Intelligent molten iron casting system and method

Abstract

The invention provides an intelligent molten iron casting system and method, and the method comprises the steps: adjusting the rotating speed of a tipping motor according to a deviation value, and enabling a tipping angle to tend to a target value; the distribution state of the molten iron in the mold is monitored, and a molten iron flowing temperature distribution diagram is generated; analyzing defects according to the molten iron flowing temperature distribution diagram to generate a defect distribution diagram; tipping parameters are adjusted according to the defect distribution diagram, and a molten iron flowing path is optimized; the relative positions of the ladle and the exhaust fume collecting hood are monitored, and a distance change curve is generated; the position of the exhaust fume collecting hood is adjusted according to the distance change curve, and dynamic sealing connection is achieved; monitoring the flue gas concentration in the exhaust fume collecting hood, and generating a concentration change data set; a negative pressure system is adjusted according to the concentration change data set, and the flue gas capture efficiency is enhanced; recording the movement track of the floor mold vehicle, and generating a displacement change sequence; adjusting the position of the exhaust fume collecting hood according to the displacement change sequence, and keeping synchronous movement; and monitoring the leakage condition of the dynamic sealing area, and generating a leakage intensity data set.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of information technology, and in particular to an intelligent molten iron pouring system and method. BACKGROUND

[0002] The molten iron ladle tilting process is a key link in steel smelting, involving multiple complex technical challenges. The primary problem is how to achieve precise control of the tilting angle and speed to ensure the stability and uniformity of molten iron flow. This requires overcoming the limitations of traditional control methods in dynamic environments, particularly the accuracy and reliability issues under high temperature and high pressure conditions. At the same time, defects such as cold shut and gas hole that may occur during molten iron flow seriously affect the quality of castings. How to monitor and prevent these defects in real time during the tilting process becomes a technical problem that needs to be solved urgently. In addition, the molten iron ladle tilting process produces a large amount of high-temperature flue gas, causing serious environmental pollution. How to effectively capture and treat these flue gases while not affecting the flexibility of tilting operations constitutes a difficult technical contradiction. The solution to these problems involves multiple disciplines such as mechanical control, thermodynamics, and fluid mechanics. It also needs to consider the complexity and variability of the actual production environment. How to organically combine these factors to achieve intelligent, precise, and environmentally friendly molten iron ladle tilting process is a comprehensive technical challenge. SUMMARY

[0003] The present application provides an intelligent molten iron pouring system and method, mainly including: Through the sensor network, the angle information during the molten iron ladle tilting process is collected to generate a dynamic angle data set. The deviation value of the tilting angle from the target angle is calculated based on the dynamic angle data set to determine the tilting speed adjustment direction. The tilting motor speed is adjusted based on the deviation value to make the tilting angle tend towards the target value. The distribution state of molten iron in the mold is monitored to generate a molten iron flow temperature distribution map. Defects are analyzed based on the molten iron flow temperature distribution map to generate a defect distribution map. The tilting parameters are adjusted based on the defect distribution map to optimize the molten iron flow path. The relative position of the molten iron ladle and the smoke collection hood is monitored to generate a distance change curve. The smoke collection hood position is adjusted based on the distance change curve to achieve dynamic sealing connection. The smoke concentration in the smoke collection hood is monitored to generate a concentration change data set. The negative pressure system is adjusted based on the concentration change data set to enhance the smoke capture efficiency. The displacement change sequence is generated by recording the movement trajectory of the mold car. The smoke collection hood position is adjusted based on the displacement change sequence to maintain synchronous movement. The leakage condition of the dynamic sealing area is monitored to generate a leakage intensity data set. The sealing device pressure is adjusted based on the leakage intensity data set to optimize the sealing effect.

[0004] Further, the angle information in the tilting process of the ladle is collected through the sensor network to generate a dynamic angle data set, including: obtaining angle data in the tilting process of the ladle through the sensor network, and storing as an initial dynamic data set; if the collection frequency of the initial dynamic data set is lower than a preset threshold, supplementing missing data through an interpolation algorithm to obtain a completed data set; for the completed data set, a time series analysis algorithm is used to detect the change trend of the angle data to determine an abnormal point set; according to the abnormal point set, a clustering algorithm is used to group abnormal data to obtain an abnormal data classification result; deviation features are extracted from the abnormal data classification result, a mean filter algorithm is used to smooth the deviation data, and a smoothed deviation data set is obtained; for the smoothed deviation data set, if the deviation value exceeds a preset threshold, it is marked as an abnormal deviation to obtain an abnormal deviation marking set; a dynamic angle deviation analysis result is generated through the abnormal deviation marking set, and a final deviation data set is output.

[0005] Further, the deviation value between the tilting angle and the target angle is calculated according to the dynamic angle data set to determine the tilting speed adjustment direction, including: current tilting angle data is extracted from the dynamic angle data set, compared with a preset target angle, and the deviation value between the two is calculated; according to the deviation value, a pre-established mapping rule is used to judge the tilting speed adjustment requirement corresponding to the deviation value; if the deviation value exceeds a preset threshold range, the adjustment direction of the tilting speed is determined through a logical mapping table; for the adjustment direction, the relevant speed parameter interval is obtained to obtain the specific speed adjustment amplitude; the speed adjustment amplitude and the current tilting speed are superimposed to determine the updated tilting speed value; according to the updated tilting speed value, a regression model is used to predict the speed change trend to judge the stability of subsequent adjustment; if the speed change trend shows instability, the deviation value is calibrated again through a feedback mechanism to obtain a final adjustment scheme.

