Intelligent monitoring method, system and storage medium for the whole process of hot working

Abstract

The present application relates to the technical field of intelligent monitoring, and more particularly to an intelligent monitoring method and system for the whole process of hot processing and a storage medium. The method comprises the following steps: S1, collecting temperature, stress and other data through a sensor array, extracting space-time features, generating a defect initiation probability graph, and determining a potential defect area; S2, optimizing processing parameters according to the characteristics of the potential defect area, applying them to a simulation environment, and calculating a defect risk assessment value; S3, predicting a defect propagation path based on risk assessment and defect classification, and constructing a dynamic defect propagation model; S4, updating genetic algorithm refinement and optimizing processing parameters by comparing predicted defect propagation data and real-time data; and S5, adjusting processing equipment control instructions according to the optimized parameters, and generating a stable set of processing parameters. The method significantly improves processing quality and reduces defect occurrence through intelligent monitoring and dynamic optimization.

Description

Technical Field

[0001] This invention relates to the field of intelligent monitoring technology, and in particular to an intelligent monitoring method, system and storage medium for the entire process of thermal processing. Background Technology

[0002] In modern industrial manufacturing, hot working is a common processing method widely used in the processing of metals and other materials. The hot working process includes stages such as heating, holding, and cooling, and the parameters of each stage (such as temperature, time, and cooling rate) directly affect the performance and quality of the processed material. However, traditional hot working processes often rely on experience and manual operation, lacking precise real-time monitoring and automated adjustment, which easily leads to defects such as microscopic defects like porosity and cracks. These defects can significantly affect the quality and performance of processed products, especially under requirements of high precision and high reliability, such as in the processing of steel components for high-speed rail tracks.

[0003] While some existing technologies utilize sensors to collect processing data for real-time monitoring, most are limited to monitoring single parameters, such as temperature or stress, failing to consider multi-dimensional time-series data analysis and lacking systematic defect prediction and optimization control methods. Furthermore, many methods still rely on empirical adjustments, lacking intelligent processing capabilities, resulting in the inability to effectively predict and optimize defects during the processing.

[0004] Therefore, the main problems with existing technologies are: the lack of comprehensive and multi-dimensional monitoring methods, the inability to predict and assess potential defects in the processing process in real time and accurately, and the reliance on human experience in handling defects after they occur, resulting in low production efficiency and unstable product quality.

[0005] To address the aforementioned issues, this invention proposes an intelligent monitoring method that can predict and optimize the occurrence of defects during the hot processing process by real-time acquisition and intelligent analysis of multi-dimensional time-series data, combined with neural networks and genetic algorithms. This enables intelligent control of the processing process, ensuring the stability and high reliability of the processing quality. Summary of the Invention

[0006] This invention proposes an intelligent monitoring method, system, and storage medium for the entire hot working process, covering the entire process from data acquisition to processing parameter optimization, and achieving precise control of the hot working process through multi-dimensional time-series data analysis and intelligent algorithms.

[0007] Firstly, the present invention provides an intelligent monitoring method for the entire hot working process, mainly comprising: Step S1: Collect multi-dimensional time-series datasets through a sensor array, extract spatiotemporal features based on the multi-dimensional time-series datasets through a neural network, obtain a defect initiation probability map by analyzing the spatiotemporal features, and determine potential defect areas based on the defect initiation probability map; Step S2: Determine the optimal parameter combination based on the characteristics of the potential defect area, apply the optimal parameter combination to the processing simulation environment, obtain the defect type and occurrence probability based on the probability calculation model, and calculate the defect risk assessment value; Step S3: Predict the defect propagation path based on the defect risk assessment value and defect classification, and construct a dynamic defect propagation model based on the predicted defect propagation path and the multidimensional time series dataset; Step S4: Based on the dynamic defect propagation model and the collected data, obtain the predicted defect propagation data, compare the predicted defect propagation data with the real-time collected defect propagation data, update the genetic algorithm according to the comparison results, and determine the refined optimization parameter combination; Step S5: Adjust the control commands of the processing equipment according to the combination of refining and optimization parameters to obtain a stable set of processing parameters.

[0008] As a preferred embodiment of the present invention, step S1 involves acquiring a multi-dimensional time-series dataset using a sensor array, including: The heating temperature, holding time, cooling rate, material crystal structure, ambient humidity, and equipment vibration index of the hot-worked parts are collected in real time by a sensor array to obtain a time-series dataset containing multi-dimensional features. The multi-dimensional time-series dataset includes at least time-series data of the hot-worked temperature field, the stress field of the hot-worked parts, and environmental parameters. The multi-dimensional time-series dataset is preprocessed to obtain a standardized multi-dimensional time-series dataset.

[0009] As a preferred embodiment of the present invention, step S1 involves using a neural network to extract spatiotemporal features, including: A multidimensional time-series dataset is input into a convolutional neural network, and the spatiotemporal distribution features of the temperature field and stress field are extracted. By analyzing the temperature field features and stress field features, the spatial temperature field gradient and stress field peak distribution are obtained. Based on the temperature field gradient and stress field peak distribution, a defect initiation probability map is generated. High-risk areas in the defect initiation probability map where the temperature gradient exceeds a preset threshold and overlaps with the stress peak area are identified as potential porosity defect initiation areas and are used as potential defect areas.

[0010] As a preferred embodiment of the present invention, step S2, determining the optimal parameter combination based on the characteristics of the potential defect region, includes: If the temperature gradient in the potential defect area exceeds a preset threshold, the population is initialized using a genetic algorithm, a fitness function is defined, and the population is iteratively optimized according to the fitness function to obtain an optimized combination of heating temperature, holding time, and cooling rate. The fitness function is designed based on the temperature gradient and the probability of defect initiation and is used to evaluate the effectiveness of the processing parameters.

