A dynamic temperature controlled injection molding process for high speed welding

Abstract

The application relates to the field of laser welding, and discloses a dynamic temperature control injection welding method for high-speed welding, which comprises the following steps: S1, integrating a fiber grating array and acquiring welding temperature data; S2, establishing a multi-physical field coupling model containing a heat conduction equation and a fluid dynamics equation, and simulating a temperature field; S3, constructing a control strategy based on reinforcement learning, and optimizing welding parameters; S4, dynamically calibrating temperature data measured by the fiber grating according to a real-time welding state to improve measurement accuracy; and S5, using model predictive control technology to perform real-time temperature prediction, and adjusting welding parameters according to a prediction result. The fiber grating array is used to monitor the temperature of a molten pool in real time, and a dynamic calibration technical scheme is combined, so that the effect of closed-loop control of the whole welding process is achieved. Compared with the prior art which depends on manual sampling and offline correction, the defects of temperature feedback lag and error accumulation are solved, and the molten pool is directly prevented from being overheated or from being insufficiently penetrated.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of laser welding, in particular to a dynamic temperature control injection welding method for high-speed welding. BACKGROUND

[0002] Currently, the quality control of laser welding process mainly relies on contact or non-contact temperature detection means such as thermocouples and infrared temperature measurement, and combines with heat conduction simulation model for process parameter presetting. In terms of control strategy, the traditional method mainly adopts PID control or expert system based on rule base to realize the steady-state regulation and control of the welding process through preset parameter combination. At the same time, for the calibration technology of temperature measurement error, the existing scheme usually corrects the measurement deviation caused by environmental interference based on constant compensation coefficient or offline calibration data.

[0003] However, with the development of high-power laser welding process towards high speed and high precision, the above-mentioned technologies gradually face challenges in practical application. For example, the traditional temperature detection means is difficult to capture the dynamic change of the molten pool in real time due to the limitation of sampling frequency, resulting in the hysteresis of temperature regulation; the heat conduction simulation model cannot accurately predict the weld appearance deviation caused by molten pool flow due to the uncoupling of fluid dynamics effect; and the PID control with fixed parameters is prone to parameter mismatching problem under the fluctuation of material properties or environmental interference, which needs frequent adjustment relying on human experience. In addition, the static calibration coefficient is difficult to adapt to the influence of dynamic variables such as protective gas flow field and substrate surface state in the welding process, resulting in the accumulation of measurement error and further causing welding defects. SUMMARY

[0004] In view of the deficiencies of the prior art, the present application provides a dynamic temperature control injection welding method for high-speed welding, which solves the problems of unstable welding quality and high defect rate caused by temperature monitoring hysteresis, molten pool dynamic behavior prediction error and process parameter solidification in the laser welding process.

[0005] To achieve the above purpose, the present application is realized by the following technical scheme: a dynamic temperature control injection welding method for high-speed welding, comprising the following steps:

[0006] S1, integrating a high-temperature-resistant fiber grating array at the head of the welding torch, acquiring temperature data in the welding process by monitoring the temperature change of the molten pool in real time;

[0007] S2, establishing a multi-physics field coupling model containing heat conduction equation and fluid dynamics equation, and simulating the temperature field;

[0008] S3, constructing a control strategy based on reinforcement learning, controlling the temperature of the welding process through state space, action space and reward function, and optimizing the welding parameters;

[0009] S4, dynamically calibrating the temperature data measured by the real-time welding state to improve the measurement accuracy;

[0010] S5, using model predictive control technology for real-time temperature prediction, and adjusting the welding parameters according to the prediction results.

[0011] Preferably, in the S1 step, the fiber grating array is at least one fiber grating with a spacing of 1mm and a sampling frequency of 8-12kHz to obtain high-resolution temperature data;

[0012] The real-time monitoring specifically includes the following steps:

[0013] Collecting real-time data of at least 50 temperature measurement points to form a temperature field data set;

[0014] Transmitting the collected temperature data to a central control unit for subsequent data processing and analysis.

[0015] Preferably, in the S2 step, the multi-physical field coupling model containing the heat conduction equation and the fluid dynamics equation includes the following steps in the establishment process:

[0016] Determining the thermal physical parameters of the welding material, including density, specific heat capacity and thermal conductivity;

[0017] Based on the thermal physical parameters of the welding material and the set welding conditions, constructing a numerical model of three-dimensional transient heat conduction equation and fluid dynamics equation;

[0018] The solving step of the numerical model includes discretizing the heat conduction equation and fluid dynamics equation using finite element method, and using iterative solution algorithm to obtain the distribution of temperature field.