[0006] Further, the deviation value between the tilting angle and the target angle is calculated according to the dynamic angle data set to determine the tilting speed adjustment direction, including: current tilting angle data is extracted from the dynamic angle data set, compared with a preset target angle, and the deviation value between the two is calculated; according to the deviation value, a pre-established mapping rule is used to judge the tilting speed adjustment requirement corresponding to the deviation value; if the deviation value exceeds a preset threshold range, the adjustment direction of the tilting speed is determined through a logical mapping table; for the adjustment direction, the relevant speed parameter interval is obtained to obtain the specific speed adjustment amplitude; the speed adjustment amplitude and the current tilting speed are superimposed to determine the updated tilting speed value; according to the updated tilting speed value, a regression model is used to predict the speed change trend to judge the stability of subsequent adjustment; if the speed change trend shows instability, the deviation value is calibrated again through a feedback mechanism to obtain a final adjustment scheme.

[0007] Furthermore, the monitoring of the distribution state of molten iron in the mold and the generation of a molten iron flow temperature distribution map include: collecting thermal radiation data of the molten iron flow in the mold by an infrared thermal imaging device to generate original thermal imaging data; using an image processing algorithm to denoise and enhance the original thermal imaging data to obtain a clear thermal image; if there is an abnormal regional brightness in the clear thermal image, extracting the molten iron distribution area by a threshold segmentation algorithm to determine the molten iron flow boundary; calculating the uniformity of the molten iron distribution in the mold according to the molten iron flow boundary to generate a distribution state parameter; comparing the distribution state parameter with a preset uniformity threshold, if the uniformity is lower than the threshold, dynamically adjusting the tilting angle to generate an adjustment instruction; using the adjustment instruction to drive the tilting device to execute the angle change to obtain new thermal imaging data; repeating the processing flow according to the new thermal imaging data to generate a real-time temperature distribution map.

[0008] Furthermore, the defect analysis based on the molten iron flow temperature distribution map and the generation of a defect distribution map include: collecting data of the molten iron flow temperature distribution map, extracting key characteristic values, and constructing an initial temperature distribution matrix; based on the temperature distribution matrix, performing preliminary screening using a preset threshold range, and if the temperature value of a certain area is lower than the threshold, it is determined that the area has the possibility of a cold shut defect, and a preliminary defect area set is obtained; for the preliminary defect area set, combining the molten iron flow characteristic data, analyzing the relationship between the flow velocity and the temperature change, and if the flow velocity is lower than the preset standard and the temperature change is abnormal, it is determined that the area has a porosity defect, and the potential porosity defect area is determined; obtaining the intersection of the potential porosity defect area and the cold shut defect area, and classifying the intersection area using a support vector machine algorithm to determine the defect type and distribution location; extracting defect location information from the classified data, and combining the severity assessment standard to determine the defect severity and obtain a grading result; and constructing a defect distribution diagram through the grading result and the location data to indicate the defect location and severity level.

[0009] Furthermore, the method of adjusting the tipping parameters according to the defect distribution map and optimizing the molten iron flow path includes: obtaining abnormal area data through the defect distribution map, performing preliminary classification based on the defect location and severity, and obtaining abnormal distribution characteristics; determining the initial adjustment parameters of the tipping angle and speed based on the abnormal distribution characteristics using preset mapping rules; if the initial adjustment parameters exceed the preset threshold range, performing a secondary calibration on the parameters through the control system to obtain the calibrated adjustment values; obtaining real-time simulation data of the molten iron flow based on the calibrated adjustment values ​​to determine whether the flow path meets the optimization target; if the flow path does not meet the optimization target, determining new flow path parameters by iteratively adjusting the tipping angle and speed; predicting the defect generation probability based on the new flow path parameters using a support vector machine model to obtain a probability evaluation result; if the probability evaluation result shows that the defect generation probability is higher than the preset threshold, fine-tuning the flow path in combination with historical data to obtain a final optimization solution.

[0010] Furthermore, the position of the smoke hood is adjusted according to the displacement change sequence to maintain synchronous movement, including: collecting displacement change data of the ground model vehicle through sensors, obtaining real-time position information, and obtaining an initial displacement sequence; according to the initial displacement sequence, using a filtering method to smooth the data, eliminate noise interference, and determine the processed displacement trend data; applying a Kalman filter algorithm to predict the displacement trend data, obtaining a future short-term displacement change prediction value, and judging the reliability of the prediction result; if the deviation between the prediction result and the actual collected data exceeds a preset threshold, adjusting the parameters of the prediction model to obtain an optimized displacement prediction sequence; according to the optimized displacement prediction sequence, calculating the adjustment instruction of the smoke hood moving device and determining the target position parameters; sending the target position parameters to the moving device through the control system, driving the smoke hood to adjust its position in real time, and obtaining the synchronous movement execution status; comparing the degree of matching between the execution status and the actual position of the ground model vehicle, if the matching degree is lower than the preset threshold, triggering the feedback mechanism, re-collecting the displacement data, and judging the synchronization effect after adjustment.

[0011] The technical solution provided by the embodiment of the present invention may have the following beneficial effects: The present invention discloses an intelligent control method for the tilting process of a molten iron ladle. By collecting angle data in real time and filtering it, the deviation from the target angle is calculated and the tilting speed is dynamically adjusted. At the same time, infrared thermal imaging is used to analyze the temperature distribution of the molten iron flow, identify cold shut or air hole defects, and optimize the tilting parameters to improve the molten iron flow path. In addition, the present invention also realizes a dynamic sealing connection of the smoke hood, adjusts the position of the smoke hood according to the distance change curve, and optimizes the negative pressure system by monitoring the smoke concentration to improve the smoke capture efficiency. This method realizes the organic combination of precise control of the molten iron ladle tilting process, defect prevention and environmental protection treatment, improves casting quality and production efficiency, and reduces environmental pollution. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 The present invention provides a flow chart of an intelligent molten iron casting system and method. DETAILED DESCRIPTION

[0013] In order to make the objectives, technical solutions and advantages of the present invention more clear, the present invention is described in detail below with reference to the accompanying drawings and specific embodiments.

