Comprehensive multi-dimensional infrared heating automation control method and system

Abstract

This invention provides a comprehensive, multi-dimensional infrared heating automation control method and system, covering the field of infrared heating technology. This system acquires temperature and power data from infrared heating equipment, as well as workpiece surface temperature distribution data, performs feature extraction based on deep learning, constructs a dynamic compensation matrix, and establishes an adaptive heating optimization model to generate an optimal control strategy, dynamically adjusting lamp layout, power, and angle. This invention achieves precise control of infrared heating, improves heating uniformity, reduces energy consumption, and extends equipment life.

Description

Technical Field

[0001] The present invention relates to infrared heating technology, and in particular to a comprehensive multi-dimensional infrared heating automatic control method and system. Background Art

[0002] Infrared heating technology is a heating method that converts electrical energy into thermal energy using the principle of infrared radiation. It is widely used in industrial production, material processing, electronic manufacturing and other fields. As industrial manufacturing requires more precision control, infrared heating technology is gaining more and more attention and application due to its advantages such as non-contact, fast response, and high energy conversion efficiency. Traditional infrared heating control is mainly based on feedback control of a single temperature parameter. The temperature data of the heating area is collected by a temperature sensor, and the output power of the infrared lamp is adjusted according to a preset temperature curve. In recent years, with the development of artificial intelligence and deep learning technologies, intelligent infrared heating control systems have begun to be explored and applied in the industrial field. By introducing multi-source data analysis and model prediction functions, the accuracy and adaptability of the infrared heating process are improved.

[0003] However, existing infrared heating control technologies still have many shortcomings. First, most control systems only focus on single-dimensional temperature parameter monitoring, which makes it difficult to fully capture the complex heat transfer phenomena in the infrared heating process and the non-uniform distribution characteristics of the workpiece surface temperature, resulting in heating accuracy and consistency that are difficult to meet the requirements of high-precision manufacturing. Secondly, traditional control methods lack the ability to dynamically optimize the spatial layout and irradiation angle of infrared lamps, and are unable to adjust the irradiation strategy in real time according to the characteristics of workpieces of different shapes and materials, resulting in energy waste and unsatisfactory heating effects. Finally, existing systems generally lack self-learning and adaptive functions, and are unable to accumulate experience and optimize control strategies from historical heating processes. When faced with changes in production conditions or new workpieces, parameters need to be manually readjusted, which increases production preparation time and reduces production efficiency. Summary of the Invention

[0004] The embodiments of the present invention provide a comprehensive multi-dimensional infrared heating automatic control method and system, which can solve the problems in the prior art.

[0005] A first aspect of an embodiment of the present invention provides a comprehensive multi-dimensional infrared heating automatic control method, comprising:

[0006] Obtain real-time temperature data of infrared heating equipment, infrared lamp power data and surface temperature distribution data of heated workpieces;

[0007] Perform feature extraction on the real-time temperature data, the infrared lamp power data, and the surface temperature distribution data based on a deep learning algorithm to obtain a temperature feature vector, a power feature vector, and a temperature distribution feature vector of the infrared heating process;

[0008] Based on the temperature eigenvector, the power eigenvector and the temperature distribution eigenvector, an infrared heating dynamic compensation matrix is ​​constructed, and heating parameters of the infrared heating device are dynamically adjusted in real time by using the infrared heating dynamic compensation matrix;

[0009] collecting real-time heating state information of the heated workpiece, and establishing an adaptive heating optimization model based on the real-time heating state information. The adaptive heating optimization model, based on the deep learning algorithm, evaluates and optimizes the state-action sequence during the heating process to generate an optimal heating control strategy. The optimal heating control strategy is used to dynamically adjust the spatial layout, power output, and irradiation angle of the infrared lamps;

[0010] The optimal heating control strategy is converted into control instructions for infrared heating equipment, and automatic and precise control of the infrared heating process is achieved by executing the control instructions.

[0011] The real-time temperature data, the infrared lamp power data, and the surface temperature distribution data are subjected to feature extraction based on a deep learning algorithm to obtain a temperature feature vector, a power feature vector, and a temperature distribution feature vector of the infrared heating process. The feature vector includes:

[0012] Normalizing the real-time temperature data, the infrared lamp power data, and the surface temperature distribution data;

[0013] Constructing a multi-channel input deep learning network structure, the deep learning network structure includes a temperature feature extraction branch, a power feature extraction branch, and a temperature distribution feature extraction branch, inputting the normalized real-time temperature data into the temperature feature extraction branch, inputting the normalized infrared lamp power data into the power feature extraction branch, and inputting the normalized surface temperature distribution data into the temperature distribution feature extraction branch;

[0014] A channel attention mechanism is introduced into the deep learning network structure to adaptively assign attention weights according to the importance of features output by each feature branch, and feature fusion is performed on the outputs of the temperature feature extraction branch, the power feature extraction branch, and the temperature distribution feature extraction branch through the channel attention mechanism;

[0015] Based on the result of the feature fusion, a temperature feature vector, a power feature vector and a temperature distribution feature vector are generated respectively.

[0016] Constructing an infrared heating dynamic compensation matrix based on the temperature eigenvector, the power eigenvector, and the temperature distribution eigenvector includes:

[0017] Based on the temperature eigenvector, the power eigenvector, and the temperature distribution eigenvector, an adaptive weight allocation algorithm is used to construct a three-dimensional dynamic compensation matrix, wherein each dimension of the three-dimensional dynamic compensation matrix corresponds to a temperature compensation coefficient, a power compensation coefficient, and a temperature distribution uniformity compensation coefficient, respectively. The adaptive weight allocation algorithm dynamically adjusts the weight of each compensation coefficient according to real-time feedback data during the heating process;

[0018] The three-dimensional dynamic compensation matrix is ​​subjected to matrix operation with real-time heating parameters to construct an infrared heating dynamic compensation matrix.

[0019] Collecting real-time heating status information of the heated workpiece and establishing an adaptive heating optimization model based on the real-time heating status information includes:

[0020] Collecting real-time heating status information of the heated workpiece, the real-time heating status information including workpiece surface temperature distribution data, workpiece surface heat absorption rate data and workpiece material stress distribution data;

[0021] The real-time heating state information is constructed as a state vector, which includes a temperature state component, a heat absorption state component and a stress state component. An adaptive heating optimization model is established based on the mapping relationship between the state vector and the heating control action.