[0011] As a preferred embodiment of the present invention, step S2 involves obtaining an optimized parameter combination and applying it to a processing simulation environment to obtain a defect risk assessment value, including: The optimized combination of heating temperature, holding time, and cooling rate is obtained and input into the processing simulation environment. The simulation processing process is executed, and the simulation results, simulation environment parameters, and material response characteristics during the simulation processing are input into the probability calculation model to calculate the defect type and the corresponding probability of occurrence. The defect type and the corresponding probability of occurrence are weighted to obtain the defect risk assessment value. The simulation results include the stress distribution and temperature change data of the hot-processed material.

[0012] As a preferred embodiment of the present invention, step S3 includes: Based on the defect risk assessment value, defects are classified using a defect type classification method. Based on the classification results and real-time data feedback, the propagation path direction and speed of defects are predicted to obtain the defect propagation mechanism. A dynamic defect extension model is constructed by analyzing the multidimensional time-series dataset and the defect propagation mechanism. The dynamic defect extension model is used to characterize the evolution behavior of defects during the hot processing process.

[0013] As a preferred embodiment of the present invention, step S4 includes: The dynamic defect expansion model is compared with the real-time acquired multidimensional time-series dataset to calculate the model deviation. If the model deviation is greater than a preset threshold, the crossover and mutation operation and selection mechanism of the genetic algorithm are updated, and the population is optimized again to obtain a refined and optimized parameter combination. The refined and optimized parameter combination includes the adjusted heating temperature, holding time and cooling rate.

[0014] As a preferred embodiment of the present invention, step S5 includes: Control commands for the processing equipment are generated based on the refined and optimized parameter combination. The heating temperature, holding time and cooling rate in the actual processing process are adjusted. Data is collected in real time based on the adjusted processing process to determine the prediction accuracy of micropore defects and obtain a stable processing parameter set. The stable processing parameter set is used to guide the hot processing process of high-speed railway track steel components.

[0015] Secondly, the present invention also provides an intelligent monitoring system for the entire hot working process, for implementing the above-mentioned method, the system comprising: The defect determination unit is used to collect a multi-dimensional time-series dataset through a sensor array, extract spatiotemporal features based on the multi-dimensional time-series dataset through a neural network, obtain a defect initiation probability map by analyzing the spatiotemporal features, and determine potential defect areas based on the defect initiation probability map. The risk assessment unit is used to determine the optimal parameter combination based on the characteristics of the potential defect area, apply the optimal parameter combination to the processing simulation environment, obtain the defect type and occurrence probability based on the probability calculation model, and calculate the defect risk assessment value. The model building unit is used to predict the defect propagation path based on the defect risk assessment value and defect classification, and to build a dynamic defect propagation model based on the predicted defect propagation path and the multidimensional time series dataset. The parameter optimization unit is used to obtain predicted defect propagation data based on the dynamic defect expansion model and the collected data, compare the predicted defect propagation data with the real-time collected defect propagation data, update the genetic algorithm according to the comparison results, and determine the refined and optimized parameter combination. The equipment control unit is used to adjust the control commands of the processing equipment according to the combination of refining and optimization parameters to obtain a stable set of processing parameters.

[0016] Thirdly, the present invention also provides a computer-readable storage medium storing instructions that, when executed by a processor, implement the above-described method.

[0017] The technical solutions provided by the embodiments of the present invention may include the following beneficial effects: This invention uses a sensor array to collect multi-dimensional data such as temperature, stress, and cooling rate in real time during hot processing. A convolutional neural network is employed to extract spatiotemporal features, accurately capturing the changing patterns of temperature and stress fields and generating a defect initiation probability map to help locate potential defect areas and identify possible porosity defects. Based on the characteristics of these potential defect areas, a genetic algorithm is used to optimize processing parameters. Through a processing simulation environment, the optimized parameter combination is applied to the simulated processing to assess defect risk and probability of occurrence. A probability calculation model further refines the defect risk assessment, accurately predicting the types of defects and their likelihood of occurrence during processing. The optimized processing parameters effectively reduce defect risk and ensure processing quality. Combining the defect risk assessment results, a dynamic defect propagation model predicts the expansion path of defects. This model, by comparing real-time acquired data with the defect propagation mechanism, can not only predict the direction and speed of defect expansion but also provide guidance for early defect identification and handling. The model can dynamically adjust its defect propagation predictions, optimizing and adjusting the model based on actual data feedback to ensure prediction accuracy. By comparing predicted defect propagation data with real-time acquired defect propagation data, the genetic algorithm is updated and processing parameters are refined and optimized, ensuring the effectiveness of real-time parameter adjustments. This allows for real-time optimization of the processing process based on changing operating conditions, further reducing the probability of defects. Finally, control commands for the processing equipment are generated based on the refined and optimized parameters, adjusting heating, heat preservation, and cooling parameters during the actual processing to ensure precise control. Real-time data feedback continuously optimizes the processing process, ensuring the stability and consistency of processing parameters, thereby improving processing quality. Through the synergy of these technical solutions, intelligent monitoring and optimization of the hot processing process are achieved, enabling not only accurate prediction of defects during processing but also dynamic adjustment of processing parameters, significantly improving processing quality and product stability. Attached Figure Description

[0018] Figure 1 This is a flowchart of an intelligent monitoring method for the entire hot working process according to an embodiment of the present invention; Figure 2 This is a structural diagram of an intelligent monitoring system for the entire hot working process according to an embodiment of the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and thoroughly described below with reference to the accompanying drawings. The described embodiments are merely some embodiments of the present invention.