[0014] like Figure 1 In this embodiment, an intelligent molten iron casting system and method may specifically include: Step S101: collect angle data during the ladle tipping process in real time through a sensor network to generate an initial dynamic angle data set. The initial dynamic angle data set contains time-series angle information for subsequent deviation analysis and control adjustment. In response to possible noise interference during the acquisition process, a filtering algorithm is used to smooth the initial dynamic angle data set to obtain a smoothed angle data set.

[0015] An initial dynamic angle data set is generated through a sensor network, containing time-series angle information, by collecting angle data in real time during the ladle's tilting process. Based on the initial dynamic angle data set, the data collection frequency is checked to see if it meets a preset threshold, thereby determining the completeness of the collection. If the collection frequency is lower than the preset threshold, an interpolation algorithm is used to supplement the missing data, resulting in a completed angle data set. A Kalman filter algorithm is applied to the completed angle data set to perform data denoising, resulting in a denoised angle data set. Based on the denoised angle data set, a time series analysis algorithm is used to detect the changing trend of the angle data and identify a set of outliers. A clustering algorithm is applied to the set of outliers to group the outliers and obtain an outlier data classification result. Based on the outlier data classification result, deviation features are extracted, and the deviation data is smoothed using a mean filter algorithm to obtain a smoothed deviation data set. If a deviation value in the smoothed deviation data set exceeds a preset threshold, it is marked as an abnormal deviation, resulting in an abnormal deviation tag set. Using the abnormal deviation tag set, dynamic angle deviation analysis results are generated, and a final deviation data set is output for subsequent control adjustments.

[0016] For example, a high-precision tilt sensor (model BMS-360, 0.01° accuracy) deployed on the ladle collects tilt angle data in real time at a frequency of 100Hz. This generates a dynamic dataset consisting of timestamps (format: YYYY-MM-DD HH:mm:ss.SSS) and angle values ​​(unit: degrees) and stores it on a cloud-based MQTT server. If the acquisition frequency falls below a preset threshold (e.g., 80Hz), a cubic spline interpolation algorithm is used to supplement missing data points, ensuring a complete record of 100 records per second. A Kalman filter is applied to the supplemented dataset, with measurement noise covariance set to 0.02 and process noise set to 0.001. This filter improves the smoothness of the angle data by 30%. Based on the filtered data, an ARIMA time series model is used to analyze angle trends, with a confidence interval of ±3σ. Data points outside this range are marked as outliers. Outliers are grouped using the DBSCAN clustering algorithm, with a neighborhood radius ε = 0.5° and a minimum sample size min_samples = 3, to identify isolated noise points and continuous outlier segments. Deviation features (such as maximum deviation and duration) are extracted from the clustering results. A mean filter algorithm with a window size of 5 is used to smooth the deviation sequence. If the smoothed deviation value exceeds a preset threshold (e.g., ±1.5°), it is marked as an abnormal deviation and the time interval is recorded. Finally, all abnormal flags are integrated to generate a dynamic angle deviation analysis report containing the deviation type, magnitude, and time distribution, which is then output to the control system adjustment module.

[0017] In step S102, for the smooth angle data set, the deviation value between the current tilting angle and the preset target angle is calculated, the mapping rule is used to judge the speed adjustment requirement corresponding to the deviation value, the tilting speed adjustment direction is determined to be accelerated or slowed down, and the speed adjustment parameter is generated, which is used to guide the subsequent motor control.

[0018] The dynamic angle data set and the preset target angle are obtained, the deviation value between the current tilting angle and the preset target angle is calculated by comparison, and the deviation value is obtained. According to the deviation value, the mapping rule is used to analyze the corresponding relationship between the deviation value and the speed adjustment requirement, and the speed adjustment direction is determined to be accelerated or slowed down. If the deviation value is greater than the preset threshold value, the adjustment parameter for accelerating the tilting speed is generated; if the deviation value is less than or equal to the preset threshold value, the adjustment parameter for slowing down the tilting speed is generated, and the speed adjustment parameter is obtained. Through the speed adjustment parameter, the speed change amount required for motor control is calculated by combining the current motor state data, and the speed adjustment instruction is obtained. According to the speed adjustment instruction, the motor control algorithm is used to generate specific motor driving signals, and the motor driving parameters are determined. The motor driving parameters are obtained, the motor driving parameters are converted into pulse width modulation signals of the motor through the signal processing module, and the pulse width modulation signals are obtained. If the frequency of the pulse width modulation signal meets the motor operation requirement, it is transmitted to the motor control unit; if not, the signal frequency is adjusted to obtain the pulse width modulation signal meeting the requirement. According to the adjusted pulse width modulation signal, the motor speed adjustment is executed through the motor control unit, and the real-time motor operation state data is obtained. Through the closed-loop feedback of the real-time motor operation state data and the dynamic angle data set, the calculation of the deviation value is updated, and the new deviation value is obtained.

[0019] For example, the dynamic angle data set is obtained from the sensor, such as the current tilting angle of 45° and the preset target angle of 50°, and the deviation value of 5° is calculated. Using the mapping rule table, if the deviation value is in the range of 0°-10° and the speed adjustment requirement is acceleration, the adjustment direction is determined to be acceleration. The threshold value is set to 3°, and since 5° is greater than 3°, the speed adjustment parameter is generated as increasing the speed by 10%. Combined with the current motor speed of 2000 rpm, the target speed is calculated to be 2200 rpm, and the speed adjustment instruction is output. Using the PID control algorithm, the speed difference of 200 rpm is input, and the motor driving parameter with a duty cycle of 60% is generated after calculation by the proportional coefficient of 0.5. The duty cycle is converted into a 10 kHz pulse width modulation signal through the PWM module, and if the signal frequency deviates to 9.5 kHz, the frequency coefficient is adjusted to restore the frequency to 10 kHz. The corrected signal is input into the motor controller to drive the motor to accelerate to 2150 rpm, and the real-time speed data is collected. Combined with the newly collected tilting angle of 48°, the deviation value of 2° is recalculated, and the next adjustment cycle is entered.