[0019] Preferably, in the S3 step, the control strategy based on reinforcement learning includes the following steps:

[0020] Defining the state space, including real-time temperature data, welding speed and modal coefficient;

[0021] Defining the action space, including the adjustment range of laser power, welding speed and protective gas flow;

[0022] Designing a reward function with temperature error and temperature gradient as optimization target, which is used to guide the update of the control strategy;

[0023] The optimization process of the control strategy uses deep deterministic policy gradient algorithm to improve the temperature control accuracy of the welding process by iteratively updating the policy network and Q network.

[0024] Preferably, in the S4 step, the dynamic calibration step includes the following specific steps:

[0025] According to the temperature data measured by the real-time welding state collection fiber grating;

[0026] The temperature data is compared with the preset target temperature to calculate the temperature deviation;

[0027] The temperature data measured by the fiber grating is corrected by the calculated temperature deviation, thereby improving the accuracy of temperature measurement;

[0028] The process of correcting the temperature data adopts a weighted average method, and the measured temperature value is adjusted by a weighting factor to realize more accurate temperature data output.

[0029] Preferably, in the S5 step, the model predictive control includes the following specific steps:

[0030] According to the real-time temperature data and the historical temperature data, a temperature prediction model is established;

[0031] Using the temperature prediction model, the future temperature trend is predicted at the current time, and the temperature values at multiple future times are calculated;

[0032] Based on the prediction result, the laser power, welding speed and gas flow parameters in the welding process are adjusted;

[0033] The temperature prediction model adopts a recursive least squares method to update online according to the real-time temperature data, ensuring the reliability and accuracy of the prediction result.

[0034] Preferably, the boundary conditions of the heat conduction equation and the fluid dynamics equation of the multi-physical field coupling model are:

[0035] T(x,t)=T ambient In

[0036] Wherein, T ambient is the ambient temperature; is the temperature boundary domain.

[0037] Preferably, the state space definition includes a plurality of characteristics, specifically:

[0038] S={T measured ,v welding ,mode1,mode2,…,mode m};

[0039] Wherein, S is the state space; T measured is the measured temperature data; v welding is the welding speed; mode m is the mth welding mode characteristic.

[0040] Preferably, the dynamic calibration weight factor is calculated as:

[0041]

[0042] where w i is the weight factor of the i-th measurement point, used to give different importance to different measurement points when correcting the temperature data; T i is the actual measured temperature of the i-th measurement point; T target is the target temperature, i.e. the temperature value desired to be reached during the welding process; and ∈ is a small constant to avoid division by zero errors.

[0043] Preferably, the online update gain matrix of the temperature prediction model is:

[0044]

[0045] where K k is the gain matrix at the k-th iteration, used to adjust the weights of the prediction model; P k is the prediction error covariance matrix at the k-th iteration, representing a measure of the prediction uncertainty; Φ k is the current input vector at the k-th iteration, used for the state representation of the model; and R is the variance of the observation noise, representing the uncertainty in the observation data, which influences the calculation of the gain.

[0046] The present application provides a dynamic temperature control injection molding method for high-speed welding. It has the following

[0047] Advantages:

[0048] 1. The present application monitors the molten pool temperature in real time through the fiber grating array, and combines the dynamic calibration technical solution to achieve the effect of closed-loop control throughout the welding process. Compared with the existing technology which relies on manual sampling and offline correction, the defects of temperature feedback lag and error accumulation are solved, and the molten pool overheating or lack of penetration is directly avoided.

[0049] 2. The present application realizes accurate collaborative simulation of temperature field and flow field through multi-physical field coupling modeling of heat conduction equation and fluid dynamics equation. Traditional methods only consider a single factor of temperature, leading to a disconnection between process parameters and actual behavior of the molten pool, while the present application can accurately predict the molten pool fluctuation and optimize the weld forming quality.

[0050] 3. The present application uses the DDPG algorithm to optimize welding parameters in real time through the control strategy of reinforcement learning, solving the pain points of traditional PID control relying on experience parameter tuning and poor adaptability. The existing technology has a rigid response to complex working conditions, while the present application dynamically matches the best parameter combination through the self-learning mechanism, significantly improving the welding stability and consistency.