[0022] The adaptive heating optimization model is based on the deep learning algorithm and generates the optimal heating control strategy by evaluating and optimizing the state-action sequence during the heating process.

[0023] Acquire a state-action sequence during the heating process, wherein the state-action sequence includes workpiece heating state data and heating control action data;

[0024] Constructing a deep learning optimization model, the deep learning optimization model including a state evaluation network and an action generation network, the state evaluation network extracting features from the workpiece heating state data and outputting a state evaluation value, and the action generation network predicting the optimal heating control action at the next moment based on the state evaluation value;

[0025] A reward function is set for the state-action sequence, and network parameters of the deep learning optimization model are optimized and updated according to calculation results of the reward function to generate an optimal heating control strategy.

[0026] Converting the optimal heating control strategy into control instructions for infrared heating equipment, and implementing automated and precise control of the infrared heating process by executing the control instructions includes:

[0027] Converting the optimal heating control strategy into a device control instruction, wherein the device control instruction includes a power control instruction, an angle control instruction, and a position control instruction;

[0028] monitoring the execution process of the device control instruction according to the real-time execution status of the infrared heating device, and compensating and adjusting the execution parameters of the device control instruction when an execution deviation is detected;

[0029] The device control instruction after compensation adjustment is sent to the control unit of the infrared heating device for execution, thereby realizing automatic and precise control of the infrared heating process.

[0030] A second aspect of the present invention provides a comprehensive multi-dimensional infrared heating automation control system, including:

[0031] The first unit is used to obtain real-time temperature data of the infrared heating equipment, infrared lamp power data and surface temperature distribution data of the heated workpiece;

[0032] The second unit is used to perform feature extraction on the real-time temperature data, the infrared lamp power data and the surface temperature distribution data based on a deep learning algorithm to obtain a temperature feature vector, a power feature vector and a temperature distribution feature vector of the infrared heating process;

[0033] A third unit is configured to construct an infrared heating dynamic compensation matrix based on the temperature eigenvector, the power eigenvector, and the temperature distribution eigenvector, and to dynamically adjust the heating parameters of the infrared heating device in real time using the infrared heating dynamic compensation matrix;

[0034] A fourth unit is configured to collect real-time heating status information of the heated workpiece and establish an adaptive heating optimization model based on the real-time heating status information. The adaptive heating optimization model is based on the deep learning algorithm and generates an optimal heating control strategy by evaluating and optimizing the state-action sequence during the heating process. The optimal heating control strategy is used to dynamically adjust the spatial layout, power output, and irradiation angle of the infrared lamps.

[0035] The fifth unit is used to convert the optimal heating control strategy into control instructions for the infrared heating equipment, and realize automatic and precise control of the infrared heating process by executing the control instructions.

[0036] According to a third aspect of an embodiment of the present invention, an electronic device is provided, including:

[0037] processor;

[0038] a memory for storing processor-executable instructions;

[0039] The processor is configured to call the instructions stored in the memory to execute the aforementioned method.

[0040] According to a fourth aspect of an embodiment of the present invention, a computer-readable storage medium is provided, on which computer program instructions are stored. When the computer program instructions are executed by a processor, the method described above is implemented.

[0041] The beneficial effects of this application are as follows:

[0042] The comprehensive multi-dimensional infrared heating automation control method provided by the present invention uses a deep learning algorithm to extract features of temperature data, power data and temperature distribution data and construct a dynamic compensation matrix, thereby realizing precise regulation of heating parameters during the infrared heating process and effectively solving problems such as uneven temperature and low energy utilization in traditional control methods.

[0043] The adaptive heating optimization model established by the present invention can dynamically adjust the spatial layout, power output and irradiation angle of the lamp tube according to the real-time heating status information of the workpiece, significantly improving the heating uniformity, reducing energy waste, and enhancing the consistency and stability of product quality. It is particularly suitable for industrial production scenarios with high requirements for heating accuracy.

[0044] The present invention realizes closed-loop control and self-optimization of the heating process through multi-dimensional data fusion and intelligent decision-making mechanism. The system has strong environmental adaptability and anti-interference ability, can cope with changes in different workpiece materials, shapes and production requirements, improves production efficiency, reduces production costs, and provides a more intelligent solution for the application of infrared heating technology in the industrial field. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 Schematic diagram of the flow of a comprehensive multi-dimensional infrared heating automatic control method according to an embodiment of the present invention;

[0046] Figure 2 This is a flowchart of infrared heating multidimensional data feature extraction based on deep learning in an embodiment of the present invention;

[0047] Figure 3 This is a flow chart of infrared heating workpiece state information collection and optimization model construction in an embodiment of the present invention;

[0048] Figure 4 This is a flow chart of automatic control instruction conversion and execution for infrared heating equipment according to an embodiment of the present invention. DETAILED DESCRIPTION

[0049] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0050] The technical solution of the present invention is described in detail below with reference to specific embodiments. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described in detail in some embodiments.

[0051] Figure 1 FIG. 1 is a flow chart of a multi-dimensional infrared heating automatic control method according to an embodiment of the present invention. Figure 1 As shown, the method includes:

[0052] Obtain real-time temperature data of infrared heating equipment, infrared lamp power data and surface temperature distribution data of heated workpieces;

[0053] Perform feature extraction on the real-time temperature data, the infrared lamp power data, and the surface temperature distribution data based on a deep learning algorithm to obtain a temperature feature vector, a power feature vector, and a temperature distribution feature vector of the infrared heating process;

[0054] Based on the temperature eigenvector, the power eigenvector and the temperature distribution eigenvector, an infrared heating dynamic compensation matrix is ​​constructed, and heating parameters of the infrared heating device are dynamically adjusted in real time by using the infrared heating dynamic compensation matrix;

[0055] collecting real-time heating state information of the heated workpiece, and establishing an adaptive heating optimization model based on the real-time heating state information. The adaptive heating optimization model, based on the deep learning algorithm, evaluates and optimizes the state-action sequence during the heating process to generate an optimal heating control strategy. The optimal heating control strategy is used to dynamically adjust the spatial layout, power output, and irradiation angle of the infrared lamps;

[0056] The optimal heating control strategy is converted into control instructions for infrared heating equipment, and automatic and precise control of the infrared heating process is achieved by executing the control instructions.