[0020] Step S103, if the deviation value exceeds the preset threshold range, a speed adjustment instruction is sent to the tilting motor through the control system, the speed is dynamically adjusted according to the speed adjustment parameter, the tilting angle gradually approaches the target value, and real-time angle feedback data is collected to determine whether the adjusted angle change trend meets the expectation.

[0021] The actual value of the tilting angle is collected in real time by the sensor, compared with the preset target value, and the deviation value is obtained. If the deviation value exceeds the preset threshold range, a speed adjustment instruction is generated by the control system to determine the specific parameters of the adjustment instruction. According to the generated speed adjustment instruction, a dynamic adjustment signal is sent to the tilting motor to obtain real-time change data of the motor speed. The feedback data of the tilting angle is collected by the real-time monitoring system to determine whether the angle change gradually approaches the target value. If the angle change trend does not meet the expectation, the speed adjustment parameter is recalibrated by the control system to obtain the calibrated adjustment value. According to the calibrated adjustment value, a new speed adjustment instruction is generated to determine a new motor control signal. The new adjustment instruction is sent to the tilting motor by the control system to obtain the adjusted angle change data. According to the adjusted angle change data, the deviation value is continuously compared with the target value to determine whether the deviation value is still within the threshold range. If the deviation value still exceeds the threshold range, the parameter calibration and speed adjustment process is executed by the control system to obtain the final angle stabilization data.

[0022] For example, the current tilting angle value is collected in real time by the tilting angle sensor installed on the device at a sampling frequency of 100 Hz, and the difference value is calculated with the preset target value of 45°. When the deviation value exceeds the threshold of ±2°, the control logic is triggered. The PID control algorithm is used to generate the speed adjustment instruction, the proportional coefficient Kp is set to 0.8, the integral time Ti is 0.1 seconds, and the output PWM duty cycle adjustment range is 20%-80%. The adjustment instruction is converted into a motor driving signal, which is sent to the tilting motor controller through the CAN bus at a rate of 500 kbps. The linear change curve of the motor speed from 1200 rpm to 1500 rpm is recorded in real time. The tilting angle feedback data is collected synchronously, and the change rate is calculated every 50 ms. If the angle change does not reach the expected slope of 0.5° / s within 10 seconds, the secondary calibration process is started. The least squares method is used to fit the historical adjustment data, and the PID parameters are recalculated as Kp=1.2 and Ti=0.05 seconds to generate a new PWM duty cycle control signal of 65%. The updated instruction increases the motor speed to 1800 rpm, and the tilting angle approaches the target value at a rate of 0.8° / s. The deviation between the real-time angle and the target value is continuously compared, and when the deviation value is still greater than ±1°, the iterative optimization process is triggered. The control parameters are dynamically corrected by the gradient descent method until the angle stabilizes in the range of 44.9°-45.1°.

[0023] In step S104, while adjusting the tilting angle, the infrared thermal imaging device is used to collect thermal radiation data of the molten iron in the mold to generate original temperature distribution data, and the original temperature distribution data is subjected to denoising and image enhancement processing to obtain a clear molten iron flow temperature distribution map, which is used to analyze the molten iron distribution state.

[0024] The infrared thermal imaging device is used to collect thermal radiation data of the molten iron in the mold at a fixed frequency to generate an original temperature distribution data set. The median filter algorithm is used to remove noise points for the original temperature distribution data set to obtain a denoised temperature distribution data set. The image enhancement algorithm is applied to adjust the contrast and brightness according to the denoised temperature distribution data set to generate an enhanced temperature distribution data set. The molten iron flow temperature distribution map is generated by performing gray value mapping on the enhanced temperature distribution data set. If there is a temperature abnormal region in the molten iron flow temperature distribution map, the edge detection algorithm is used to extract the boundary of the abnormal region to obtain a temperature abnormal region data set. According to the temperature abnormal region data set, the molten iron distribution unevenness feature is calculated in combination with the ladle tilting angle data to generate a distribution unevenness feature set. The clustering algorithm is used to classify the molten iron flow state through the distribution unevenness feature set to obtain a molten iron distribution state classification result. If the molten iron distribution state classification result shows that there is a cold shut or gas hole defect, the defect position and degree are predicted through the regression algorithm to generate a defect analysis data set. According to the defect analysis data set, the ladle tilting angle parameter is adjusted to generate an optimized tilting control parameter set.

[0025] For example, the infrared thermal imaging device collects the thermal radiation data of the molten iron in the mold at a frequency of 30 frames per second, generates a raw data set containing a temperature matrix, and each pixel point corresponds to a 0.5°C resolution. For the raw data, a 3*3 window median filtering algorithm is used to eliminate impulse noise, and a threshold is set to remove abnormal values of ±50°C, and a smooth temperature matrix is output. The histogram equalization algorithm is applied to the denoised data to adjust the gray scale range to 0-255, and the contrast of the low temperature area is enhanced. Through bilinear interpolation, the enhanced data is mapped to a temperature distribution map of 512*512 pixels, and the pseudo-color coding is displayed in the 400-1600°C interval. If a region with a local temperature difference exceeding 200°C is detected in the temperature map, a Canny operator is used to extract the contour, and an abnormal area data set containing coordinates and temperature difference values is generated. Combined with the inclination angle data (accuracy 0.01°) collected by the inclination sensor in real time, the correlation coefficient of the flow rate of the molten iron front and the rate of change of the angle is calculated, and a feature vector containing the flow rate deviation and the temperature gradient is output. The K-means clustering algorithm (k=3) is used to classify the feature vector, and the normal flow, vortex and stagnation are divided into three states. When the clustering result shows that the stagnation state accounts for more than 15%, the position of the blowhole defect is predicted through a polynomial regression model, and a defect probability distribution map is output. Based on the defect distribution result, a PID control algorithm is used to dynamically adjust the rotation speed of the tilting motor, and an optimized parameter set containing the target angle and acceleration is generated.