[0020] like Figure 1 This embodiment provides an intelligent monitoring method for the entire hot working process, which may specifically include: Step S1: Collect a multi-dimensional time-series dataset using a sensor array and extract spatiotemporal features using a neural network; wherein, in step S1, collecting the multi-dimensional time-series dataset using a sensor array includes: The heating temperature, holding time, cooling rate, material crystal structure, ambient humidity, and equipment vibration index of the hot-worked parts are collected in real time by a sensor array to obtain a time-series dataset containing multi-dimensional features. The multi-dimensional time-series dataset includes at least time-series data of the hot-worked temperature field, the stress field of the hot-worked parts, and environmental parameters. The multi-dimensional time-series dataset is preprocessed to obtain a standardized multi-dimensional time-series dataset.

[0021] Specifically, in one embodiment, by integrating a multi-type sensor array into the heat treatment equipment, comprehensive acquisition and processing of multi-dimensional time-series data of the heat treatment process can be achieved. The data collected by each type of sensor are arranged in chronological order to form a multi-dimensional time-series dataset of the heat treatment temperature field, the stress field of the heat-treated part, and environmental parameters. The heat treatment temperature field is obtained based on sensor data such as heating temperature, holding time, and cooling rate. Infrared temperature sensors collect the temperature distribution of the component in real time during the heating and cooling processes. The holding time recorded by the timer module is used to construct the time-series data of the isothermal stage. The cooling rate sensor or thermocouple monitors the cooling rate, which characterizes the dynamic changes of the temperature gradient during the cooling stage. The combination of these three can form a temperature field reflecting the temperature distribution and evolution of the component at different processing stages. The stress field of the heat-treated part is derived by combining the material crystal structure detection data and the temperature field data. An X-ray diffractometer monitors the material crystal of the component in real time. The phase transformation process of the structure, such as the transformation from austenite to martensite, combined with thermal conditions such as heating temperature and cooling rate, allows for the derivation of the stress distribution and changes caused by thermal expansion and contraction within the component, thereby constructing a stress field sequence for the hot-worked part. Environmental parameters are mainly acquired based on monitoring data from environmental humidity sensors and equipment vibration sensors. Environmental humidity sensors collect real-time data on changes in air humidity, reflecting the influence of humidity on the cooling rate and crystal structure stability. Equipment vibration sensors or accelerometers monitor the vibration intensity and frequency of the processing equipment during operation, characterizing the potential impact of external disturbances on processing stability and stress distribution. Together, they constitute the environmental parameter sequence. For example, during the heating stage, infrared temperature sensors can achieve second-level sampling and combine with vibration sensor data to capture dynamic changes, thereby ensuring a one-to-one correspondence between the time series information of temperature distribution gradient and internal stress distribution of the hot-worked part and the environmental humidity change sequence in the same dataset.After the aforementioned multidimensional data acquisition is completed, the multidimensional time-series dataset undergoes preprocessing. This preprocessing includes normalization or standardization to eliminate dimensional differences between different physical quantities. For example, min-max normalization maps temperature data from the actual range of 200 to 1000 degrees Celsius to a range of 0 to 1, or Z-score standardization converts the equipment vibration index sequence into a sequence with a mean of zero and a standard deviation of one. This ensures that data from different dimensions enters the subsequent algorithm processing module under a unified dimension. To address potential data gaps or noise disturbances, preprocessing also includes interpolation methods to fill in missing points and filtering methods to remove outliers. For example, when acquiring humidity data... Median filtering is used to eliminate transient interference, thus ensuring the continuity and stability of the final standardized dataset. The standardized multidimensional time-series dataset serves as the input data source for subsequent intelligent analysis. Different types of data maintain correspondence in both spatial and temporal dimensions, ensuring that temperature field changes, stress field responses, and environmental parameter fluctuations can be accurately identified and extracted by the convolutional neural network within a unified data framework. This technical solution, through multidimensional data acquisition and unified preprocessing using the aforementioned sensor array, ensures the temporal integrity and numerical comparability of the multi-source data input to the neural network, thereby providing a reliable data foundation for subsequent defect identification, risk assessment, and parameter optimization.

[0022] Further, in step S1, a neural network is used to extract spatiotemporal features, including: A multidimensional time-series dataset is input into a convolutional neural network, and the spatiotemporal distribution features of the temperature field and stress field are extracted. By analyzing the temperature field features and stress field features, the spatial temperature field gradient and stress field peak distribution are obtained. Based on the temperature field gradient and stress field peak distribution, a defect initiation probability map is generated. High-risk areas in the defect initiation probability map where the temperature gradient exceeds a preset threshold and overlaps with the stress peak area are identified as potential porosity defect initiation areas and are used as potential defect areas.

[0023] Specifically, in one implementation, the aforementioned multidimensional time-series dataset is formed into a unified input matrix on the time axis and input into a convolutional neural network. Multiple convolutional layers perform convolution operations on the input multidimensional time-series matrix to extract local spatiotemporal correlations. The convolutional kernel of each convolutional layer slides simultaneously in both spatial and temporal dimensions, thereby capturing the temperature gradient change pattern of the temperature field during the heating stage and the dynamic distribution pattern of the stress field during the cooling stage in the hot processing process. The convolutional output is represented as a preliminary feature map, which numerically corresponds to the coupling relationship between the temperature gradient and stress distribution over time and space. Subsequently, the network introduces a pooling layer to downsample the convolutional feature map. By using max pooling or average pooling, the dimensionality of the feature map is reduced while maintaining the key features, thereby reducing computational complexity and highlighting the persistence of the high-temperature region and the dynamic evolution characteristics of the stress concentration region during the processing. The pooled features numerically correspond to the temporal evolution trend of the temperature gradient peak and the stress concentration region.