[0057] In an optional embodiment, the real-time temperature data, the infrared lamp power data, and the surface temperature distribution data are subjected to feature extraction based on a deep learning algorithm to obtain a temperature feature vector, a power feature vector, and a temperature distribution feature vector of the infrared heating process, including:

[0058] Normalizing the real-time temperature data, the infrared lamp power data, and the surface temperature distribution data;

[0059] Constructing a multi-channel input deep learning network structure, the deep learning network structure includes a temperature feature extraction branch, a power feature extraction branch, and a temperature distribution feature extraction branch, inputting the normalized real-time temperature data into the temperature feature extraction branch, inputting the normalized infrared lamp power data into the power feature extraction branch, and inputting the normalized surface temperature distribution data into the temperature distribution feature extraction branch;

[0060] A channel attention mechanism is introduced into the deep learning network structure to adaptively assign attention weights according to the importance of features output by each feature branch, and feature fusion is performed on the outputs of the temperature feature extraction branch, the power feature extraction branch, and the temperature distribution feature extraction branch through the channel attention mechanism;

[0061] Based on the result of the feature fusion, a temperature feature vector, a power feature vector and a temperature distribution feature vector are generated respectively.

[0062] This embodiment provides a method for extracting features from real-time temperature data, infrared lamp power data, and surface temperature distribution data based on a deep learning algorithm. This method obtains the temperature feature vector, power feature vector, and temperature distribution feature vector of the infrared heating process for subsequent control decisions.

[0063] Normalize the collected real-time temperature data, infrared lamp power data, and surface temperature distribution data. For real-time temperature data, assuming the raw data range is 25°C to 350°C, normalization maps it to a range of 0 to 1. This is achieved by subtracting the minimum value of 25°C from each temperature value and then dividing it by the difference between the maximum and minimum values, 325°C. For example, if the temperature is 157°C, the normalized value is (157-25) / 325 = 0.406.

[0064] For infrared lamp power data, assuming the original power range is 0 to 5000 watts, normalization is similarly performed to map it to the range of 0 to 1. This is achieved by dividing each power value by the maximum power value of 5000 watts. For example, for a power value of 2750 watts, the normalized value is 2750 / 5000 = 0.55. For surface temperature distribution data, assuming a 320×240 pixel temperature matrix captured by an infrared thermal imager, with each pixel having a temperature range of 20°C to 400°C, normalization is performed by subtracting 20°C from each pixel's temperature value and then dividing by 380°C, resulting in a normalized temperature distribution matrix within the range of 0 to 1.

[0065] After normalization, a multi-channel input deep learning network structure is constructed, which includes a temperature feature extraction branch, a power feature extraction branch, and a temperature distribution feature extraction branch. The temperature feature extraction branch consists of five fully connected layers, with 64, 128, 256, 128, and 64 neurons in each layer, respectively, and uses the ReLU function as the activation function. The normalized real-time temperature data is organized into a time series format. For example, the temperature values ​​of 100 consecutive time points are taken to form an input vector, which is then input into the temperature feature extraction branch. The power feature extraction branch also consists of five fully connected layers, with 64, 128, 256, 128, and 64 neurons in each layer, respectively, and uses the ReLU function as the activation function.

[0066] The normalized infrared lamp power data is organized into a time series format. For example, the power values ​​at 100 consecutive time points are taken to form an input vector, which is then fed into the power feature extraction branch. The temperature distribution feature extraction branch uses a convolutional neural network structure consisting of four convolutional layers and two fully connected layers. The four convolutional layers have 32, 64, 128, and 256 convolution kernels, respectively. The kernel size is 3×3 with a stride of 1. Each convolutional layer is followed by a maximum pooling layer with a pooling window size of 2×2 and a stride of 2. The two fully connected layers have 512 and 256 neurons, respectively, and the ReLU function is used as the activation function. The normalized surface temperature distribution data is directly fed into the temperature distribution feature extraction branch.

[0067] To improve the efficiency and accuracy of feature extraction, a channel attention mechanism is introduced into the deep learning network structure, adaptively assigning attention weights based on the importance of the features output by each feature branch. The specific implementation process of the channel attention mechanism is as follows: the output features of the temperature feature extraction branch are subjected to global average pooling and global maximum pooling operations to obtain two feature descriptors respectively; these two feature descriptors are respectively passed through a multi-layer perceptron with shared weights. The multi-layer perceptron contains two fully connected layers. The first layer reduces the feature dimension to 1 / 16 of the original, and the second layer restores the feature to the original dimension. The two feature descriptors processed by the multi-layer perceptron are added together and mapped to the range of 0 to 1 using the Sigmoid function to obtain the channel attention weight of the temperature feature.

[0068] The channel attention weight calculation method for power and temperature distribution features is the same as that for temperature features. In this example, assuming the output feature dimensions of the temperature feature extraction branch, power feature extraction branch, and temperature distribution feature extraction branch are 64, 64, and 256, respectively, the attention weights calculated using the channel attention mechanism are 0.75, 0.60, and 0.82, respectively.

[0069] The outputs of the three feature extraction branches are fused using a channel-wise attention mechanism. Specifically, the output features of each feature branch are multiplied by the corresponding attention weights, and the three weighted feature vectors are then concatenated to form a fused feature vector. In this example, the fused feature vector has a dimension of 64 + 64 + 256 = 384. To further extract valid information, the fused feature vector is passed through a feature extraction network consisting of three fully connected layers, with 256, 128, and 64 neurons in each layer, respectively, and a ReLU activation function.

[0070] Based on the results of feature fusion, temperature feature vectors, power feature vectors, and temperature distribution feature vectors are generated. Specifically, the output of the feature extraction network is passed through three parallel fully connected layers to generate temperature feature vectors, power feature vectors, and temperature distribution feature vectors, respectively. In this example, the dimension of the temperature feature vector is 32, the dimension of the power feature vector is 32, and the dimension of the temperature distribution feature vector is 64. These feature vectors contain key characteristic information about temperature changes, power regulation, and surface temperature distribution during the infrared heating process, which can be used for subsequent heating control strategy optimization and temperature distribution prediction.

[0071] Through verification of experimental data, the feature vectors extracted by the above method can accurately reflect the temperature variation law, power regulation characteristics and surface temperature distribution characteristics of the infrared heating process, providing effective data support for achieving precise infrared heating control.

