Intelligent welding system

Abstract

The invention discloses an intelligent welding system which comprises a multi-dimensional sensing module, an intelligent control module and an execution module. The multi-dimensional sensing module collects welding current, interelectrode voltage and micro-vibration frequency spectrum of electrode contact in real time through a resistor and a vibration detection unit; the intelligent control module calculates the dynamic resistance and the attenuation rate thereof, the vibration energy entropy value and the oxide layer thickness, and predicts the residual life of the electrode by using a random forest algorithm; and meanwhile, a dynamic resistance-nugget growth kinetic model is established to estimate the nugget diameter, a self-adaptive optimizer is constructed in combination with a Q-learning algorithm, strategies such as dynamic resistance control and dynamic compensation are integrated, and the welding current and time are optimized in real time according to the nugget diameter and the electrode service life. The execution module executes welding parameters through a servo pressure system and a high-frequency inverter power supply, and the nugget size is accurately controlled; according to the method, the health state and the welding quality of the welding gun can be monitored in real time, dependence on experience of operators is reduced, predictive maintenance is achieved, and the overall efficiency is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of resistance welding, and particularly relates to an intelligent welding system. BACKGROUND

[0002] Resistance spot welding is a material joining process that is efficient, economical and widely used in the automotive, aerospace, household appliance and other industries. Its basic principle is to use the resistance heat generated by the current flowing through the contact surface and the adjacent area of the workpiece to locally heat it to a molten or plastic state, and at the same time, under the pressure of the electrode, a nugget is formed to achieve connection.

[0003] A typical resistance spot welding system includes a transformer, an electrode, a controller and other components. During welding, pressure is applied to the stacked metal sheets through the upper and lower electrodes, and a large current of several thousand to tens of thousands of amperes is passed. The current concentrates on the contact point between the sheets (i.e. the workpiece-workpiece contact surface), and due to the relatively high contact resistance at this point, a large amount of heat is rapidly generated under the Joule effect, forming a molten metal nugget. After power-off, the nugget cools and solidifies under pressure to form a firm weld.

[0004] Although the resistance spot welding technology is mature, in actual industrial mass production, especially in the face of the increasingly stringent requirements of modern manufacturing for high quality, high efficiency and low cost, its inherent technical bottlenecks and limitations are increasingly prominent, which are specifically manifested in the following aspects: 1. Electrode wear is uncontrollable, affecting production and cost: The electrode is subjected to the cyclic impact of high temperature, high pressure and strong current during the welding process. The electrode surface will undergo alloying, oxidation and thermal softening with the workpiece (especially galvanized sheet, high-strength steel, etc.), causing changes in its morphology (such as mushroom-shaped deformation) and an increase in contact resistance. This wear is a complex dynamic process that is difficult to accurately control. In order to ensure the quality of the weld, the electrode must be frequently ground to restore its end face shape, and the electrode cap is scrapped as a whole after a certain number of grinding times (for example, the electrode cap needs to be replaced with a new one after about 40 times of grinding, on average once every 200 spot welds), which not only increases the cost of electrode consumables, but also seriously reduces the overall efficiency of the production line due to frequent downtime for maintenance.

[0005] 2. Welding quality fluctuates greatly, consistency is difficult to guarantee: The core indicators of weld quality are nugget size and whether spatter occurs. However, the state of the workpiece surface (such as oil stains, oxides) and the thickness of the plating material (such as zinc, aluminum) inevitably fluctuate, which will significantly change the dynamic resistance of the workpiece-workpiece contact surface. When fixed welding parameters (current, time, pressure) are used, changes in dynamic resistance will cause unstable heat generation, which can easily result in insufficient nugget size development (false welding) or excessive heating leading to metal liquid spatter (spatter).

[0006] 3. Low energy utilization and significant energy waste: Traditional resistance welding control systems usually adopt constant "current-time" mode. This open-loop control method cannot perceive and adapt to the actual impedance changes caused by material properties, surface conditions and electrode wear during each welding. In order to ensure the formation of qualified nuggets under the worst working conditions, the process parameters are often set to be conservative, resulting in excessive energy supply in most welding cycles. Studies have shown that the effective utilization rate of electric energy in traditional resistance welding systems is generally less than 65%, and a large amount of electric energy is dissipated in the electrode, workpiece and circuit in the form of invalid heat, causing serious energy waste.