[0051] 4、The present application realizes the robustness calibration of anti-environmental interference through temperature deviation real-time correction and multi-parameter linkage feedback. The prior art uses fixed correction coefficient, which cannot adapt to welding transient change. The present application directly solves the one-sidedness of single parameter adjustment by dynamically adjusting the correction coefficient, and greatly reduces the defect rate of spatter, porosity and the like. BRIEF DESCRIPTION OF DRAWINGS

[0052] Figure 1 The method flowchart of the present application. DETAILED DESCRIPTION

[0053] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the specification of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0054] Please refer to the drawings in the specification of the present application Figure 1 , the embodiment of the present application provides a dynamic temperature control injection molding method for high-speed welding, comprising the following steps:

[0055] S1, a high-temperature-resistant fiber grating array is integrated at the head of the welding gun, and temperature data in the welding process is obtained by real-time monitoring of the temperature change of the molten pool.

[0056] Specifically, the head of the welding gun comprises a plurality of fiber grating units. The fiber grating array is arranged side by side with a spacing of 1mm, and the sampling frequency is set to 8-12kHz to ensure high resolution of real-time monitoring of the molten pool temperature. At the same time, the fiber grating array is connected with the central control unit to realize real-time transmission of data.

[0057] First, the temperature measurement data of the fiber grating unit is collected. The temperature measurement principle of the fiber grating is based on the Bragg grating effect, and the temperature information is obtained by the change of the reflected wavelength. The temperature data reflects the instantaneous temperature change of the molten pool in the welding process.

[0058] The construction process of the model is as follows. Set the temperature data as T i , which represents the temperature measurement value of the first T1 fiber grating. The temperature data set is defined by the following formula:

[0059] T data ={T1,T2,…,T n};

[0060] Where T data is the collected temperature data set; n is the number of measurement points, usually 50. The data set will be input to the central control unit for subsequent data processing and analysis.

[0061] Subsequently, the central control unit processes the received temperature data. By analyzing the temperature data set, temperature field data is calculated to form a complete temperature distribution model. The collected temperature data is subjected to time series analysis by the central control unit, and the data is optimized using the weighted average method. The weight w i is set as:

[0062]

[0063] where w i is the weight factor of the i-th measurement point, used to give different importance to different measurement points when correcting the temperature data; T i is the actual measured temperature of the i-th measurement point; T target is the target temperature, i.e. the temperature value desired to be reached during the welding process; and ∈ is a small constant to avoid division by zero error.

[0064] The corrected temperature data will be presented by the following formula:

[0065]

[0066] where T corrected is the corrected temperature output value, which is used for further welding process control.

[0067] Then, the processed temperature data is transmitted in real time to the welding control system, and is used for dynamic adjustment of welding parameters after confirming the validity of the data. The system can perform feedback control according to real-time temperature data, thereby optimizing the welding process and achieving the expected welding quality.

[0068] Through the above implementation steps, the fiber grating array integrated in the welding torch head can effectively monitor the temperature change of the molten pool, ensure high-precision control of the welding process, and has good practical application effect. The system optimally combines feedback mechanism and control strategy to improve the efficiency and quality of welding, and can realize real-time temperature control of the welding process.

[0069] S2, a multi-physical field coupling model containing heat conduction equation and fluid dynamics equation is established, and temperature field simulation is performed.

[0070] Specifically, in step S2, a multi-physical field coupling model containing heat conduction equation and fluid dynamics equation is established to perform temperature field simulation.

[0071] First, the thermal physical parameters of the welding material are determined, including density, specific heat capacity, thermal conductivity, and fluid dynamics parameters. The parameters are obtained through experimental measurement or material manual and input into the model construction module.

[0072] Subsequently, a three-dimensional transient heat conduction equation is constructed to describe the temperature variation in the molten pool region. The heat conduction equation is defined as:

[0073]

[0074] where T is the temperature field distribution; t is the time; is the laser input heat source; is the heat loss term, including radiation and convection heat dissipation.

[0075] At the same time, a fluid dynamics equation is established to simulate the flow behavior of the molten pool. The Navier-Stokes equation is used to describe the fluid motion:

[0076]

[0077] where u is the flow velocity vector; p is the pressure; v is the kinematic viscosity; g is the gravitational acceleration vector.

[0078] Further, the heat conduction equation and the fluid dynamics equation are discretized by the finite element method. The discretization process converts the continuous equation into a linear equation system, and an iterative solution algorithm (such as the conjugate gradient method) is used for numerical calculation to obtain the distribution results of the temperature field and the flow field.