[0072] Figure 2 This is a flowchart of infrared heating multidimensional data feature extraction based on deep learning in an embodiment of the present invention:

[0073] This flowchart details the deep learning-based infrared heating data processing and feature extraction process. First, the three types of raw data collected by the system—real-time temperature data, infrared lamp power data, and surface temperature distribution data—are standardized and preprocessed to ensure data consistency and comparability. Subsequently, a deep learning network architecture with multiple feature extraction branches is constructed. This network comprises three independent feature extraction channels: a temperature feature extraction branch processes the normalized real-time temperature data, a power feature extraction branch processes the normalized lamp power data, and a temperature distribution feature extraction branch processes the normalized surface temperature distribution data. The network architecture also innovatively incorporates a channel attention mechanism, which adaptively assigns attention weights based on the importance of the features output by each feature branch. This weighted attention approach enables intelligent fusion of multi-channel features. Finally, based on the feature fusion results, the system generates temperature feature vectors, power feature vectors, and temperature distribution feature vectors. These feature vectors provide high-quality feature representations for subsequent heating control optimization. This deep learning-based multidimensional feature extraction scheme effectively captures key feature information in the infrared heating process, laying a solid data foundation for precise control.

[0074] In an optional embodiment, constructing an infrared heating dynamic compensation matrix based on the temperature eigenvector, the power eigenvector, and the temperature distribution eigenvector includes:

[0075] Based on the temperature eigenvector, the power eigenvector, and the temperature distribution eigenvector, an adaptive weight allocation algorithm is used to construct a three-dimensional dynamic compensation matrix, wherein each dimension of the three-dimensional dynamic compensation matrix corresponds to a temperature compensation coefficient, a power compensation coefficient, and a temperature distribution uniformity compensation coefficient, respectively. The adaptive weight allocation algorithm dynamically adjusts the weight of each compensation coefficient according to real-time feedback data during the heating process;

[0076] The three-dimensional dynamic compensation matrix is ​​subjected to matrix operation with real-time heating parameters to construct an infrared heating dynamic compensation matrix.

[0077] To collect temperature sensor data during the infrared heating process, a multi-point temperature acquisition device was used to place temperature sensors at different locations within the heating area. The temperature sensors were evenly distributed within the heating area in a 5×5 grid. Each sensor collected local temperature data in real time at a sampling frequency of 10 Hz. The collected temperature data formed a temperature feature vector T = [t1, t2, t3, ..., t25], where t1 to t25 represent the temperature values ​​at 25 measurement points in degrees Celsius.

[0078] Monitor the input power parameters of infrared heating equipment, including voltage, current, and their waveform characteristics. The power monitoring module collects voltage signals in the range of 0-380V and current signals in the range of 0-16A, with a sampling frequency of 50Hz. By processing the voltage and current data, it calculates the real-time power value and its rate of change, forming a power feature vector P = [p1, p2, p3, ..., p10], where p1 to p10 represent characteristic parameters such as average power, power fluctuation, power rise rate, and power fall rate within different time windows, respectively, and are expressed in watts or watts per second.

[0079] The temperature distribution characteristics of the heating area are calculated based on the collected temperature data. The temperature distribution characteristic calculation module first calculates basic parameters such as the mean temperature, maximum temperature, minimum temperature, and temperature gradient. The temperature gradient is calculated by calculating the temperature difference between adjacent temperature measurement points, forming 20 gradient values. The system also calculates the temperature standard deviation to indicate the uniformity of the temperature distribution. A smaller standard deviation indicates a more uniform temperature distribution. These characteristics are combined to form the temperature distribution feature vector D = [d1, d2, d3, ..., d30], where d1 to d30 include parameters such as the mean temperature, standard deviation, maximum temperature, minimum temperature, and temperature gradient in each direction.

[0080] An adaptive weight allocation algorithm is used to construct a three-dimensional dynamic compensation matrix, M. This matrix has three dimensions, corresponding to the temperature compensation coefficient, the power compensation coefficient, and the temperature distribution uniformity compensation coefficient. The core of the adaptive weight allocation algorithm is to dynamically adjust the compensation weights of each dimension based on real-time feedback data from the heating process. The system initializes the weights of the three dimensions to Wt=0.4, Wp=0.4, and Wd=0.2, respectively. These initial values ​​are empirically derived from a large number of heating experiments.

[0081] During the heating process, the system calculates the deviation between the current temperature and the target temperature. If the deviation exceeds 20 degrees Celsius, the system increases the weight of the temperature compensation dimension to Wt=0.6, Wp=0.3, and Wd=0.1 to speed up temperature adjustment. When the temperature approaches the target temperature (deviation less than 5 degrees Celsius), the system increases the weight of the temperature distribution uniformity compensation to Wt=0.3, Wp=0.3, and Wd=0.4 to ensure heating uniformity. If power fluctuations exceeding 10% are detected, the system increases the weight of the power compensation dimension to Wt=0.3, Wp=0.6, and Wd=0.1 to stabilize heating power.

[0082] Based on the weights, a three-dimensional compensation matrix M = [Mt, Mp, Md] is constructed, where Mt, Mp, and Md are the temperature compensation coefficient matrices, power compensation coefficient matrices, and temperature uniformity compensation coefficient matrices, respectively. Each compensation coefficient matrix in each dimension is a 5×5 matrix, corresponding to the grid layout of the heating area. Each element in the temperature compensation coefficient matrix Mt is calculated as the ratio of the actual temperature at that location to the target temperature, multiplied by a temperature adjustment factor of 0.8. Each element in the power compensation coefficient matrix Mp is calculated as the ratio of the required power at that location to the actual output power, multiplied by a power adjustment factor of 1.2. Each element in the temperature uniformity compensation coefficient matrix Md is calculated as the percentage deviation of the temperature at that location from the average temperature of the area.

[0083] A matrix operation is performed on the three-dimensional dynamic compensation matrix and the real-time heating parameters to construct the final infrared heating dynamic compensation matrix C. The operation is as follows: C = Wt × Mt + Wp × Mp + Wd × Md. "×" represents matrix multiplication by a scalar, and "+" represents matrix addition. The resulting compensation matrix C is a 5 × 5 matrix, corresponding to the comprehensive compensation coefficients at each position in the heating area.

[0084] During the actual heating process, the system uses the compensation matrix C to adjust the output power of each zone of the infrared heater. For example, when the compensation coefficient for a certain zone is 1.2, the heating power in that zone increases by 20%; when the compensation coefficient is 0.8, the heating power in that zone decreases by 20%. This dynamic compensation effectively addresses the heating needs of different materials and environmental conditions.