[0007] 4. Maintenance strategy lags behind and lacks predictive management capability: Currently, the replacement and maintenance of electrodes mainly rely on the experience of operators or simple welding cycle counting, and the actual health status of the electrode cannot be accurately predicted. This passive maintenance mode cannot predict sudden failures of the electrode (such as cracking and severe adhesion), which can easily lead to batch quality accidents or non-planned production line shutdowns due to sudden electrode failure during continuous production, resulting in significant economic losses.

[0008] In summary, the existing resistance spot welding technology has obvious shortcomings in stability, efficiency, energy consumption and maintenance due to factors such as electrode wear and workpiece state changes. Therefore, an intelligent control method and system that can perceive the welding process state in real time, adaptively adjust parameters and predict electrode life is urgently needed to break through the above technical bottlenecks and achieve intelligent welding production with high quality, high efficiency and low energy consumption. SUMMARY

[0009] The present application provides an intelligent welding system, comprising: a multi-dimensional sensing module, an intelligent control module and an execution module; The multi-dimensional sensing module is used to collect the welding current and the voltage between the electrodes through the resistance detection unit, and to detect the micro-vibration frequency spectrum of the electrode contact instant through the vibration sensor; The intelligent control module is used to calculate the dynamic resistance and the dynamic resistance decay rate according to the welding current and the voltage between the electrodes, to calculate the vibration energy entropy value and the thickness of the electrode surface oxide layer according to the micro-vibration frequency spectrum; to construct an electrode monitoring degree prediction model based on the random forest algorithm to output the electrode remaining life according to the cumulative welding times, the dynamic resistance decay rate, the vibration energy entropy value and the thickness of the electrode surface oxide layer; to construct a dynamic resistance-nugget growth kinetics model to calculate the nugget diameter according to the dynamic resistance, to construct a dynamic resistance control strategy and a dynamic compensation control strategy, to construct an adaptive welding parameter simulation optimizer based on the Q-learning learning algorithm, and to dynamically adjust the welding current and the welding time according to the dynamic resistance, the nugget diameter and the electrode remaining life through the dynamic resistance control strategy, the dynamic compensation control strategy and the adaptive welding parameter optimizer; The execution module is configured to perform welding according to the welding current and the welding time through the servo pressure control system and the high-frequency inverter power supply to reach the target nugget diameter.

[0010] Further, the dynamic resistance and the dynamic resistance decay rate are calculated according to the welding current and the voltage between the electrodes, including: The dynamic resistance is calculated according to the welding current and the voltage between the electrodes, as shown in formula (1), (1) wherein, is a function of the resistance changing with time, and the value is the dynamic resistance, is the voltage between the electrodes, is the welding current, is the total time of the welding process; The dynamic resistance sequence in a single welding process is set as , and the fitting curve slope is calculated as the dynamic resistance decay rate, as shown in formula (2), (2) wherein, is the dynamic resistance decay rate, is the number of dynamic resistances, is the sampling point, is the dynamic resistance of the sampling point.

[0011] Further, the vibration energy entropy value and the thickness of the oxide layer on the electrode surface are calculated according to the micro-vibration spectrum, including: The frequency band energy distribution is obtained by wavelet packet decomposition of the vibration signal in the micro-vibration spectrum, and the vibration energy entropy value is calculated, as shown in formula (3), (3) wherein, is the vibration energy entropy value, is the decomposition layer number, is the energy of the th frequency band, is the percentage of the energy of the th frequency band in the total signal energy, is the energy of the th frequency band; A linear relationship between the main frequency offset and the oxide layer thickness based on the vibration signal is constructed to obtain the thickness of the oxide layer on the electrode surface, as shown in formula (4), (4) wherein, is the thickness of the oxide layer on the electrode surface, and are undetermined coefficients of the linear model, a frequency offset of the main frequency of the vibration signal, a current main frequency, an initial main frequency.