[0079] Exemplarily, the boundary conditions are set as follows:

[0080] Temperature boundary condition: At the surface of the molten pool, the laser heat source input is set as a Gaussian distribution:

[0081]

[0082] where P laser is the laser power; r is the laser beam radius; σ is the standard deviation of the Gaussian distribution.

[0083] Fluid boundary condition: At the interface between the molten pool and the solid metal, the flow velocity is set to zero (no-slip boundary), and the free surface uses a surface tension model.

[0084] The calculation results of the model are output by the central control unit to form temperature field distribution maps and flow field dynamic maps. The output data are further used to guide the dynamic adjustment of the welding parameters, such as optimizing the welding path by predicting the temperature gradient.

[0085] The present embodiment realizes the collaborative simulation of the temperature field and the flow field through the multi-physical field coupling model, can accurately simulate the dynamic behavior of the molten pool in the high-speed welding process, and effectively improves the reliability of the welding quality control.

[0086] S3, a control strategy based on reinforcement learning is constructed to control the temperature of the welding process through the state space, action space and reward function, and optimize the welding parameters.

[0087] Specifically, in step S3, a reinforcement learning-based control strategy is constructed to control the temperature of the welding process and optimize the welding parameters through the state space, action space, and reward function.

[0088] First, the state space is defined, including real-time temperature data, welding speed, and modal coefficients. The mathematical representation of the state space is:

[0089] S = {T measured ,v welding ,mode1, mode2, …, mode m};

[0090] where S is the state space; T measured is the measured temperature data; v welding is the welding speed; mode m is the mth welding mode feature.

[0091] Subsequently, the action space is defined, including the laser power adjustment range, welding speed adjustment range, and shielding gas flow. The action is output by the control policy network and directly affects the welding equipment.

[0092] Further, the reward function is designed, with temperature error and temperature gradient as optimization objectives. The specific definition is:

[0093]

[0094] where α and β are weight coefficients used to balance the punishment intensity of temperature error and temperature gradient; is the temperature gradient, calculated from the temperature field dataset.

[0095] The control policy uses the deep deterministic policy gradient algorithm for optimization. The specific process is as follows:

[0096] Initialize the policy network and Q network: the network structure is fully connected layers, with the input layer dimension matching the state space and the output layer dimension matching the action space.

[0097] Interactive sampling: during the welding process, the policy network generates actions based on the current state, observes the next state after executing the action, and calculates the reward.

[0098] Experience replay: store the interaction data in the experience pool, and randomly extract batches of data for network training.

[0099] Network update: update the network parameters by minimizing the network loss function, and update the policy network parameters by policy gradient ascent. The Q network loss function is:

[0100]

[0101] Wherein, the target value y t The calculation is as follows:

[0102] y t =r t +γQ′(s t+1 ,μ′(s t+1 |θ μ′ )|θ Q′ );

[0103] Where γ is the discount factor; Q′ and μ′ are the parameters of the target network, which are periodically updated synchronously from the main network.

[0104] For example, the training frequency of the policy network is set to be updated once every 10 interactions, and the learning rate is preferably 0.001 to ensure convergence stability.

[0105] Finally, the optimized control strategy outputs welding parameter adjustment commands to adjust the laser power, welding speed, and shielding gas flow rate in real time.

[0106] S4. Dynamically calibrate the temperature data measured by the fiber Bragg grating based on the real-time welding status to improve measurement accuracy.

[0107] Specifically, in step S4, the temperature data measured by the fiber optic grating is dynamically calibrated to improve measurement accuracy, and accurate temperature feedback is ensured through real-time welding status.

[0108] First, during the welding process, real-time temperature data collected by a fiber Bragg grating array is input to the central control unit. This temperature data, based on the principle of Bragg gratings, monitors the temperature changes of the molten pool in real time, and the recorded temperature values ​​are in the following format:

[0109] T measured =f(T) raw );

[0110] Among them, T raw This is the uncalibrated raw temperature data, and f represents the calibration function.

[0111] Next, the target temperature is obtained. This is the ideal temperature value set during the welding process, usually determined by pre-established process parameters. Then, the deviation ΔT between the measured temperature and the target temperature is calculated, as expressed below:

[0112] ΔT=T measured -T target ;

[0113] Wherein, ΔT is the temperature deviation, indicating the difference between the actual measured temperature and the set target temperature.

[0114] To improve the measurement accuracy, the temperature deviation is used to correct the temperature data measured by the fiber Bragg grating. The correction formula uses the weighted average method, and the correction process can be described as:

[0115] T corrected = T measured -kΔT;

[0116] where T corrected is the corrected temperature value; k is the correction coefficient, representing the calibrated intensity, usually determined by historical data analysis or experiment to adapt to different welding conditions.