[0085] For example, a real-world heating process involves heating a ceramic substrate with a heating area of ​​1 square meter and a target temperature of 350°C. In the initial phase, temperature compensation takes precedence, allowing the system to rapidly increase the temperature. In the mid-stage, power compensation is weighted more heavily, stabilizing the system's output power. In the later stages, temperature uniformity compensation is weighted more heavily, and the system adjusts power across each area, reducing the standard deviation of the temperature distribution from an initial ±15°C to within ±3°C, achieving uniform heating. The entire heating process shortens the temperature rise time by 22% compared to traditional methods, reduces energy consumption by 18%, and improves temperature uniformity by 35%.

[0086] In an optional embodiment, collecting real-time heating status information of the heated workpiece and establishing an adaptive heating optimization model based on the real-time heating status information includes:

[0087] Collecting real-time heating status information of the heated workpiece, the real-time heating status information including workpiece surface temperature distribution data, workpiece surface heat absorption rate data and workpiece material stress distribution data;

[0088] The real-time heating state information is constructed as a state vector, which includes a temperature state component, a heat absorption state component and a stress state component. An adaptive heating optimization model is established based on the mapping relationship between the state vector and the heating control action.

[0089] The surface temperature distribution data of the heated workpiece is collected using a multi-point temperature sensor array. This array consists of 25 high-precision thermocouples arranged in a 5×5 matrix, covering key areas of the workpiece surface. Each sensor has a sampling frequency of 10 Hz, a measurement accuracy of ±0.5°C, and a temperature measurement range of 0–1200°C. The system records the temperature values ​​at each point in real time. For example, during the heating of a metal workpiece, the temperature in the center was 850°C, the temperature in the edge was 720°C, and the temperature in the corner was 680°C, thus generating a complete temperature distribution dataset.

[0090] An infrared thermal imager is used to obtain heat absorptivity data on the workpiece surface. The thermal imager uses a medium-wave infrared detector with a wavelength range of 8 to 14 μm, a resolution of 640 × 480 pixels, and a thermal sensitivity better than 0.05°C. The heat absorptivity is calculated by measuring the emissivity and reflectivity of the workpiece surface. For example, for a certain alloy material, the heat absorptivity is 0.85 in the center, 0.78 in the edge, and 0.72 in uneven areas. The system collects heat absorptivity data every 500 milliseconds and stores it synchronously with the temperature data.

[0091] Workpiece material stress distribution data is acquired using a strain gauge array and an acoustic emission sensor network. The strain gauge array consists of 16 gauges with a measurement range of ±5000με, a sensitivity of 2.1mV / V, and a linearity better than 0.1%. The acoustic emission sensors have a frequency response range of 100kHz to 1MHz and a sensitivity of 75dB. These sensors monitor changes in stress state on the workpiece surface and within it in real time. For example, during heating, a stress value of 240MPa was detected in a certain area, approaching 80% of the material's yield strength, indicating potential thermal stress risk in that area.

[0092] The collected real-time heating state information is constructed as a state vector. The temperature state component, T, is represented as a 25-dimensional vector, with each component corresponding to the value of a temperature sensing point. The heat absorption state component, A, is represented as a 16×16 matrix after surface meshing, with each element representing the heat absorption rate at the corresponding location. The stress state component, S, is represented as a three-dimensional tensor containing information on the magnitude and direction of the principal stresses.

[0093] After constructing the state vector, the system establishes a mapping relationship between the state vector and heating control actions. Heating control actions include adjusting heating power, heating position, and heating time. For example, if the temperature of a certain area of ​​the workpiece reaches 900°C and the heat absorption rate drops below 0.65, the system automatically reduces the heating power in that area by 25% and moves the heating focus to an adjacent area with a higher heat absorption rate.

[0094] The adaptive heating optimization model is built based on a deep reinforcement learning algorithm. It utilizes a three-layer neural network structure. The number of input layer nodes matches the dimensionality of the state vector, the hidden layer contains 128 neurons, and the output layer corresponds to the control action space. During model training, a reward function is set that comprehensively considers temperature uniformity, heating efficiency, and stress safety. For every 10% improvement in temperature uniformity, the reward increases by 0.5 points; for every 5% improvement in energy efficiency, the reward increases by 0.3 points; and for every 15% reduction in maximum stress, the reward increases by 0.8 points.

[0095] Through real-time state acquisition and model optimization, the system can dynamically adjust heating parameters. For example, during the heating process of a steel plate, the system initially uses 60% power for uniform heating. When it detects that the temperature at the edge is 120°C lower than that in the center, the system automatically increases the power to the edge to 85%. When the stress value in a local area exceeds the threshold of 200MPa, the system immediately reduces the heating power to 40% and adjusts the heating angle to achieve more even heat distribution.

[0096] To verify the effectiveness of the adaptive heating optimization model, the system was tested on a specific workpiece. An alloy plate with a length of 300 mm, a width of 200 mm, and a thickness of 25 mm was used, with an initial temperature of 25°C and a target heating temperature of 850°C. Under the traditional heating method, it takes 25 minutes for the workpiece to reach the target temperature, the temperature unevenness reaches ±85°C, and the maximum residual stress is 320 MPa. After adopting the adaptive heating optimization method of this embodiment, the heating time is shortened to 18 minutes, the temperature unevenness is reduced to ±35°C, the maximum residual stress is reduced to 160 MPa, and the energy consumption is reduced by 22%.

[0097] The system also incorporates a multi-level safety mechanism. If the temperature at any monitoring point exceeds the preset threshold of 1000°C, the system automatically reduces total power to 30%. If the stress exceeds 90% of the material's yield strength, the system immediately suspends heating in that area and initiates buffer cooling. If a sensor fails, the system automatically switches to a backup sensor and issues an alarm.

[0098] Figure 3 This is a flow chart of infrared heating workpiece status information collection and optimization model construction according to an embodiment of the present invention:

[0099] This flowchart details the process of collecting workpiece state information and constructing an optimization model during infrared heating. First, the system uses multiple sensors to perform comprehensive, real-time state monitoring of the heated workpiece, collecting data from three key dimensions: surface temperature distribution data, which reflects heating uniformity; surface heat absorption data, which indicates energy efficiency; and material stress distribution data, which reflects the internal stress state of the material. The system then integrates and transforms this collected real-time heating state information into a state vector containing multiple state components: a temperature state component, which characterizes the temperature field distribution; a heat absorption state component, which indicates energy transfer efficiency; and a stress state component, which characterizes the material stress state. Based on these state vectors, the system analyzes the correlation between state variables and heating control actions, establishing an adaptive heating optimization model that automatically optimizes heating parameters based on the workpiece's real-time state, achieving precise control. This adaptive optimization scheme, based on multi-dimensional state information, provides a reliable data foundation and model support for high-quality infrared heating control.

