, characterized in that, A predictive control model is constructed to calculate the carbon potential change trend within a preset future time period. Using the third target carbon potential as a constraint, a decreasing adjustment parameter for the enrichment gas valve opening is generated as the output, including: The time-domain features of the third carbon potential deviation are extracted to obtain the deviation amplitude and the cumulative deviation, which are denoted as the third carbon potential deviation features; the real-time temperature data are statistically analyzed to generate the temperature fluctuation coefficient; the historical carbon potential change sequence is trend-fitted to extract the historical carbon potential change rate and the change rate fluctuation rate, which are denoted as the historical carbon potential change features. The structure of the predictive control model is designed; the predictive control model includes a mechanism layer and a data correction layer: the mechanism layer constructs a basic predictive model based on the in-furnace carbon transport equation, and the input parameters of the mechanism layer include enriched gas concentration, temperature, and carbon diffusion coefficient; the data correction layer trains a BP neural network based on the third carbon potential deviation feature, temperature fluctuation coefficient, and carbon potential historical change feature to compensate for the error of the carbon potential prediction value output by the mechanism layer. Input the current enrichment gas valve opening, real-time temperature data, third carbon potential deviation characteristics, temperature fluctuation coefficient, and historical carbon potential change characteristics into the predictive control model to obtain the initial carbon potential change curve within a preset future time period. An optimization objective function is constructed with the third target carbon potential as a constraint. The objective function has the following constraints: the total deviation between the predicted carbon potential value and the carbon potential decline curve within a preset future time period is minimized; the rate of change of carbon potential does not exceed the preset maximum decline rate; the change in the opening of the enrichment gas valve does not exceed the preset single adjustment upper limit; and the third target carbon potential is used as the terminal constraint. The objective function is solved using a rolling optimization algorithm to obtain the sequence of enriched gas valve openings within the future prediction step. The first element of the enriched gas valve opening sequence is taken as the current enriched gas valve opening reduction adjustment parameter.
8. The system according to claim 7, characterized in that, A transition sub-stage is set between adjacent carburizing process stages. Within the transition sub-stage, a weighted fusion of the control strategies of adjacent stages is used to achieve a smooth switching of the control strategy, including: Extract control strategy parameters from two adjacent stages to form a set of control parameters for adjacent stages; the control strategy parameters include the opening degree of enrichment gas valve and the opening degree of air valve. If the adjacent stage includes a strong infiltration mid-term, the control strategy parameters also include heating element power adjustment parameters. Calculate the difference between corresponding parameters in the control parameter sets of adjacent stages to form a parameter difference dataset; if the difference of any parameter in the parameter difference dataset exceeds the preset parameter difference threshold, the mean filtering algorithm is used to smooth the parameter difference dataset to obtain a smoothed difference dataset; if the differences do not exceed the limit, the parameter difference dataset is directly used as the smoothed difference dataset. Obtain the preset duration of the transition sub-stage; the duration of the transition sub-stage is preset based on the process characteristics of adjacent stages; Based on the smoothed difference dataset and the duration of the transition sub-stage, the initial value of the weighted fusion coefficient is calculated; the control parameter sets of adjacent stages are weighted and calculated using the initial value of the weighted fusion coefficient to obtain the preliminary control parameter set of the transition sub-stage. Calculate the deviation between the actual carbon potential corresponding to the carbon potential adjustment parameters in the initial control parameter set and the target carbon potential of the transition sub-stage; if the deviation value is greater than the preset carbon potential deviation threshold, optimize the carbon potential adjustment parameters through a linear regression algorithm to obtain an optimized carbon potential adjustment parameter set; The optimized carbon potential adjustment parameter set and the preliminary control parameter set are fused together according to a preset weight to obtain the transition sub-stage parameter set; the transition sub-stage parameter set is the control parameter for the transition sub-stage. The data processing unit sends the parameter set of the transition sub-stage to the PLC control unit to achieve a smooth switching of control strategies between adjacent stages.