[0026] In step S105, the temperature feature values are extracted and the temperature distribution matrix is constructed for the molten iron flow temperature distribution map. The abnormal area is screened through the preset temperature threshold range, and it is judged whether there is a cold shut or blowhole defect. A defect distribution map is generated, which labels the specific position and severity of the defect, providing a basis for subsequent optimization.

[0027] The molten iron flow temperature distribution map data is obtained, and the original temperature data points are extracted. According to the extracted temperature data points, a two-dimensional temperature distribution matrix is constructed. Through the preset cold shut defect temperature threshold range, the abnormal low temperature area in the temperature distribution matrix is screened. If the abnormal low temperature area is screened out, it is marked as a potential cold shut defect area, and the preliminary position of the cold shut defect is obtained. Through the preset blowhole defect temperature threshold range, the high temperature fluctuation area in the temperature distribution matrix is analyzed. If the high temperature fluctuation area is analyzed, it is marked as a potential blowhole defect area, and the preliminary position of the blowhole defect is obtained. The severity of the cold shut and blowhole defects is calculated by combining the defect position and temperature gradient algorithm. According to the calculated severity, a defect distribution map containing defect position and severity labels is generated. Through the defect distribution map data, an image file labeled with the specific position and severity of the cold shut and blowhole defects is output.

[0028] Exemplarily, the molten iron flow temperature distribution map data is acquired, the temperature data is collected by using an infrared thermal imager, and the temperature value of each pixel point is extracted, for example, 1 million temperature data points are extracted in a 1000x1000 pixel image. According to the extracted temperature data points, a two-dimensional temperature distribution matrix is constructed, the matrix rows and columns correspond to the image coordinates, and the matrix elements store the temperature values of the corresponding positions, such as matrix T(i,j) representing the temperature of the i-th row and the j-th column. By presetting a cold shut defect temperature threshold range, setting a low temperature threshold of 1300°C or lower, traversing the temperature distribution matrix, and screening all continuous regions below 1300°C as abnormal low temperature regions. If an abnormal low temperature region is screened out, a region growing algorithm is used to mark the connected domain to obtain the preliminary position of the cold shut defect, such as the rectangular region with coordinates (x1, y1) to (x2, y2). By presetting a pore defect temperature threshold range, setting a high temperature fluctuation threshold of 1500°C or higher and an adjacent temperature difference of more than 50°C, analyzing the mutation points in the temperature distribution matrix, and marking the regions that meet the conditions. If a high temperature fluctuation region is analyzed, an edge detection algorithm is used to identify discontinuous high temperature points to obtain the preliminary position of the pore defect, such as the circular region with coordinates (x3, y3) as the center. A defect position and temperature gradient combined algorithm is used to calculate the severity of the cold shut defect as (1300-Tmin) / 100 and the severity of the pore defect as AT / 50, where Tmin is the minimum temperature and AT is the maximum temperature difference. According to the calculated severity, a defect distribution map is generated, and the cold shut defect is marked with a blue translucent region on the image, the pore defect is marked with a red circle, and the severity level 1-5 is displayed with a numerical label. Through the defect distribution map data, a PNG format image file is output, with a resolution of 1920x1080, containing a legend to explain the correspondence between the color and the severity.

[0029] In step S106, if the defect distribution map shows that there is a defect, a fine adjustment instruction of the tilting angle and speed is generated by the control system according to the defect position and severity to adjust the molten iron flow path, and the adjusted flow simulation data is obtained to determine whether the defect generation probability is reduced to determine the optimized parameters.

[0030] The defect distribution map data is obtained, the defect position and severity are identified through image processing technology, the coordinates and classification level of the defect area are obtained. According to the coordinates and classification level of the defect area, combined with the molten iron flow characteristic data, the flow velocity and temperature change in the defect area are analyzed to determine the potential defect type. If the flow velocity is lower than the preset standard and the temperature change is abnormal, the control system generates a preliminary tilting angle and speed fine tuning instruction to obtain an adjustment parameter set. The adjustment parameter set is used to perform simulation calculation of the molten iron flow path to obtain adjusted flow simulation data. By comparing the adjusted flow simulation data with the original flow data defect generation probability, it is judged whether the adjustment reduces the defect generation probability, and the probability change value is obtained. If the probability change value shows that the defect generation probability is reduced, the optimized tilting angle and speed parameters are extracted according to the adjusted flow simulation data, and the optimization parameter set is determined. According to the optimization parameter set, the tilting angle and speed settings of the control system are updated to generate new molten iron flow path data. Through the new molten iron flow path data, the defect distribution map is regenerated to judge whether there is a new defect area, and the updated defect distribution data is obtained. According to the updated defect distribution data, the defect position and severity analysis is circularly executed to obtain the final molten iron flow optimization parameters.

[0031] For example, the generation step is as follows: The defect distribution map is scanned by an image processing algorithm (such as edge detection and threshold segmentation) to identify the coordinates of the defect area (such as X=120mm, Y=80mm) and classify the level (such as severity level 3). Combined with the molten iron flow characteristic database (such as flow rate 2.5m / s, temperature gradient ΔT=15℃ / s), the flow data of the defect area is analyzed. If the flow rate is detected to be lower than 1.8m / s and the temperature fluctuation exceeds ±20℃, it is determined as a gas hole defect. The control system generates a tilting angle adjustment instruction (such as increasing 2°) and a speed fine tuning instruction (such as decreasing 0.3m / s) based on the PID algorithm to form a parameter set. The adjusted flow path is simulated by computational fluid dynamics (CFD) to output flow velocity field and temperature field data. If the defect probability is reduced from 12% to 8% by comparing the simulation data with the original data, the optimization parameters (such as tilting angle 15°, flow rate 2.2m / s) are extracted. After updating the control system parameters, real-time flow simulation is performed to generate a new defect distribution map. If the new map shows that the defect area is reduced to one, the coordinate analysis and parameter optimization are circularly executed to finally output the stable parameters (such as tilting angle 16°, flow rate 2.1m / s).