[0100] Through the above adaptive heating optimization method, the system can dynamically adjust the heating strategy according to the real-time status information of the workpiece to achieve the goals of precise temperature control, uniform heating, stress control and energy saving.

[0101] In an optional embodiment, the adaptive heating optimization model generates an optimal heating control strategy based on the deep learning algorithm by evaluating and optimizing the state-action sequence during the heating process, including:

[0102] Acquire a state-action sequence during the heating process, wherein the state-action sequence includes workpiece heating state data and heating control action data;

[0103] Constructing a deep learning optimization model, the deep learning optimization model including a state evaluation network and an action generation network, the state evaluation network extracting features from the workpiece heating state data and outputting a state evaluation value, and the action generation network predicting the optimal heating control action at the next moment based on the state evaluation value;

[0104] A reward function is set for the state-action sequence, and network parameters of the deep learning optimization model are optimized and updated according to calculation results of the reward function to generate an optimal heating control strategy.

[0105] Acquire the state-action sequences during the heating process. These sequences contain workpiece heating state data and heating control action data. The workpiece heating state data can include information such as the temperature distribution of the workpiece, heating time, and physical properties of the workpiece material. These data can be collected through multiple temperature sensors installed on the heating equipment. For example, during the heating process of a metal plate, 5 temperature measuring points can be installed on the surface of the plate, and temperature data can be collected every 2 seconds to record the temperature change curve. The heating control action data includes operating parameters such as heating power setting, heating area control, and heating time adjustment. For example, the process data of adjusting the heating power from 50 kilowatts to 65 kilowatts and the power distribution in different heating areas are recorded.

[0106] The constructed deep learning optimization model consists of two core components: a state assessment network and an action generation network. The state assessment network utilizes a multi-layer convolutional neural network architecture, consisting of four convolutional layers and two fully connected layers. The network receives workpiece heating status data as input, extracts spatial features of the temperature distribution through convolution operations, integrates temporal features through fully connected layers, and ultimately outputs a state assessment value. For example, a temperature distribution map of size 100×100×3 is input. After processing through the convolutional and pooling layers, 64 feature maps are extracted. These feature maps are then converted through fully connected layers into a state assessment value ranging from -1 to 1, which indicates the quality of the current heating state.

[0107] The action generation network uses a deep reinforcement learning architecture and consists of an input layer, three hidden layers, and an output layer. Each hidden layer contains 128, 256, and 128 neurons, respectively. Based on the state evaluation value output by the state evaluation network, the network predicts the optimal heating control action at the next moment. For example, for a device with four heating zones, the action generation network outputs a 4-dimensional vector representing the power adjustment value for each zone. For example, [10, -5, 8, 3] indicates that the power of the first zone is increased by 10 kilowatts, the power of the second zone is reduced by 5 kilowatts, and so on.

[0108] Setting a reward function for the state-action sequence is a key step in achieving adaptive optimization. The design of the reward function takes into account multiple factors: temperature uniformity, heating rate, energy consumption, and product quality requirements. In practical applications, the reward function can be defined as a weighted sum of these factors. For example, when temperature uniformity improves, the reward value increases; when energy consumption decreases, the reward value increases; and when the time to reach the target temperature is shortened, the reward value increases. Specifically, temperature uniformity can be evaluated by calculating the standard deviation of the temperature at each temperature measurement point on the workpiece surface. The smaller the standard deviation, the more uniform the temperature distribution; energy consumption is calculated by accumulating power usage; and the heating rate is evaluated by the time required to reach the target temperature.

[0109] Based on the calculated reward values, the system uses a policy gradient descent algorithm to update the network parameters of the deep learning optimization model. During training, the system performs multiple rounds of heating simulations or actual heating tests, calculating the cumulative reward and updating the network parameters after each round. For example, during a parameter update, if the cumulative reward value of the action sequence generated by the current policy is 85 points (out of 100), the system adjusts the network weights based on the gradient direction so that the action sequence generated in the next round will receive a higher reward value. By setting a learning rate of 0.001 and iterating through 500 rounds of training, the model parameters stabilized and the reward value increased to above 95 points, indicating that the model has learned an effective heating control strategy.

[0110] The resulting optimal heating control strategy is a mapping function from workpiece state to heating control actions. It outputs the optimal heating control parameters in real time based on the current workpiece heating state. In practical applications, this control strategy enables adaptive adjustments to the heating process. For example, if the temperature in a certain area of ​​the workpiece is detected to be too high, the heating power in that area will be automatically reduced; if the overall heating speed is detected to be too slow, the total power output will be appropriately increased. In this way, the system can maximize heating efficiency and reduce energy consumption while ensuring product quality.

[0111] After applying this method on the plate heating production line of a steel company, compared with the traditional fixed parameter control solution, product heating uniformity was improved by 18%, energy consumption was reduced by 12%, and heating time was shortened by 15%, significantly improving production efficiency and product quality.

[0112] In an optional embodiment, converting the optimal heating control strategy into control instructions for an infrared heating device, and implementing automated and precise control of the infrared heating process by executing the control instructions includes:

[0113] Converting the optimal heating control strategy into a device control instruction, wherein the device control instruction includes a power control instruction, an angle control instruction, and a position control instruction;

[0114] monitoring the execution process of the device control instruction according to the real-time execution status of the infrared heating device, and compensating and adjusting the execution parameters of the device control instruction when an execution deviation is detected;

[0115] The device control instruction after compensation adjustment is sent to the control unit of the infrared heating device for execution, thereby realizing automatic and precise control of the infrared heating process.

[0116] The process of converting the optimal heating control strategy into machine control instructions involves the generation of three core instructions. Power control instructions are calculated based on the target workpiece heating curve and material properties. The system first extracts the heat demand for each heating stage from the optimal strategy and, combined with the infrared heater's power conversion efficiency (typically 75%-85%), calculates the actual power output.