[0024] After convolution and pooling, the neural network integrates features from different dimensions to form an abstract representation of the spatiotemporal distribution characteristics of the temperature and stress fields. This representation describes the spatial distribution pattern of temperature changes and the temporal evolution of stress fluctuations throughout the entire hot working process. Further analysis of these spatiotemporal distribution characteristics involves comparing the gradient value of the temperature field with a preset threshold. If the gradient value exceeds the threshold, it is marked as a high-risk area; otherwise, it is marked as a low-risk area. Simultaneously, it analyzes whether the peak region of the stress field overlaps with the high temperature gradient region to identify potential porosity initiation locations. The rate of change of the temperature field in space is calculated using the gradient calculation formula and combined with the local maximum value of the stress field to form a defect initiation probability map. This probability map shows areas where porosity may occur in the hot-worked part, such as areas with inhomogeneous material crystal structures. In this defect initiation probability map, high-risk areas where the temperature gradient exceeds the preset threshold and overlaps with the stress peak region are identified as potential porosity defect initiation areas. Each high-probability region in the probability map corresponds to the probability of defect occurrence, and areas exceeding the set threshold are selected as key monitoring areas. For example, in the hot processing scenario of high-speed rail track steel components, assuming that the multidimensional time series dataset is collected from the welding process, the heating temperature rises from 20 degrees Celsius to 800 degrees Celsius, the holding time is 30 minutes, and the cooling rate is 5 degrees Celsius per minute, the spatiotemporal distribution features extracted by the convolutional neural network show that the temperature field has a high gradient region at the weld joint, and the stress field has a peak in this region. Through analysis, it is determined that the potential porosity initiation region is located at the center of the joint, which helps to optimize processing parameters and reduce the probability of defects.

[0025] The above technical solution, through the multi-layer convolution, pooling, and integration mechanism of the convolutional neural network, establishes a mapping relationship between the input multi-dimensional time-series data in different dimensions and processing levels. That is, the original sensor data corresponds to the local spatiotemporal pattern in convolution processing, the compressed representation of key features in pooling processing, and the probability distribution of defect initiation regions in integration analysis. This enables real-time prediction and accurate identification of potential defects throughout the entire hot processing process, providing a scientific basis for the dynamic optimization and stability improvement of subsequent processing parameters.

[0026] Step S2: Determine the optimal parameter combination based on the characteristics of the potential defect area, apply the optimal parameter combination to the processing simulation environment, obtain the defect type and occurrence probability based on the probability calculation model, and calculate the defect risk assessment value; In step S2, determining the optimal parameter combination based on the characteristics of the potential defect area includes: If the temperature gradient in the potential defect area exceeds a preset threshold, the population is initialized using a genetic algorithm, a fitness function is defined, and the population is iteratively optimized according to the fitness function to obtain an optimized combination of heating temperature, holding time, and cooling rate. The fitness function is designed based on the temperature gradient and the probability of defect initiation and is used to evaluate the effectiveness of the processing parameters.

[0027] Specifically, in one implementation, the parameter optimization process is triggered by identifying the characteristics of potential defect areas. When the temperature gradient of the potential microscopic porosity defect initiation area (i.e., the aforementioned potential defect area) exceeds a preset threshold, a genetic algorithm is initiated to dynamically optimize the processing parameters. The temperature gradient is obtained based on spatiotemporal features extracted from a multidimensional time-series dataset. For example, if the temperature gradient calculated during the hot processing of the component is 55 degrees Celsius per centimeter and exceeds the preset threshold of 50 degrees Celsius per centimeter, it is automatically determined that the optimization process needs to be entered. First, the population is initialized using a genetic algorithm. Each individual corresponds to a combination of heating temperature, holding time, and cooling rate. The initial population size can be set to 100 individuals, each randomly generated within a certain numerical range, so that the initial population can cover multiple possible controllable parameter ranges of the processing technology. Subsequently, a fitness function is defined, which is jointly determined by the temperature gradient and the defect initiation probability. This fitness function is used to quantify the effectiveness of different parameter combinations in reducing defect risk. The temperature gradient is directly calculated from the aforementioned spatiotemporal features, while the defect initiation probability is obtained from the aforementioned defect initiation probability map. The genetic algorithm then iteratively optimizes the initial population, including selection, crossover, and mutation operations. Selection prioritizes individuals with higher fitness values ​​to ensure the optimization direction is correct. Crossover generates new individuals through parameter exchange to increase solution diversity. Mutation introduces random perturbations to individual parameters to avoid getting trapped in local optima. After multiple generations of iteration, the optimized parameter combination converges. Each iteration recalculates the fitness value of each individual. For example, a parameter combination of heating temperature 1000°C, holding time 30 minutes, and cooling rate 10°C per minute corresponds to a temperature gradient of 40°C per centimeter and a defect probability of 0.02, resulting in a calculated fitness value of 20.015. This value is higher than most individuals and is therefore prioritized for retention, gradually approaching the optimal solution during population evolution. The final optimized output might be a combination of heating temperature 950°C, holding time 25 minutes, and cooling rate 8°C per minute. This combination has been verified in simulations to significantly reduce the probability of porosity defects. For different types of hot-worked parts, the optimization logic can be adjusted according to the varying temperature sensitivities of the components. For example, in the machining of track beams, the temperature gradient threshold can be set to 45 degrees Celsius per centimeter, and the weight of the temperature gradient term in the fitness function can be increased to obtain optimization results that better reflect the characteristics of beam-type components. Furthermore, the fitness function can be extended to include a cooling rate influence factor. This design ensures that the cooling rate is directly coupled with the temperature gradient and defect probability during the evaluation process, thus allowing the optimization parameters to comprehensively reflect machining stability across multiple dimensions.The above technical solution, through the dynamic evolution and optimization of the genetic algorithm, can realize the logical closed loop between multi-source time-series data and processing parameters, so that the correspondence between different physical quantities is mapped to a unified evaluation value through the fitness function, and gradually converges to the optimal combination in the iterative evolution, thereby ensuring the intelligence and stability of the monitoring and optimization of the entire hot processing process in the overall technical solution.