[0032] In step S107, while optimizing the molten iron flow path, the relative distance data between the molten iron ladle and the smoke collecting hood is collected by a laser ranging device to generate an initial distance data set. The initial distance data set is smoothed to obtain a distance change curve, which is used for subsequent dynamic position adjustment and sealing control.

[0033] Using laser ranging equipment combined with dust and high-temperature protection measures, the relative distance data between the ladle and the fume hood is collected in real time to generate an initial distance data set. Based on this initial distance data set, a smoothing algorithm is used to remove noise and extract trends from the data, resulting in a smoothed distance curve. The distance curve is used to analyze the relative position change trend between the ladle and the fume hood and determine the baseline parameters for dynamic adjustment. Based on the baseline parameters and preset distance range thresholds, a position deviation is determined to determine whether the position deviation exceeds the allowable range, resulting in a deviation determination result. If the deviation determination result indicates that the deviation exceeds the allowable range, a position adjustment instruction is generated to determine the movement direction and amplitude of the fume hood follower. Based on the position adjustment instruction, the fume hood follower performs dynamic position adjustment, acquiring adjusted real-time distance data. The adjusted real-time distance data is compared with the preset distance range to determine whether the target distance range is reached, generating a comparison result. If the comparison result indicates that the target distance range is not reached, the deviation determination and position adjustment steps are repeated to obtain new adjustment data. Based on the final adjustment data, the stable relative position of the ladle and the fume hood is recorded, and dynamic sealing control parameters are generated to determine the sealing connection status.

[0034] For example, a laser ranging device uses an infrared laser with a wavelength of 905 nanometers and a sampling frequency of 100Hz to collect real-time distance data between the edge of the ladle and the bottom edge of the fume hood. Combined with a compressed air dust control system embedded in the high-temperature protective hood, this ensures the acquisition of an initial distance data set with millimeter-level accuracy in a 1200°C environment. The raw data is smoothed using a Savitzky-Golay filter algorithm, with a window width of 15 sampling points and a quadratic polynomial fit to eliminate ±5mm random noise caused by molten iron sloshing, generating a continuous distance curve. The position change rate is calculated and analyzed using the curve derivative. When the rate of change exceeds 0.2m / s, the ladle is determined to have begun tipping, triggering dynamic adjustment of the baseline parameters. A preset distance threshold of 200±10mm is used, and the real-time deviation is calculated using the sliding window standard deviation. If three consecutive sampling points exceed the threshold, an adjustment command is generated. Based on the deviation, a PID control algorithm is used to calculate the adjustment amount, with a proportional coefficient set to 0.8. X / Y axis movement commands are then output to the servo motor-driven follower. After adjustment, distance data is collected again and multi-sensor data is fused using a Kalman filter. The standard is considered met when five consecutive sampling values ​​fall within the 195-205mm range. If the standard is not met, the adjusted data is used as the new input, and the PID calculation and motion control are repeated until the cumulative number of adjustments reaches 10 or a stable state is reached. The final recorded stable distance value is 203mm, and a sealing pressure control curve is generated. The linear solenoid valve outputs a dynamic pressure of 0.5-1.2MPa according to the curve.

[0035] Step S108, according to the distance change curve, the adjustment parameters of the fume hood follower device are calculated, the position of the fume hood is dynamically adjusted by the control system to drive the follower device to ensure a predetermined distance range with the ladle, to achieve dynamic sealing connection, and real-time position data is collected to determine whether the sealing state meets the preset standard.

[0036] Real-time distance data between the ladle and the fume hood is collected to obtain a distance change curve. According to the distance change curve, the adjustment parameters of the fume hood follower device are calculated. Adjustment signals are sent through the control system to drive the follower device to change the position of the fume hood. Real-time position data of the fume hood after adjustment is obtained to update the position information. If the real-time position data does not meet the predetermined distance range, the adjustment parameters are recalculated. The follower device is driven again by the adjusted parameters to update the position of the fume hood. The sealing state data of the fume hood after adjustment is collected to obtain sealing state information. If the sealing state data does not meet the preset standard, a deviation correction signal is generated. The follower device parameters are adjusted according to the deviation correction signal to determine the dynamic sealing connection of the fume hood and the ladle.

[0037] For example, a laser ranging sensor is used to collect real-time distance data between the ladle and the fume hood at a sampling frequency of 100 Hz, and a Kalman filter algorithm is used to eliminate noise interference to generate a smooth distance change curve. Based on the slope change rate of the curve, a PID control algorithm is used to calculate the displacement increment parameters of the follower device, where the proportional coefficient Kp is set to 0.8 and the integral time Ti is 0.1 s. A 4-20 mA current signal is output through a PLC control system to drive a servo motor to adjust the height of the fume hood at a rate of 0.05 m / s. An encoder installed on the follower device feeds back real-time position data, which is updated to the database and compared with the preset range (±10 mm). If the deviation is detected to exceed the threshold, the PID parameters are re-optimized using the gradient descent method with an iteration step size of 0.01. The adjusted parameters trigger the servo motor to position again, and a pressure sensor collects the contact force data of the sealing surface with a sampling period of 50 ms. When the contact force is less than 15 N, a correction signal containing direction and amplitude is generated based on a fuzzy logic algorithm (e.g. +5 mm / 0.2 N). Finally, the servo motor angle is corrected according to the correction signal to stabilize the sealing gap within the range of 3-5 mm.