[0012] Further, an electrode monitoring degree prediction model is constructed based on a random forest algorithm to output the electrode remaining life according to the cumulative welding times, dynamic resistance decay rate, vibration energy entropy value and electrode surface oxide layer thickness, including: The cumulative welding times, dynamic resistance decay rate, vibration energy entropy value and electrode surface oxide layer thickness are taken as a training set, K sample subsets are randomly extracted from the training set with replacement and input into the random forest model for training, the MSE criterion is used for splitting, and the splitting is stopped until the set maximum depth is reached, to obtain the electrode monitoring degree prediction model, as shown in formula (5), (5) wherein, the electrode remaining life, the number of decision trees, , the cumulative welding times, the dynamic resistance decay rate, the vibration energy entropy value, the electrode surface oxide layer thickness.

[0013] Further, a dynamic resistance-nugget growth kinetics model is constructed to calculate the nugget diameter according to the dynamic resistance, including: The integral area of the dynamic resistance curve is constructed, as shown in formula (6), (6) wherein, the integral area of the dynamic resistance curve in the welding process, a function of the resistance changing with time, , the voltage between the electrodes, the welding current, the total time of the welding process; A nonlinear model of the nugget diameter and the integral area of the dynamic resistance curve is constructed, as shown in formula (7), (7) wherein, the nugget diameter, , and the workpiece material related coefficient.

[0014] Further, an adaptive welding parameter optimizer is constructed based on a Q-learning learning algorithm, including: Step one: initialize Q table, define state space, action space and reward function, the state space includes dynamic resistance average, plate thickness combination and electrode remaining life; the action space is to adjust the welding current and pressurization time , the reward function is shown in formula (8), (8) wherein, is the target nugget diameter; Step two: determine the current state , select an action according to the ε-greedy strategy, that is, select a set of adjusted welding current values and pressurization time values; Step three: execute the action, use the new parameters for welding; measure the new dynamic resistance average after welding, and calculate the nugget diameter and the new reward function, and then determine the new state ; Step four: update the Q table using step two until the Q table converges, and obtain the trained Q table, that is, the adaptive welding parameter optimizer.

[0015] Further, the dynamic resistance control strategy includes: derivative method is used to calculate the resistance change rate of dynamic resistance , if the rising rate of exceeds the preset value, the welding current is increased by w%; if the average value of is lower than the reference value within a certain time, the welding is stopped and the electrode is dressed.

[0016] Further, the dynamic compensation control strategy includes: preset peak time deviation according to the material type and plate thickness combination; set the point of as the peak point of dynamic resistance, record the appearance time of the peak point , calculate the peak time deviation of from the theoretical peak time, if , adjust the welding current through the current adjustment formula and adjust the pressurization time through the pressurization time adjustment formula; the current adjustment formula is shown in formula (9), (9) wherein, is the adjusted welding current, is the welding current before adjustment, is the preset gain coefficient; the pressurization time adjustment formula is shown in formula (10), (10) in, The adjusted pressurization time. The pressurization time before adjustment. The gain coefficient is the preset pressurization time.

[0017] Furthermore, the welding current and welding time are dynamically adjusted based on the dynamic resistance and weld nugget diameter using a dynamic resistance control strategy, a dynamic compensation control strategy, and an adaptive welding parameter optimizer, including: The initial welding current and pressurization time are obtained by an adaptive welding parameter optimizer based on the dynamic resistance average, actual weld nugget diameter, plate thickness combination, and remaining electrode life. Welding is performed using the initial welding current and pressure application time. The resistance change rate of the dynamic resistance is detected during the welding process. If the rate of increase of the resistance change rate exceeds the preset value or the average value within a certain period of time is lower than the reference value, the initial welding current and pressure application time are adjusted using the dynamic resistance control strategy to obtain the welding current and pressure application time after secondary adjustment. Welding is performed using a secondary-adjusted welding current and pressurization time. The peak delay of the dynamic resistance is detected. If the peak delay is greater than the set time deviation, a dynamic compensation control strategy is adopted to readjust the secondary-adjusted welding current and pressurization time to obtain the final welding current and pressurization time.