[0028] Further, in step S2, the optimized parameter combination is applied to the processing simulation environment to obtain a defect risk assessment value, including: The optimized combination of heating temperature, holding time, and cooling rate is obtained and input into the processing simulation environment. The simulation processing process is executed, and the simulation results, simulation environment parameters, and material response characteristics during the simulation processing are input into the probability calculation model to calculate the defect type and the corresponding probability of occurrence. The defect type and the corresponding probability of occurrence are weighted to obtain the defect risk assessment value. The simulation results include the stress distribution and temperature change data of the hot-processed material.

[0029] Specifically, in this embodiment, the optimized parameter combination, namely the combination of heating temperature, holding time, and cooling rate, is first obtained through genetic algorithm optimization. For example, the heating temperature is 800 degrees Celsius, the holding time is 30 minutes, and the cooling rate is 5 degrees Celsius per minute. This combination is extracted and input as the initial conditions into the processing simulation environment. The above parameter combination is processed by simulation software to simulate the temperature and stress changes during the heating, holding, and cooling stages of the processing. Next, the simulation software executes the hot processing flow, simulates the thermal response of the material corresponding to the hot-processed part under given parameters, and calculates the probability of defects occurring. During the simulation, the stress distribution and temperature change data of the material are extracted from the simulation output, and the probability of crack defects and porosity defects is calculated using a probability calculation model based on finite element analysis. The material is divided into multiple small units using the finite element analysis method, and the types of defects that may occur in the material under different hot processing conditions and their probability of occurrence are evaluated by solving the stress and temperature equations of each unit. The overall defect risk assessment value is calculated by applying a weighted average formula to the occurrence probabilities corresponding to different crack types. The aforementioned probability calculation model is a calculation model trained with training data. The training data is multi-dimensional time-series data, including temperature field, stress field, environmental parameters, material properties of hot-processed parts such as thermal expansion coefficient and elastic modulus, response characteristics, and corresponding defect types and occurrence probabilities during the simulated heating process.

[0030] The aforementioned probabilistic calculation model can adjust to changes in external environmental conditions, such as the impact of ambient humidity on material response. When ambient humidity increases, the shrinkage effect during the cooling stage intensifies, leading to a higher probability of porosity. In this case, the probabilistic calculation model adjusts relevant parameters to adapt to humidity changes, thereby ensuring reliability during the hot working process. For example, when the cooling rate is adjusted to 10 degrees Celsius per minute, the probabilistic calculation model can display the probability of cracks caused by stress concentration due to rapid cooling and the overall risk value, further supporting refined adjustment of processing parameters and reducing the risk of defects. This technical solution not only achieves real-time simulation and defect prediction during the processing but also ensures the stability and quality of hot-processed components under complex working conditions by dynamically adjusting and optimizing processing parameters. This effectively improves the intelligent monitoring and optimization capabilities of the entire hot working process, ensuring precise adjustment of processing parameters and minimizing defect risks.

[0031] Step S3: Predict the defect propagation path based on the defect risk assessment value and defect classification, and construct a dynamic defect propagation model based on the predicted propagation path and the multidimensional time series dataset; Specifically, this includes: classifying defects according to defect risk assessment values ​​using a defect type classification method; predicting the direction and speed of defect propagation paths based on classification results and real-time data feedback; obtaining defect propagation mechanisms; and constructing a dynamic defect extension model through analysis of the multi-dimensional time-series dataset and the defect propagation mechanism. The dynamic defect extension model is used to characterize the evolutionary behavior of defects during the hot processing process.

[0032] Specifically, in this embodiment, microscopic porosity defects are classified according to their defect risk assessment values ​​using a defect type classification method. Specifically, based on the magnitude of the defect risk assessment value, microscopic porosity defects are first divided into low-risk and high-risk categories. Low-risk categories correspond to assessment values ​​below 0.5, while high-risk categories correspond to assessment values ​​above or equal to 0.5. This classification allows for the rapid identification of defect types with different risk levels. Based on the classification results, further analysis of multidimensional data (such as temperature field, stress field, and ambient humidity) locates the specific positions of potential porosity defects and marks their dimensions. These markings serve as input data for subsequent prediction models. Next, data such as heating temperature, cooling rate, ambient humidity, and equipment vibration during the processing are collected in real time and input into the prediction model. This real-time data allows for dynamic adjustments to the understanding of defect development. For example, when the cooling rate or ambient humidity changes, the path and speed of porosity or crack propagation may also change; the real-time data provides a basis for these changes in the model. Based on classification results and real-time data feedback, the propagation path and speed of pores or cracks during processing are predicted by analyzing the stress and temperature fields of the pores. This prediction is based on the defect propagation mechanism, namely the propagation law of pores in stress concentration areas. In this process, the possible propagation direction and speed of the pores are calculated according to the temperature changes and stress distribution around the pores, and the prediction results are adjusted according to environmental factors (such as humidity and cooling rate).

[0033] The construction of the dynamic defect propagation model further characterizes the evolutionary behavior of defects during processing by combining time series analysis and defect propagation mechanisms. Specifically, time series analysis techniques are applied to process multidimensional time series datasets to identify the spatiotemporal patterns of temperature and stress fields, thereby obtaining key data features during processing. Based on these data, combined with defect propagation mechanisms, such as the propagation law of pores in stress concentration areas, model equations are constructed to describe the change of defect size over time. The time series analysis results are then integrated with the defect propagation mechanism to form a dynamic model to simulate the complete process of defect initiation and propagation.