[0117] Next, the correction coefficient k in dynamic calibration is calculated as follows, by setting the reference temperature T baseline to determine the correction strategy, which can be expressed as:

[0118]

[0119] T baseline is the reference temperature under stable conditions; it can be obtained through long-term historical data or standard welding experiments.

[0120] Through the above formula, we can dynamically update the correction coefficient. When the initial conditions change in each welding cycle, the reference temperature T baseline may need to be re-evaluated to ensure the effectiveness of k.

[0121] After obtaining the corrected temperature value T corrected , it is passed to the welding control system, and the control system adjusts the laser power, welding speed and protective gas flow in the welding process according to these data, so as to optimize the welding quality and ensure the quality satisfaction. This control process can be expressed as:

[0122] P output =g(T corrected ,v welding ,Q gas );

[0123] where P output is the welding control instruction output according to the corrected temperature and other parameters;

[0124] g represents the function of generating control instructions, involving the comprehensive control strategy of welding equipment.

[0125] The whole process of dynamic calibration is: first, collect real-time temperature data, then calculate temperature deviation and correction coefficient, and finally output corrected temperature data to guide the operation of welding equipment. This process ensures the maximum accuracy of real-time temperature measurement, thereby improving the welding quality, and the weight factor calculation of dynamic calibration is:

[0126]

[0127] Among them, w i T is the weighting factor for the i-th measurement point, used to assign different importance to different measurement points when correcting temperature data; i T represents the actual measured temperature at the i-th measurement point. target is the target temperature, i.e., the temperature value that is desired to be achieved during the welding process; ∈ is a small constant to avoid division by zero errors.

[0128] S5. Real-time temperature prediction is performed using model predictive control technology, and welding parameters are adjusted based on the prediction results.

[0129] Specifically, in step S5, quality monitoring and optimization control of the welding process are implemented to ensure that the quality of the final welded joint meets the standard requirements.

[0130] First, key parameter data during the welding process are collected, including corrected temperature data, welding speed, laser power, and shielding gas flow rate. This data is sent to the monitoring system in real time for analysis to determine whether the welding process is within the set quality standards.

[0131] The setting of quality standard range should take into account material properties and process requirements, as follows:

[0132] Temperature standard range T min ≤T corrected ≤T max ;

[0133] Welding speed v min ≤v welding ≤v max ;

[0134] Laser power P laser,min ≤P laser ≤P laser,max ;

[0135] Next, the monitoring system uses statistical process control technology to analyze real-time data and identify potential quality problems. When a parameter deviates from the standard range, the system alerts the operator to make adjustments through an alarm mechanism.

[0136] The temperature deviation is calculated using the following formula:

[0137] ΔT=T corrected -T target ;

[0138] To monitor the difference between the actual welding temperature and the target temperature.

[0139] To optimize the welding process, the control system dynamically adjusts welding parameters based on feedback information. This control strategy includes:

[0140] Automated adjustment of welding parameters:

[0141] If ΔT > 0, reduce laser power to prevent overheating.

[0142] If ΔT < 0, increase laser power or adjust welding speed appropriately.

[0143] Closed-loop control mechanism:

[0144] Using feedback adjustment, ensure that the welding process remains in the best state, achieving real-time optimization of parameters.

[0145] Introduce temperature prediction model, so as to predict and optimize future welding process;

[0146] The online update gain matrix of the temperature prediction model is:

[0147]

[0148] Where K k is the gain matrix in the kth iteration, used to adjust the weight of the prediction model; P k is the prediction error covariance matrix in the kth iteration, indicating the measure of prediction uncertainty; Φ k is the current input vector in the kth iteration, used for the state representation of the model; R is the variance of the observation noise, indicating the uncertainty in the observation data, affecting the calculation of the gain.

[0149] Finally, the monitoring system archives all the data of the welding process, generates quality records and analysis reports, for subsequent traceability and process improvement.