[0117] For example, for a heating task that requires heating an aluminum alloy workpiece from 25°C to 180°C within 240 seconds, the system generates a segmented power control curve: 0-60 seconds for the power ramp-up phase (20%-80%), 60-180 seconds for the constant power heating phase (80%), and 180-240 seconds for the power reduction phase (80%-30%), ensuring that the target temperature is accurately reached and avoiding temperature overshoot.

[0118] The generation of angle control instructions depends on the geometric characteristics of the workpiece surface and the requirements for irradiation uniformity. The system calculates the energy distribution density at different angles through the infrared radiation energy distribution model to determine the optimal irradiation angle of the infrared heater. For flat workpieces, the system usually sets the irradiation angle to be perpendicular to the surface; for curved workpieces, such as car door panels, the system will generate dynamic angle adjustment instructions to change the heater irradiation angle between 30°-60° to ensure that the energy distribution uniformity reaches more than 90%. The angle control instruction includes the initial angle value, the angle change rate and the angle limit value, and the accuracy is controlled within ±1°.

[0119] The position control instruction determines the spatial position of the infrared heater relative to the workpiece, including the horizontal position and the distance height. The system calculates the optimal irradiation distance and movement trajectory based on the size, shape and heating area of ​​the workpiece. For example, for a rectangular workpiece of 800mm×600mm, the system will generate a scanning position control instruction: at a height of 200mm, scan along the X-axis at a speed of 100mm / s, and offset the Y-axis by 50mm after each scan to ensure that the energy density deviation received in each area is controlled within ±5%. For fixed-point heating tasks, a fixed position instruction is generated and the optimal heating distance is specified (usually 1.5-3 times the diameter of the infrared lamp tube).

[0120] During the execution of device control commands, the system performs real-time monitoring and deviation compensation. The monitoring module collects the infrared heating device's actual execution parameters (power output, angular position, and spatial coordinates) and compares them with the command setpoints. If a deviation is detected that exceeds the allowable range, the compensation adjustment mechanism is triggered. For example, if the actual power output is more than 5% lower than the setpoint, the system calculates a compensation value based on the PID control algorithm and increases the power input command by a corresponding percentage.

[0121] For angle control, if the executed angle deviates by more than 2° from the commanded angle, the system sends an angle calibration command to drive the servo motor for fine-tuning. For position control, the system uses laser ranging feedback. If the position deviation exceeds 3mm, a position compensation command is immediately triggered to ensure the optimal irradiation distance between the heater and the workpiece.

[0122] After compensation and adjustment, the device control instructions are sent to the infrared heating device's control unit via a communication interface. The control unit then interprets the instructions into specific hardware drive signals: power control instructions are converted into PWM signals for the power regulation module, angle control instructions into rotation angle signals for the servo motor, and position control instructions into displacement pulse signals for the X, Y, and Z axis stepper motors.

[0123] The control unit continuously receives and executes updated instructions within a 10ms execution cycle, ensuring that the equipment's operating status remains highly consistent with the optimal heating control strategy. In a practical application, during the composite material curing process, the infrared heating control system implemented using this method can control the workpiece surface temperature within ±2.5°C of the set value, with a temperature uniformity of ±3°C, significantly better than the ±8°C error range of traditional control methods.

[0124] This technical solution achieves automated and precise control of the infrared heating process by converting abstract optimal heating control strategies into specific device executable instructions, and combining real-time monitoring and deviation compensation mechanisms. It can meet the needs of high-precision and high-uniformity heating and is suitable for advanced manufacturing processes in aerospace, automotive manufacturing, electronic packaging and other fields.

[0125] Figure 4 This is a flow chart of the conversion and execution of automatic control instructions for infrared heating equipment according to an embodiment of the present invention:

[0126] This flowchart illustrates the complete infrared heating control strategy conversion and execution process. First, the system converts the optimal heating control strategy into specific device control instructions. These instructions primarily encompass three core dimensions: power control instructions for precisely adjusting the infrared lamp's output power, angle control instructions for adjusting the lamp's illumination angle, and position control instructions for controlling the lamp's spatial placement. The system then monitors the infrared heating device's execution status in real time. By comparing actual performance with expected results, any deviations detected during execution trigger a compensation mechanism to dynamically adjust and optimize the device control instruction's execution parameters in real time. Finally, the compensated device control instructions are transmitted to the infrared heating device's control execution unit, which is responsible for executing the specific instructions. This enables automated and precise control of the entire infrared heating process, ensuring heating stability and uniformity. This hierarchical, closed-loop control scheme effectively improves the control accuracy and reliability of the infrared heating process.

[0127] A second aspect of the present invention provides a comprehensive multi-dimensional infrared heating automation control system, including:

[0128] The first unit is used to obtain real-time temperature data of the infrared heating equipment, infrared lamp power data and surface temperature distribution data of the heated workpiece;

[0129] The second unit is used to perform feature extraction on the real-time temperature data, the infrared lamp power data and the surface temperature distribution data based on a deep learning algorithm to obtain a temperature feature vector, a power feature vector and a temperature distribution feature vector of the infrared heating process;

[0130] A third unit is configured to construct an infrared heating dynamic compensation matrix based on the temperature eigenvector, the power eigenvector, and the temperature distribution eigenvector, and to dynamically adjust the heating parameters of the infrared heating device in real time using the infrared heating dynamic compensation matrix;

[0131] A fourth unit is configured to collect real-time heating status information of the heated workpiece and establish an adaptive heating optimization model based on the real-time heating status information. The adaptive heating optimization model is based on the deep learning algorithm and generates an optimal heating control strategy by evaluating and optimizing the state-action sequence during the heating process. The optimal heating control strategy is used to dynamically adjust the spatial layout, power output, and irradiation angle of the infrared lamps.

[0132] The fifth unit is used to convert the optimal heating control strategy into control instructions for the infrared heating equipment, and realize automatic and precise control of the infrared heating process by executing the control instructions.

[0133] According to a third aspect of an embodiment of the present invention, an electronic device is provided, including:

[0134] processor;

[0135] a memory for storing processor-executable instructions;

[0136] The processor is configured to call the instructions stored in the memory to execute the aforementioned method.

[0137] According to a fourth aspect of an embodiment of the present invention, a computer-readable storage medium is provided, on which computer program instructions are stored. When the computer program instructions are executed by a processor, the method described above is implemented.

[0138] The present invention may be a method, an apparatus, a system and / or a computer program product. The computer program product may include a computer-readable storage medium carrying computer-readable program instructions for executing various aspects of the present invention.