[0034] In one embodiment, the defect type classification method is implemented through threshold-based division. For example, when the defect risk assessment value is 0.3, it is classified as low-risk. This type of defect usually occurs in processing scenarios with short heat preservation times. The classification result helps to quickly screen out porosity or cracks that need to be monitored first, avoiding interruptions in the processing. In specific applications, such as in the processing of high-speed rail track steel components, under the conditions of heating temperature of 800 degrees Celsius and heat preservation time of 30 minutes, the classification result shows that the proportion of high-risk porosity increases. This helps to improve the accuracy of subsequent predictions. The prediction process is further optimized through real-time data feedback, taking into account the situation where the equipment vibration index exceeds 5 Hz, the propagation direction deflects by 10 degrees, and the speed increases by 15%, thereby reducing the risk of defect propagation and ensuring the stability of the track structure.

[0035] By incorporating a scenario with a cooling rate of 10 degrees Celsius per minute, the predictive model shows that the propagation path tilts towards the high gradient region of the stress field, with an expansion rate of 0.1 mm per hour. This prediction provides a basis for timely adjustment of processing parameters, ensuring the stability of the hot processing process. In step S3, the constructed dynamic defect propagation model combines time series analysis to extract features, such as using the Autoregressive Integrated Moving Average (ARIMA) method to process temperature data, and incorporates defect propagation mechanisms, such as the diffusion principle of pores at crystal boundaries, to characterize the evolution behavior of defects. In the simulation, when the ambient humidity is 60%, the model's prediction of the defect propagation path deviates by less than 5%, significantly improving the reliability of the prediction and reducing the probability of crack occurrence in actual processing.

[0036] For example, when the vibration index in the processing environment is 3 Hz, the model predicts a porosity expansion rate of 0.05 mm / h, and the deviation is 2% after comparison with real-time data, proving the effectiveness of the model. This process helps optimize the combination of heating temperature, holding time, and cooling rate to ensure the processing stability and safety of high-speed railway track steel components. Through this intelligent monitoring method, combined with real-time data and predictive models, the defect propagation path can be effectively predicted, and processing parameters can be dynamically adjusted, thereby minimizing the risk of defect occurrence and improving processing quality and efficiency.

[0037] Step S4: Based on the dynamic defect propagation model and collected data, obtain predicted defect propagation data, compare the predicted defect propagation data with the real-time collected defect propagation data, update the genetic algorithm according to the comparison results, and determine the refined optimization parameter combination; specifically including: The model bias is calculated by comparing the propagation path direction and velocity of micropores predicted by the dynamic defect extension model with the actual defect propagation data, i.e., the propagation direction and velocity. If the model bias is greater than a preset threshold, the crossover and mutation operation and selection mechanism of the genetic algorithm are updated, and the population is optimized again to obtain a refined and optimized parameter combination. The refined and optimized parameter combination includes the adjusted heating temperature, holding time and cooling rate.

[0038] Specifically, in this embodiment, the intelligent monitoring method compares a dynamic defect propagation model with a real-time acquired multidimensional time-series dataset to calculate model bias and adjust processing parameters. First, the model numerically matches the predicted micropore propagation path direction and velocity based on real-time acquired multidimensional time-series data (such as vibration, ambient humidity, and cooling rate) with the actual real-time defect propagation data, and calculates the average absolute error between the two over time as the model bias. This multidimensional time-series dataset originates from temperature and stress field data collected by a sensor array, providing accurate real-time input for subsequent model calculations.

[0039] If the calculated model bias is less than or equal to a preset threshold (e.g., 0.05), the current parameter combination remains unchanged without optimization. If the bias exceeds the threshold, the system updates the genetic algorithm, adjusting its crossover and mutation probabilities, and re-optimizes the population. Specifically, the crossover probability is increased from the initial 0.6 to 0.8 to enhance population diversity, while the mutation probability is increased from 0.01 to 0.05 to improve the accuracy of local searches. Furthermore, the selection mechanism is optimized by employing tournament selection to prioritize the retention of individuals with high fitness, ensuring the quality and stability of the optimization results.

[0040] In the updated genetic algorithm, optimized parameter combinations are applied to the hot working scenarios of hot-worked parts. For example, if the temperature gradient of potential micropore defects exceeds a set threshold, the genetic algorithm will simulate more possible parameter combinations by increasing the number of exchange points in the crossover operation to obtain more accurate optimization results. Optimized parameter combinations, such as a heating temperature of 1180 degrees Celsius, a holding time of 28 minutes, and a cooling rate of 4 degrees Celsius per minute, have been verified through simulations to effectively reduce the probability of defect occurrence. For example, in the simulation, the probability of crack defects decreased from 15% to 8%, thereby improving the overall durability of the rail steel components.

[0041] Furthermore, the real-time data feedback mechanism enables the model to be finely adjusted according to actual working conditions during processing. For example, in a certain scenario, if the actual cooling rate deviates from the model's prediction, the system will dynamically adjust the variation rate based on this deviation, increasing the scope of the local search and optimizing the parameter combination. In this scenario, the optimized parameter combination is a cooling rate of 3.5 degrees Celsius per minute, combined with a heating temperature of 1160 degrees Celsius and a holding time of 26 minutes. The final simulation results show that the probability of porosity defects is reduced to 5%, significantly suppressing defect propagation caused by rapid changes in microstructure.

[0042] When handling defect control during the high-temperature insulation stage, the system detects deviations between the preset model time and the actual operation. For example, if the insulation time data does not match, the model will readjust and optimize its strategy based on the deviation. By finely adjusting the insulation time to 24 minutes, the heating temperature to 1140 degrees Celsius, and the cooling rate to 2.8 degrees Celsius per minute, the model's defect prediction accuracy can reach 96%, further enhancing the stability of the hot working process and ensuring the processing quality of rail steel components under different humidity environments.

[0043] The above technical solution, by combining real-time data with a dynamic adjustment mechanism, establishes a precise feedback and optimization cycle, thereby achieving efficient defect prediction and processing parameter optimization throughout the entire hot processing process, ensuring the quality and stability of the final product.