[0018] Beneficial effects: This invention provides an intelligent welding system with the following advantages: 1. By constructing an electrode monitoring degree prediction model based on the random forest algorithm, the electrode life and wear degree can be predicted, and the electrodes can be repaired or replaced in a timely manner to control the wear of the electrodes within a certain range and improve the overall efficiency of the production line. 2. By constructing multiple adjustment strategies, the welding current and welding time are dynamically adjusted according to the dynamic resistance and weld nugget diameter to ensure welding quality and consistency; 3. The execution module, combined with the intelligent control module, can calculate the heat demand in real time based on the dynamic resistance and accurately control the energy input, avoiding energy redundancy set up to ensure the worst operating conditions.

[0019] 4. By using multiple sensors, the health status and welding quality of the welding torch can be monitored in real time, reducing reliance on operator experience, enabling predictive maintenance, and improving overall efficiency. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This invention provides a system block diagram of an intelligent welding system. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] This embodiment provides an intelligent welding system, such as Figure 1 As shown, it includes: a multi-dimensional sensing module, an intelligent control module, and an execution module; The multidimensional sensing module is used to collect welding current and voltage between electrodes through a resistance detection unit, and to detect the micro-vibration spectrum at the moment of electrode contact through a vibration sensor. The intelligent control module is used to calculate the dynamic resistance and dynamic resistance attenuation rate based on the welding current and the voltage between the electrodes, and to calculate the vibration energy entropy and the oxide layer thickness on the electrode surface based on the micro-vibration spectrum. It constructs an electrode monitoring prediction model based on the random forest algorithm to output the remaining electrode lifespan based on the cumulative welding count, dynamic resistance attenuation rate, vibration energy entropy, and oxide layer thickness. It also constructs a dynamic resistance-melt nugget growth dynamics model to calculate the melt nugget diameter based on the dynamic resistance, and builds dynamic resistance control strategies and dynamic compensation control strategies. An adaptive welding parameter simulation optimizer is constructed based on the Q-learning algorithm. Through the dynamic resistance control strategy, dynamic compensation control strategy, and adaptive welding parameter optimizer, the welding current and welding time are dynamically adjusted based on the dynamic resistance, melt nugget diameter, and remaining electrode lifespan. The execution module is used to perform welding according to the welding current and welding time through a servo pressure control system and a high-frequency inverter power supply to achieve the target weld nugget diameter.

[0024] Specifically, the technical solution provided in this application can be used to solve the defects in existing plate welding processes. In this process, two workpiece plates are welded together, with electrodes divided into upper and lower electrodes. The two workpieces are located on the upper and lower electrodes, respectively. During welding, the electrode material and the workpiece plate diffuse into each other, affecting the shape of the electrode surface and thus reducing the welding quality. The weld nugget diameter is the most critical indicator for measuring the welding effect. This invention uses a multi-dimensional sensing module to collect welding current and voltage between electrodes at high frequency, detecting the micro-vibration spectrum at the moment of electrode contact. An electrode life prediction model is designed, along with various adjustment strategies for welding current and pressurization time. The intelligent control module uses a PID controller to adjust the welding current and pressurization time based on the calculated or adjusted strategies, and automatically switches to a low-power state (standby power < 50W) during non-welding periods. Ineffective energy consumption is reduced through current waveform optimization. Finally, a servo pressure control system in the actuator provides precise and stable electrode pressure, and a high-frequency inverter provides the current required for welding, ensuring that the final weld nugget diameter meets the requirements and guaranteeing welding quality. Simultaneously, the temperature distribution in the welding area is monitored, and infrared thermal imaging is used to verify the uniformity of the temperature distribution, ensuring that the weld nugget diameter meets the temperature requirements after welding.

[0025] In a specific embodiment, the multidimensional sensing module is specifically used to collect welding current and voltage between electrodes through a resistance detection unit, detect the micro-vibration spectrum at the moment of electrode contact through a vibration sensor, and monitor the temperature distribution in the welding area through a temperature measurement module. Specifically, the resistance detection unit samples the welding current (±0.5% accuracy) and the voltage between the electrodes at high frequency; the temperature measurement module is a 64×64 pixel infrared thermal imaging array to monitor the temperature distribution in the welding area (resolution ±5℃); and the vibration sensor is integrated into the electrode clamping end to detect the micro-vibration spectrum (20-2000Hz) at the moment of electrode contact.