[0038] Step S109, for the dynamic sealing connection state, a gas sensor is used to collect smoke concentration data inside the fume hood to generate a concentration change data set. By analyzing the fluctuation characteristics of the concentration change data set, it is determined whether the smoke collection efficiency meets the preset threshold to determine the adjustment requirements of the negative pressure system and generate corresponding control instructions.

[0039] The gas sensor collects the smoke concentration data in the smoke hood in real time at a preset frequency, generates an initial concentration data set containing a time stamp and a concentration value. According to the initial concentration data set, the time series analysis algorithm is used to extract the fluctuation characteristics of the concentration data, and the fluctuation characteristic data set is obtained. Through the fluctuation characteristic data set, the standard deviation and mean value of the concentration fluctuation are calculated by using the statistical analysis method, and the quantization index of the fluctuation characteristics is determined. According to the quantization index, combined with the preset concentration fluctuation threshold, it is judged whether the smoke collection efficiency reaches the preset standard. If the judgment result shows that the smoke collection efficiency is lower than the preset threshold, the specific time period and the concentration abnormal value of the insufficient efficiency are obtained by analyzing the abnormal points in the fluctuation characteristic data set. According to the specific time period and the concentration abnormal value of the insufficient efficiency, combined with the dynamic sealing state data of the smoke hood, the adjustment requirement of the negative pressure system is determined. Through the adjustment requirement, the target pressure value and the adjustment range of the negative pressure system are calculated by using the control algorithm, and the preliminary control instruction data set is generated. According to the preliminary control instruction data set, combined with the real-time collected sealing contact force data, the optimized final control instruction is obtained. Through the final control instruction, it is transmitted to the negative pressure system execution module to complete the automatic adjustment of the negative pressure system, and the smoke collection efficiency state data is updated.

[0040] For example, a gas sensor (e.g., MQ-135, sampling frequency 50Hz) collects smoke concentration data inside the smoke hood in real time, generating an initial concentration dataset containing a timestamp (format: YYYY-MM-DD HH:mm:ss) and concentration values ​​(unit: ppm). This data is transmitted via RS485 and stored in a cloud database, with a CRC32 checksum to ensure integrity. An ARIMA time series analysis algorithm is applied to the initial concentration dataset to extract the periodicity and trend characteristics of concentration fluctuations. A sliding window size is set to 10 seconds, and the concentration range and slope within the window are calculated to obtain a fluctuation characteristic dataset. Based on the fluctuation characteristic dataset, the concentration fluctuation dispersion is calculated using the standard deviation formula σ = √(Σ(xi - μ)² / N), where μ is the mean and N is the number of data points. Combined with a preset threshold (e.g., σ ≤ 50 ppm is considered acceptable), a quantitative indicator is generated. If the quantitative indicator shows σ > 50 ppm, the DBSCAN clustering algorithm is used to identify outliers, setting a neighborhood radius of ε = 5 ppm and a minimum sample count of min_samples = 3. Periods in which concentration suddenly increases by more than 100 ppm are marked as outliers. Based on the outlier data, combined with real-time data on the distance between the fume hood and the ladle (accuracy ±1 mm) collected by a laser rangefinder, the correlation coefficient between the seal gap and concentration fluctuations is calculated. If the correlation coefficient R² > 0.7, negative pressure adjustment is determined. A PID control algorithm is used to calculate the target pressure, setting the proportional coefficient Kp = 0.8, the integral time Ti = 10 seconds, the differential time Td = 2 seconds, and the output pressure adjustment value ΔP = Kp(e + 1 / Ti∫edt + Tdde / dt). This generates preliminary control commands. These preliminary commands are compared with real-time seal contact force data collected by a pressure sensor (range 0-10 kPa, accuracy ±0.5%). If the pressure deviation exceeds 5%, a least-squares method is used to fit the correction coefficient to generate the final control command. The final command is sent to the negative pressure system inverter via the Modbus protocol to adjust the fan speed (range 0-3000rpm). The updated concentration data is written to the database and a new round of monitoring cycle is triggered.

[0041] In step S1010, if the concentration change data set shows that the flue gas concentration is lower than the preset threshold, the fan speed of the negative pressure system is adjusted by the control system to increase the negative pressure value to enhance the flue gas capture efficiency, and the adjusted flue gas concentration data is obtained. The adjusted flue gas concentration data is continuously monitored to determine whether the capture efficiency has stably reached the target.

[0042] A concentration change data set is acquired. Sensors collect flue gas concentration data in real time to obtain the current flue gas concentration value. If the current flue gas concentration value is below a preset threshold, the control system calculates the negative pressure system adjustment parameters and determines the fan speed adjustment. Based on the adjustment value, the control system sends a command to the negative pressure system to adjust the fan speed, resulting in an increased negative pressure value. The adjusted negative pressure value enhances the flue gas capture efficiency and obtains adjusted flue gas concentration data. A monitoring algorithm is used to continuously analyze the adjusted flue gas concentration data to determine whether the capture efficiency has reached the stability target. If the capture efficiency has not reached the stability target, an iterative optimization algorithm is used to recalculate the negative pressure system adjustment parameters to obtain a new fan speed adjustment value. Based on the new adjustment value, the control system adjusts the fan speed again to obtain an updated negative pressure value. If the optimized flow path deviation exceeds the preset threshold, the ladle position is adjusted using an iterative optimization algorithm to generate new relative distance data. Based on the new relative distance data, the control system recalculates the flue gas capture efficiency and obtains adjusted flue gas concentration data.