[0150] Although embodiments of the present application have been shown and described, it will be understood by those having ordinary skill in the art that various changes, modifications, substitutions and alterations can be made thereto without departing from the principles and spirit of the present application, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A dynamic temperature controlled injection welding process for high speed welding, characterized in that, The method comprises the following steps: S1, integrating a high-temperature-resistant fiber grating array in the welding torch head, and obtaining temperature data during the welding process by monitoring the temperature change of the molten pool in real time; S2, establishing a multi-physics field coupling model containing heat conduction equation and fluid dynamics equation, and simulating the temperature field; S3, constructing a control strategy based on reinforcement learning, and controlling the welding process through state space, action space and reward function to optimize the welding parameters; S4, dynamically calibrating the temperature data measured by the fiber grating according to the real-time welding state to improve the measurement accuracy; S5, using model predictive control technology to perform real-time temperature prediction, and adjusting the welding parameters according to the prediction results; In the S4 step, the dynamic calibration step comprises the following specific steps: Collecting temperature data measured by the fiber grating according to the real-time welding state; Comparing the temperature data with the preset target temperature to calculate the temperature deviation; Using the calculated temperature deviation to correct the temperature data measured by the fiber grating, thereby improving the accuracy of temperature measurement; The process of correcting the temperature data adopts a weighted average method, which adjusts the measured temperature value through a weighting factor to realize more accurate temperature data output; In the S3 step, the control strategy based on reinforcement learning comprises the following steps: Defining the state space, including real-time temperature data, welding speed and modal coefficient; Defining the action space, including the adjustment range of laser power, welding speed and protective gas flow; Designing a reward function with temperature error and temperature gradient as optimization targets, which is used to guide the update of the control strategy; The optimization process of the control strategy adopts a deep deterministic policy gradient algorithm, which iteratively updates the policy network and Q network to improve the temperature control accuracy of the welding process.

2. A dynamic temperature controlled injection molding process for high speed welding as defined in claim 1, wherein, In the S1 step, the fiber grating array is at least one fiber grating with a spacing of 1mm and a sampling frequency of 8-12kHz to obtain high-resolution temperature data; The real-time monitoring specifically comprises the following steps: Collecting real-time data of at least 50 temperature measurement points to form a temperature field data set; Transmitting the collected temperature data to a central control unit for subsequent data processing and analysis.

3. A dynamic temperature controlled injection molding process for high speed welding as defined in claim 1, wherein, In the S2 step, the multi-physics field coupling model containing heat conduction equation and fluid dynamics equation comprises the following steps in the establishment process: Determining the thermal physical parameters of the welding material, including density, specific heat capacity and thermal conductivity; Based on the thermal physical parameters of the welding material and the set welding conditions, a numerical model of three-dimensional transient heat conduction equation and fluid dynamics equation is constructed; The solving step of the numerical model comprises discretizing the heat conduction equation and fluid dynamics equation using the finite element method, and obtaining the distribution of the temperature field by using an iterative solving algorithm.

4. A dynamic temperature controlled injection molding process for high speed welding as defined in claim 1, wherein, In the S5 step, the model predictive control comprises the following specific steps: According to the real-time temperature data and the historical temperature data, a temperature prediction model is established; Using the temperature prediction model, the future temperature trend is predicted at the current time, and the temperature values at multiple future times are calculated; Based on the prediction results, the laser power, welding speed and gas flow parameters in the welding process are adjusted; The temperature prediction model adopts a recursive least square method to perform online updating according to real-time temperature data, so as to ensure reliability and accuracy of the prediction result.

5. A dynamic temperature controlled injection molding process for high speed welding as defined in claim 2, wherein, Boundary conditions of the heat conduction equation and the fluid dynamics equation of the multi-physical field coupling model are: ； wherein is the ambient temperature; is the temperature boundary region.

6. A dynamic temperature controlled injection molding process for high speed welding as defined in claim 4, wherein, The state space definition comprises a plurality of characteristics, specifically: ； wherein, is a state space; is measured temperature data; is a welding speed; is a first welding mode feature.

7. A dynamic temperature controlled injection molding process for high speed welding as defined in claim 1, wherein, The weight factor of the dynamic calibration is calculated as: ； wherein is a weight factor for the i-th measurement point, used to give different importance to different measurement points when correcting the temperature data; is the actual measured temperature for the i-th measurement point; is the target temperature, i.e. the temperature value that is desired to be reached during the welding process; is a small constant to avoid division by zero errors.​​ 8. A dynamic temperature controlled injection molding process for high speed welding as defined in claim 4, wherein, The online updating gain matrix of the temperature prediction model is: ； wherein, is the gain matrix for adjusting the weights of the prediction model in the th iteration; is the prediction error covariance matrix in the th iteration, representing a measure of prediction uncertainty; is the current input vector in the th iteration, used for the state representation of the model; is the variance of the observation noise, representing the uncertainty in the observation data, influencing the calculation of the gain.

Images