[0026] In a specific embodiment, the intelligent control module is specifically used for: The dynamic resistance and dynamic resistance attenuation rate are calculated based on the welding current and the voltage between the electrodes. The dynamic resistance is calculated based on the welding current and the voltage between the electrodes, as shown in formula (11). (11) in, The resistance is a function of time, and its value is the dynamic resistance. The voltage between the electrodes. For welding current, This represents the total time of the welding process. Let the dynamic resistance sequence during a single welding process be: The slope of its fitted curve is calculated as the dynamic resistance attenuation rate, as shown in formula (12). (12) in, The dynamic resistance attenuation rate, This refers to the number of dynamic resistors. For sampling points, The dynamic resistance of the sampling point; The vibration energy entropy and electrode surface oxide layer thickness are calculated based on the micro-vibration spectrum, including: The frequency band energy distribution of the vibration signal in the micro-vibration spectrum is obtained by wavelet packet decomposition, and the vibration energy entropy value is calculated as shown in formula (13). (13) in, The value is the vibration energy entropy. The number of decomposition layers, For the first Energy of each frequency band For the first The percentage of energy in each frequency band relative to the total signal energy. For the first Energy in each frequency band; A linear relationship between the dominant frequency offset of the vibration signal and the oxide layer thickness is constructed to obtain the oxide layer thickness on the electrode surface, as shown in formula (14). (14) in, The thickness of the oxide layer on the electrode surface. and These are the undetermined coefficients of the linear model. This represents the dominant frequency offset of the vibration signal. The current main frequency, This is the initial main frequency; Specifically, and The determination of the undetermined coefficients of a linear model through calibration experiments is a conventional technique for those skilled in the art and will not be described or elaborated upon in detail. An electrode monitoring prediction model is constructed based on the random forest algorithm to output the remaining electrode lifetime according to the cumulative number of welding operations, dynamic resistance attenuation rate, vibration energy entropy value, and electrode surface oxide layer thickness, including: The cumulative number of welding operations, dynamic resistance attenuation rate, vibration energy entropy value, and electrode surface oxide layer thickness are used as the training set. K sample subsets are randomly selected with replacement from the training set and input into the random forest model for training. The MSE criterion is used for splitting until the set maximum depth is reached and then splitting stops to obtain the electrode monitoring degree prediction model, as shown in formula (15). (15) in, For the remaining lifespan of the electrode, For the number of decision trees, , To calculate the cumulative number of welding operations, The dynamic resistance attenuation rate, The value is the vibration energy entropy. The thickness of the oxide layer on the electrode surface; In this scheme, K=100, that is, the number of decision trees is 100, and the maximum depth is 8. The specific training process and the use of the MSE criterion for splitting are conventional techniques. Random forest is a publicly available machine learning model, so the specific training process will not be described. A dynamic resistance-melt nucleus growth kinetic model is constructed to calculate the melt nucleus diameter based on dynamic resistance, including: Construct the integral area of ​​the dynamic resistance curve, as shown in formula (16). (16) in, This represents the integral area of ​​the dynamic resistance curve during the welding process. Since is a function of resistance over time, , The voltage between the electrodes. For welding current, This represents the total time of the welding process. A nonlinear model is constructed to represent the relationship between the melt core diameter and the integral area of ​​the dynamic resistance curve, as shown in formula (17). (17) in, The diameter of the molten core, , and The correlation coefficient of the workpiece material; Specifically, the correlation coefficient of the workpiece material is determined through multiple regression calibration. Welding is performed on a set of known welding materials to produce different integrated areas. Then, the weld nugget diameter is measured to find the optimal one. , and The specific methods and experiments are conventional technical means for those skilled in the art and will not be described in detail. Construct a dynamic resistance control strategy, which includes: The rate of change of resistance of dynamic resistance is calculated using the derivative method. ,like If the rate of increase exceeds the preset value, the welding current will be increased by w%; if If the average value is lower than the benchmark value within a certain period of time, welding is stopped and the electrodes are reground. Specifically, if caused by surface contamination If there is a sudden increase, i.e. the rate of increase exceeds the preset value, the current will be automatically increased by 5%. If caused by electrode wear If the average value is lower than the reference value within a certain period of time, the welding clamp will automatically trigger the electrode grinding program to complete the electrode grinding. Construct a dynamic compensation control strategy, including: Based on the combination of material type and plate thickness, the peak time deviation is preset. In this plan The range is ; Will The point is set as the peak point of the dynamic resistance, and the occurrence time of the peak point is recorded. ,calculate Peak time deviation from theoretical peak time ,like The welding current is adjusted using the current adjustment formula, and the pressure application time is adjusted using the pressure application time adjustment formula. For example, in normal welding of galvanized steel sheets, the theoretical peak time is 60ms. However, actual measurements may vary. The delay is 12ms. The current regulation formula is shown in (16). (16) in, For the adjusted welding current, To adjust the welding current before welding, This is the preset gain coefficient; The formula for adjusting the pressurization time is shown in formula (17). (17) in, The adjusted pressurization time. The pressurization time before adjustment. The gain coefficient for the preset pressurization time; An adaptive welding parameter optimizer is constructed based on the Q-learning algorithm, including: Step 1: Initialize the Q-table, define the state space, action space, and reward function. The state space includes the dynamic resistance average, plate thickness combination, and electrode remaining lifetime; the action space is for adjusting the welding current. and pressurization time The reward function is shown in formula (18). (18) in, The target melting core diameter; Step 2: Determine the current state Choose an action based on the ε-greedy strategy. That is, to select a set of adjusted welding current values ​​and pressure application time values; Step 3: Perform the action using the new parameters; measure the new average dynamic resistance after welding, and calculate the weld nugget diameter and the new reward function to determine the new state. ; Step 4: Update the Q-table using Step 2 until the Q-table converges, thus obtaining the trained Q-table, which is the adaptive welding parameter optimizer. The welding current and welding time are dynamically adjusted based on the dynamic resistance and weld nugget diameter using a dynamic resistance control strategy, a dynamic compensation control strategy, and an adaptive welding parameter optimizer. This includes: The initial welding current and pressurization time are obtained by an adaptive welding parameter optimizer based on the average dynamic resistance, actual weld nugget diameter, plate thickness combination, and remaining electrode life; the plate thickness is the thickness of each of the two (or more) metal plates to be welded. Welding is performed using the initial welding current and pressure application time. The resistance change rate of the dynamic resistance is detected during the welding process. If the rate of increase of the resistance change rate exceeds the preset value or the average value within a certain period of time is lower than the reference value, the initial welding current and pressure application time are adjusted using the dynamic resistance control strategy to obtain the welding current and pressure application time after secondary adjustment. Welding is performed using a secondary-adjusted welding current and pressurization time. The peak delay of the dynamic resistance is detected. If the peak delay is greater than the set time deviation, a dynamic compensation control strategy is adopted to readjust the secondary-adjusted welding current and pressurization time to obtain the final welding current and pressurization time.