[0139] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A comprehensive multi-dimensional infrared heating automation control method, characterized in that: include: Obtain real-time temperature data of infrared heating equipment, infrared lamp power data and surface temperature distribution data of heated workpieces; Perform feature extraction on the real-time temperature data, the infrared lamp power data, and the surface temperature distribution data based on a deep learning algorithm to obtain a temperature feature vector, a power feature vector, and a temperature distribution feature vector of the infrared heating process; Based on the temperature eigenvector, the power eigenvector, and the temperature distribution eigenvector, an infrared heating dynamic compensation matrix is ​​constructed, and the heating parameters of the infrared heating equipment are dynamically adjusted in real time through the infrared heating dynamic compensation matrix, including: based on the temperature eigenvector, the power eigenvector, and the temperature distribution eigenvector, an adaptive weight distribution algorithm is used to construct a three-dimensional dynamic compensation matrix, wherein each dimension of the three-dimensional dynamic compensation matrix corresponds to a temperature compensation coefficient, a power compensation coefficient, and a temperature distribution uniformity compensation coefficient, respectively, and the adaptive weight distribution algorithm dynamically adjusts the weight of each compensation coefficient according to real-time feedback data during the heating process; Perform matrix operations on the three-dimensional dynamic compensation matrix and the real-time heating parameters to construct an infrared heating dynamic compensation matrix; collecting real-time heating status information of the heated workpiece, and establishing an adaptive heating optimization model based on the real-time heating status information, including: constructing a three-dimensional dynamic compensation matrix based on the temperature eigenvector, the power eigenvector, and the temperature distribution eigenvector using an adaptive weight allocation algorithm, wherein each dimension of the three-dimensional dynamic compensation matrix corresponds to a temperature compensation coefficient, a power compensation coefficient, and a temperature distribution uniformity compensation coefficient, respectively, and the adaptive weight allocation algorithm dynamically adjusts the weight of each compensation coefficient based on real-time feedback data during the heating process; Perform matrix operations on the three-dimensional dynamic compensation matrix and the real-time heating parameters to construct an infrared heating dynamic compensation matrix; The adaptive heating optimization model generates an optimal heating control strategy by evaluating and optimizing a state-action sequence during a heating process based on the deep learning algorithm, including: obtaining a state-action sequence during a heating process, wherein the state-action sequence includes workpiece heating state data and heating control action data; Constructing a deep learning optimization model, the deep learning optimization model including a state evaluation network and an action generation network, the state evaluation network extracting features from the workpiece heating state data and outputting a state evaluation value, and the action generation network predicting the optimal heating control action at the next moment based on the state evaluation value; Setting a reward function for the state-action sequence, optimizing and updating the network parameters of the deep learning optimization model according to the calculation result of the reward function, and generating an optimal heating control strategy; The optimal heating control strategy is used to dynamically adjust the spatial layout, power output and irradiation angle of the infrared lamps; The optimal heating control strategy is converted into control instructions for infrared heating equipment, and automatic and precise control of the infrared heating process is achieved by executing the control instructions.

2. The method according to claim 1, characterized in that The real-time temperature data, the infrared lamp power data, and the surface temperature distribution data are subjected to feature extraction based on a deep learning algorithm to obtain a temperature feature vector, a power feature vector, and a temperature distribution feature vector of the infrared heating process. The feature vector includes: Normalizing the real-time temperature data, the infrared lamp power data, and the surface temperature distribution data; Constructing a multi-channel input deep learning network structure, the deep learning network structure includes a temperature feature extraction branch, a power feature extraction branch, and a temperature distribution feature extraction branch, inputting the normalized real-time temperature data into the temperature feature extraction branch, inputting the normalized infrared lamp power data into the power feature extraction branch, and inputting the normalized surface temperature distribution data into the temperature distribution feature extraction branch; A channel attention mechanism is introduced into the deep learning network structure to adaptively assign attention weights according to the importance of features output by each feature branch, and feature fusion is performed on the outputs of the temperature feature extraction branch, the power feature extraction branch, and the temperature distribution feature extraction branch through the channel attention mechanism; Based on the result of the feature fusion, a temperature feature vector, a power feature vector and a temperature distribution feature vector are generated respectively.

3. The method according to claim 1, characterized in that Converting the optimal heating control strategy into control instructions for infrared heating equipment, and implementing automated and precise control of the infrared heating process by executing the control instructions includes: Converting the optimal heating control strategy into a device control instruction, wherein the device control instruction includes a power control instruction, an angle control instruction, and a position control instruction; monitoring the execution process of the device control instruction according to the real-time execution status of the infrared heating device, and compensating and adjusting the execution parameters of the device control instruction when an execution deviation is detected; The device control instruction after compensation adjustment is sent to the control unit of the infrared heating device for execution, thereby realizing automatic and precise control of the infrared heating process.

4. A comprehensive multi-dimensional infrared heating automation control system for implementing the method according to any one of claims 1 to 3, characterized in that: include: The first unit is used to obtain real-time temperature data of the infrared heating equipment, infrared lamp power data and surface temperature distribution data of the heated workpiece; The second unit is used to perform feature extraction on the real-time temperature data, the infrared lamp power data and the surface temperature distribution data based on a deep learning algorithm to obtain a temperature feature vector, a power feature vector and a temperature distribution feature vector of the infrared heating process; A third unit is configured to construct an infrared heating dynamic compensation matrix based on the temperature eigenvector, the power eigenvector, and the temperature distribution eigenvector, and to dynamically adjust the heating parameters of the infrared heating device in real time using the infrared heating dynamic compensation matrix; A fourth unit is configured to collect real-time heating status information of the heated workpiece and establish an adaptive heating optimization model based on the real-time heating status information. The adaptive heating optimization model is based on the deep learning algorithm and generates an optimal heating control strategy by evaluating and optimizing the state-action sequence during the heating process. The optimal heating control strategy is used to dynamically adjust the spatial layout, power output, and irradiation angle of the infrared lamps. The fifth unit is used to convert the optimal heating control strategy into control instructions for the infrared heating equipment, and realize automatic and precise control of the infrared heating process by executing the control instructions.

5. An electronic device, characterized in that: include: processor; a memory for storing processor-executable instructions; The processor is configured to call the instructions stored in the memory to execute the method according to any one of claims 1 to 3.

6. A computer-readable storage medium having computer program instructions stored thereon, characterized in that: When the computer program instructions are executed by a processor, the method according to any one of claims 1 to 3 is implemented.

Images