[0044] Step S5: Adjust the control commands of the processing equipment according to the refined and optimized parameter combination to obtain a stable processing parameter set; specifically including: Control commands for the processing equipment are generated based on the refined and optimized parameter combination. The heating temperature, holding time and cooling rate in the actual processing process are adjusted. Data is collected in real time based on the adjusted processing process to determine the prediction accuracy of micropore defects and obtain a stable processing parameter set. The stable processing parameter set is used to guide the hot processing process of high-speed railway track steel components.

[0045] Specifically, in this embodiment, the intelligent monitoring method generates control commands based on refined and optimized processing parameters such as heating temperature, holding time, and cooling rate. It then adjusts various parameters of the hot processing process in conjunction with real-time data feedback to ensure the hot processing quality of high-speed railway track steel components. Specifically, based on the refined and optimized combination obtained by the genetic algorithm, the values ​​of heating temperature, holding time, and cooling rate are first extracted and converted into control commands executable by the processing equipment. During this process, the values ​​are mapped to the temperature controller of the heating furnace and the rate regulator of the cooling system through the equipment interface, thereby ensuring uniform heating during processing and reducing stress concentration and the initiation of porosity defects caused by excessive temperature gradients. Based on the optimized control commands, the settings of the processing equipment are adjusted in real time, for example, adjusting the heating temperature from the initial value of 1150 degrees Celsius to 1200 degrees Celsius to reduce the potential risk of microscopic porosity defects. Parameters such as holding time and cooling rate are also dynamically adjusted during this process.

[0046] During the adjusted processing, the sensor array continuously collects multi-dimensional time-series data, including heating temperature, holding time, cooling rate, ambient humidity, and equipment vibration. This data is stored and initially filtered for noise for subsequent data analysis and judgment. The real-time acquired data is processed by a convolutional neural network to extract the spatiotemporal characteristics of the temperature and stress fields, and potential porosity defect areas are identified through a prediction model. The prediction results are compared with actual observed defects to calculate the accuracy, thereby evaluating the effectiveness of the parameter settings during processing. When the prediction accuracy reaches a set threshold, such as 95%, the optimized parameter combination is confirmed to be effective and can significantly reduce the probability of defect occurrence.

[0047] Throughout this process, the data feedback mechanism is continuously optimized to ensure that processing parameters remain within a stable range. During multiple iterations, the model adjusts processing parameters based on the processing requirements of different steel components. For example, it adjusts control commands under varying cooling rates to ensure the stable and reliable processing of hot-worked parts, such as high-speed rail track steel components. Finally, after multiple verifications and adjustments, the resulting stable processing parameter set, including heating temperature range, holding time, and cooling rate, will be used to guide subsequent production batches. These stable processing parameter sets, through optimization and iterative verification, can significantly improve processing quality, reduce the impact of microscopic porosity defects on the safety of track components, and ensure the efficiency and stability of the hot-worked process in the production of hot-worked parts, such as high-speed rail track steel components.

[0048] This invention also provides an intelligent monitoring system for the entire hot working process, used to implement the above-mentioned method, such as... Figure 2 As shown, the system includes: The defect determination unit is used to collect a multi-dimensional time-series dataset through a sensor array, extract spatiotemporal features based on the multi-dimensional time-series dataset through a neural network, obtain a defect initiation probability map by analyzing the spatiotemporal features, and determine potential defect areas based on the defect initiation probability map. The risk assessment unit is used to determine the optimal parameter combination based on the characteristics of the potential defect area, apply the optimal parameter combination to the processing simulation environment, obtain the defect type and occurrence probability based on the probability calculation model, and calculate the defect risk assessment value. The model building unit is used to predict the defect expansion path based on the defect risk assessment value and defect classification, and to build a dynamic defect expansion model based on the predicted expansion path and the multidimensional time series dataset. The parameter optimization unit is used to obtain predicted defect propagation data based on the dynamic defect expansion model and the collected data, compare the predicted defect propagation data with the real-time collected defect propagation data, update the genetic algorithm according to the comparison results, and determine the refined and optimized parameter combination. The equipment control unit is used to adjust the control commands of the processing equipment according to the combination of refining and optimization parameters to obtain a stable set of processing parameters.

[0049] The present invention also provides a computer-readable storage medium storing instructions that, when executed by a processor, implement the above-described method.

[0050] In summary, this invention uses a sensor array to collect multi-dimensional data such as temperature, stress, and cooling rate in real time during the hot processing process, providing a comprehensive data foundation for subsequent defect prediction and processing parameter optimization. A convolutional neural network (CNN) is employed to extract spatiotemporal features, accurately capturing the changing patterns of the temperature and stress fields and generating a defect initiation probability map to help locate potential defect areas and identify possible porosity defect areas. Based on the characteristics of potential defect areas, a genetic algorithm is used to optimize processing parameters, such as heating temperature, holding time, and cooling rate. Through a processing simulation environment, the optimized parameter combination is applied to the simulated processing to assess defect risk and probability of occurrence. A probability calculation model further refines the defect risk assessment, accurately predicting the types of defects and their likelihood of occurrence during processing. The optimized processing parameters effectively reduce defect risk and ensure processing quality. Combining the defect risk assessment results, a dynamic defect propagation model predicts the defect propagation path. This model, by comparing real-time collected data with the defect propagation mechanism, can not only predict the direction and speed of defect propagation but also provide guidance for early defect identification and handling. This model can dynamically adjust the prediction of defect propagation, optimizing and adjusting the model based on actual data feedback to ensure prediction accuracy. By comparing predicted defect propagation data with real-time collected defect propagation data, the genetic algorithm is updated and processing parameters are refined and optimized, ensuring the effectiveness of real-time adjustment of processing parameters. This allows for real-time optimization of the processing process based on changes in operating conditions, further reducing the probability of defects. Finally, control commands for the processing equipment are generated based on the refined and optimized parameters, adjusting parameters such as heating, heat preservation, and cooling during the actual processing to ensure precise control. Through real-time data feedback, the processing process can be continuously optimized, ensuring the stability and consistency of processing parameters, thereby improving processing quality. Through the synergy of the above technical solutions, intelligent monitoring and optimization of the hot processing process are achieved, enabling not only accurate prediction of defects during processing but also dynamic adjustment of processing parameters, significantly improving processing quality and product stability.