[0027] The adjustment of welding current and pressure application time in this module is divided into three stages: Phase 1: Initial Parameter Setting (Before Welding Begins) The adaptive optimizer queries the "knowledge base" based on the current state (plate thickness, electrode life, etc.) to provide a theoretically optimal initial welding current and pressurization time. Phase Two: Based on the rate of change of resistance ( Rapid intervention occurs within the first few milliseconds to tens of milliseconds after the welding current is turned on. This stage corresponds to the process of electrode contact with the plate, the coating being crushed, and heat generation beginning. A PID controller would monitor this process at an extremely high frequency (e.g., every 0.1 ms). The system compares the current to a preset normal range, and this judgment and adjustment are performed on a microsecond or millisecond scale. For example, if surface contamination is detected causing a sudden increase in dR / dt, the system may increase the current by 5% within 1-2 ms. This adjustment occurs before the melt nucleus actually begins to grow. Phase 3: Compensation based on peak time delay. During this phase, the melt nucleus is forming and growing. The control system will calculate the resistance curve in real time and identify the peak point. Once the peak time point is identified (e.g., 72ms) and the delay amount (12ms) from the expected value (60ms) is calculated, the dynamic compensation strategy will take effect immediately. In a specific embodiment, the execution module is used to perform welding according to the welding current and welding time through a servo pressure control system and a high-frequency inverter power supply to achieve the target weld nugget diameter; Specifically, the servo pressure control system has a pressure regulation accuracy of ±10N (compared to ±50N for traditional pneumatic systems), the high-frequency inverter power supply has a response time of ≤1ms, and supports current waveform modulation.