[0043] For example, an electrochemical sensor installed in the flue collects flue gas concentration data once per second and transmits it to a PLC control system. The current flue gas concentration is 80 mg / m³. If this value falls below the preset threshold of 100 mg / m³, a PID algorithm is used to calculate the negative pressure system adjustment parameters. Based on the 20 mg / m³ deviation, the fan speed needs to be increased by 15%. The control system sends a command to the inverter via the Modbus protocol, adjusting the fan speed from 1200 rpm to 1380 rpm, raising the system negative pressure from -200 Pa to -230 Pa. A computational fluid dynamics model is used to analyze the flow field distribution after the negative pressure increase, resulting in an adjusted flue gas concentration of 95 mg / m³. A sliding window algorithm is used to calculate the standard deviation of 30 consecutive concentration data sets. If the standard deviation exceeds 2 mg / m³, the capture efficiency is determined to be unstable. The PID parameters are then iteratively optimized using the gradient descent method, and the speed is recalculated to increase by another 5%. The control system adjusts the fan speed to 1449 rpm, bringing the negative pressure to -241 Pa. If the laser rangefinder detects a deviation of more than ±5mm between the ladle and the fume hood, a genetic algorithm is used to optimize the position parameters, generating a new relative distance of 3.2mm. Based on the updated spatial coordinates, CFD simulation recalculates the flow field characteristics, ultimately outputting flue gas data with a stable concentration of 102mg / m³.

[0044] The preferred embodiments of the present invention disclosed above are intended only to help illustrate the present invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the present invention to specific embodiments. Obviously, many modifications and variations are possible based on the contents of this specification. These embodiments are selected and described in detail in this specification to better explain the principles and practical applications of the present invention, thereby enabling those skilled in the art to better understand and utilize the present invention. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. An intelligent molten iron casting system and method, characterized in that: The following steps are involved: (S1) collecting ladle tilting angle data through a sensor network to generate a dynamic angle dataset; (S2) calculating a deviation between a tilting angle and a target angle based on the dynamic angle data set, and determining a tilting speed adjustment direction; (S3) adjusting the speed of the tilting motor according to the deviation value so that the tilting angle approaches the target value; (S4) monitoring the distribution state of the molten iron in the mold and generating a molten iron flow temperature distribution map; (S5) performing defect analysis based on the molten iron flow temperature distribution map to generate a defect distribution map; (S6) adjusting the tilting parameter according to the defect distribution map to optimize the molten iron flow path; (S7) monitoring the relative position of the ladle and the fume hood and generating a distance change curve; (S8) dynamically adjusting the position of the smoke hood based on the distance variation curve to achieve a sealed connection; (S9) monitoring the smoke concentration in the smoke hood and generating a concentration change data set; (S10) regulating the negative pressure system according to the concentration change data set to improve the flue gas capture efficiency; (S11) recording the displacement trajectory of the mold transport vehicle and generating a displacement change sequence; (S12) synchronously adjusting the position of the smoke hood based on the displacement change sequence; (S13) monitoring the leakage intensity of the dynamic sealing area and generating a leakage intensity data set; (S14) Adjusting the sealing device pressure according to the leakage intensity data set to optimize the sealing effect.

2. The system and method according to claim 1, characterized in that The step (S1) comprises: Obtaining an initial dynamic data set of rollover angles; When the acquisition frequency is lower than the preset threshold, the missing data is supplemented by an interpolation algorithm; Time series analysis is used to detect angle change trends and identify abnormal point sets; Group abnormal data through clustering algorithms; After extracting the deviation features, the mean filter is used for smoothing; If the deviation value after smoothing exceeds the preset threshold, it is marked as an abnormal deviation; Output dynamic angle deviation analysis results.

3. The system and method according to claim 1, characterized in that The step (S2) comprises: Extract the current rollover angle data and compare it with the target angle; When the deviation value exceeds the threshold range, the adjustment direction is determined through the logic mapping table; Get the speed parameter range and calculate the adjustment range; Add the adjustment amplitude to the current tipping speed; Use regression model to predict speed change trend; When the trend is unstable, secondary calibration is performed through the feedback mechanism.

4. The system and method according to claim 1, characterized in that The step (S3) comprises: When the deviation value exceeds the threshold, real-time rollover angle data is obtained; The proportional-integral-differential algorithm is used to calculate the speed adjustment; Sending speed adjustment instructions to the tilting motor; If the rollover angle does not approach the target value, the angle data is smoothed using Kalman filtering; The speed adjustment command is regenerated based on the optimized deviation value.

5. The system and method according to claim 1, characterized in that The step (S4) comprises: Collect thermal radiation data inside the mold through infrared thermal imaging equipment; De-noising and enhancing the raw thermal imaging data; When there is regional brightness anomaly, the molten iron flow boundary is extracted through threshold segmentation; Calculate the uniformity of molten iron distribution and compare it with the preset threshold; If the uniformity is lower than the threshold, a tilt angle adjustment instruction is generated; Generate real-time temperature distribution maps based on the adjusted thermal imaging data.

6. The system and method according to claim 1, characterized in that The step (S5) comprises: Construct molten iron temperature distribution matrix; When the area temperature is lower than the threshold, it is marked as a cold shut defect area; Combining the relationship between flow velocity and temperature change to identify the pore defect area; Support vector machine algorithm is used to classify defect areas; Generates graded results based on severity assessment criteria; Construct defect distribution maps with location and severity.

7. The system and method according to claim 1, characterized in that The step (S12) includes: Collect the displacement data of the mold transport vehicle and generate the initial sequence; Eliminate noise interference through filtering methods; Apply Kalman filter algorithm to predict short-term displacement changes; Optimize the displacement prediction model when the prediction deviation exceeds the threshold; Generate smoke hood position adjustment instructions based on the optimized prediction sequence; When the position matching degree between the fume hood and the mold transport vehicle is lower than a threshold, the displacement data is collected again.

8. The system and method according to claim 1, characterized in that: The step (S8) comprises: The distance sensor is used to collect the distance data between the ladle and the smoke hood in real time; constructing a distance-time variation curve based on the spacing data; When the distance change rate exceeds a preset threshold, a smoke hood displacement compensation instruction is generated; Executing the displacement compensation instruction through a hydraulic drive device; Verify the sealing connection status in real time and trigger a secondary position calibration if the seal fails.

Images