[0051] Although the present invention has been described in detail above with general descriptions and specific embodiments, modifications or improvements can be made to it, which will be obvious to those skilled in the art. Therefore, all such modifications or improvements made without departing from the spirit of the present invention fall within the scope of protection claimed by the present invention.

Claims

1. An intelligent monitoring method for the entire hot working process, characterized in that, include: Step S1: Collect heating temperature, holding time, cooling rate, material crystal structure, ambient humidity, and equipment vibration index of the hot-worked part during the hot working process to obtain a time-series dataset containing multi-dimensional features. Input the multi-dimensional time-series dataset into a convolutional neural network and extract the spatiotemporal distribution features of the temperature field and stress field. By analyzing the temperature field features and stress field features, obtain the temperature field gradient and stress field peak distribution in space. Based on the temperature field gradient and stress field peak distribution, generate a defect initiation probability map. High-risk areas in the defect initiation probability map where the temperature gradient exceeds a preset threshold and overlaps with the stress peak area are identified as potential defect initiation areas and are used as potential defect areas. The multidimensional time-series dataset includes at least time-series data of the hot working temperature field, the stress field of the hot working part, and environmental parameters. The stress field of the hot working part is obtained through material crystal structure detection data and temperature field data. Step S2: If the temperature gradient in the potential defect area exceeds a preset threshold, the population is initialized using a genetic algorithm, a fitness function is defined, and the population is iteratively optimized according to the fitness function to obtain an optimized combination of heating temperature, holding time, and cooling rate. The fitness function is designed based on the temperature gradient and the probability of defect initiation. The optimized parameter combination is then applied to the processing simulation environment, and the defect type and occurrence probability are obtained according to the probability calculation model. Finally, the defect risk assessment value is calculated. Step S3: Based on the defect risk assessment value, classify the defects using the defect type classification method. Based on the classification results and real-time data feedback, predict the direction and speed of defect propagation path, obtain the defect propagation mechanism, and construct a dynamic defect expansion model through the analysis of the multi-dimensional time series dataset and the defect propagation mechanism. Step S4: Compare the dynamic defect expansion model with the real-time acquired multidimensional time series dataset to calculate the model deviation. If the model deviation is greater than a preset threshold, update the crossover and mutation operation and selection mechanism of the genetic algorithm, and re-optimize the population to obtain a refined and optimized parameter combination. The refined and optimized parameter combination includes the adjusted heating temperature, holding time and cooling rate. Step S5: Adjust the control commands of the processing equipment according to the combination of refining and optimization parameters to obtain a stable set of processing parameters.

2. The method as described in claim 1, characterized in that, In step S1, a multidimensional time-series dataset is collected, including: The heating temperature, holding time, cooling rate, material crystal structure, ambient humidity, and equipment vibration index of the hot-worked parts are collected in real time by a sensor array to obtain a time-series dataset containing multi-dimensional features. The multi-dimensional time-series dataset is then preprocessed to obtain a standardized multi-dimensional time-series dataset.

3. The method as described in claim 1, characterized in that, In step S2, the optimized parameter combination is applied to the processing simulation environment to obtain the defect risk assessment value, including: The optimized combination of heating temperature, holding time, and cooling rate is obtained and input into the processing simulation environment. The simulation processing process is executed, and the simulation results, simulation environment parameters, and material response characteristics during the simulation processing are input into the probability calculation model to calculate the defect type and the corresponding probability of occurrence. The defect type and the corresponding probability of occurrence are weighted to obtain the defect risk assessment value. The simulation results include the stress distribution and temperature change data of the hot-processed material.

4. The method as described in claim 1, characterized in that, Step S5 includes: Control commands for the processing equipment are generated based on the refined and optimized parameter combination. The heating temperature, holding time and cooling rate in the actual processing process are adjusted. Data is collected in real time based on the adjusted processing process to determine the prediction accuracy of micropore defects and obtain a stable processing parameter set. The stable processing parameter set is used to guide the hot processing process of high-speed railway track steel components.

5. An intelligent monitoring system for the entire hot working process, used to implement the method as described in any one of claims 1-4, characterized in that, The system includes: The defect determination unit is used to collect a multi-dimensional time-series dataset through a sensor array, extract spatiotemporal features based on the multi-dimensional time-series dataset through a neural network, obtain a defect initiation probability map by analyzing the spatiotemporal features, and determine potential defect areas based on the defect initiation probability map. The risk assessment unit is used to determine the optimal parameter combination based on the characteristics of the potential defect area, apply the optimal parameter combination to the processing simulation environment, obtain the defect type and occurrence probability based on the probability calculation model, and calculate the defect risk assessment value. The model building unit is used to predict the defect propagation path based on the defect risk assessment value and defect classification, and to build a dynamic defect propagation model based on the predicted defect propagation path and the multidimensional time series dataset. The parameter optimization unit is used to obtain predicted defect propagation data based on the dynamic defect expansion model and the collected data, compare the predicted defect propagation data with the real-time collected defect propagation data, update the genetic algorithm according to the comparison results, and determine the refined and optimized parameter combination. The equipment control unit is used to adjust the control commands of the processing equipment according to the combination of refining and optimization parameters to obtain a stable set of processing parameters.

6. A computer-readable storage medium storing instructions thereon, characterized in that, When the instructions are executed by the processor, they implement the method as described in any one of claims 1-4.

Images