[0028] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions 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. An intelligent welding system, characterized in that, include: Multi-dimensional sensing module, intelligent control module, and execution module; The multidimensional sensing module is used to collect welding current and voltage between electrodes through a resistance detection unit, and to detect the micro-vibration spectrum at the moment of electrode contact through a vibration sensor. The intelligent control module is used to calculate the dynamic resistance and dynamic resistance attenuation rate based on the welding current and the voltage between the electrodes, and to calculate the vibration energy entropy and the oxide layer thickness on the electrode surface based on the micro-vibration spectrum. It constructs an electrode monitoring prediction model based on the random forest algorithm to output the remaining electrode lifespan based on the cumulative welding count, dynamic resistance attenuation rate, vibration energy entropy, and oxide layer thickness. It also constructs a dynamic resistance-melt nugget growth dynamics model to calculate the melt nugget diameter based on the dynamic resistance, and builds dynamic resistance control strategies and dynamic compensation control strategies. An adaptive welding parameter simulation optimizer is constructed based on the Q-learning algorithm. Through the dynamic resistance control strategy, dynamic compensation control strategy, and adaptive welding parameter optimizer, the welding current and welding time are dynamically adjusted based on the dynamic resistance, melt nugget diameter, and remaining electrode lifespan. The execution module is used to perform welding according to the welding current and welding time through a servo pressure control system and a high-frequency inverter power supply to achieve the target weld nugget diameter.

2. The intelligent welding system according to claim 1, characterized in that, The dynamic resistance and dynamic resistance attenuation rate are calculated based on the welding current and the voltage between the electrodes, including: The dynamic resistance is calculated based on the welding current and the voltage between the electrodes, as shown in formula (1). （1） in, The resistance is a function of time, and its value is the dynamic resistance. The voltage between the electrodes. For welding current, Let be the total time of the welding process; let the dynamic resistance sequence during a single welding process be . The slope of its fitted curve is calculated as the dynamic resistance attenuation rate, as shown in formula (2). （3） in, The dynamic resistance attenuation rate, This refers to the number of dynamic resistors. For sampling points, This represents the dynamic resistance at the sampling point.

3. The intelligent welding system according to claim 2, characterized in that, The vibration energy entropy and electrode surface oxide layer thickness are calculated based on the micro-vibration spectrum, including: The frequency band energy distribution of the vibration signal in the micro-vibration spectrum is obtained by wavelet packet decomposition, and the vibration energy entropy value is calculated as shown in formula (3). （3） in, The value is the vibration energy entropy. The number of decomposition layers, For the first Energy of each frequency band For the first The percentage of energy in each frequency band relative to the total signal energy. For the first Energy in each frequency band; A linear relationship between the dominant frequency offset of the vibration signal and the oxide layer thickness is constructed to obtain the oxide layer thickness on the electrode surface, as shown in formula (4). （4） in, The thickness of the oxide layer on the electrode surface. and These are the undetermined coefficients of the linear model. This represents the dominant frequency offset of the vibration signal. The current main frequency, This is the initial main frequency.

4. The intelligent welding system according to claim 1, characterized in that, An electrode monitoring prediction model is constructed based on the random forest algorithm to output the remaining electrode lifetime according to the cumulative number of welding operations, dynamic resistance attenuation rate, vibration energy entropy value, and electrode surface oxide layer thickness, including: The cumulative number of welding operations, dynamic resistance attenuation rate, vibration energy entropy value, and electrode surface oxide layer thickness are used as the training set. K sample subsets are randomly selected with replacement from the training set and input into the random forest model for training. The MSE criterion is used for splitting until the set maximum depth is reached, and the electrode monitoring degree prediction model is obtained, as shown in formula (5). （5） in, For the remaining lifespan of the electrode, For the number of decision trees, , To calculate the cumulative number of welding operations, The dynamic resistance attenuation rate, The value is the vibration energy entropy. The thickness of the oxide layer on the electrode surface.

5. The intelligent welding system according to claim 2, characterized in that, A dynamic resistance-melt nucleus growth kinetic model is constructed to calculate the melt nucleus diameter based on dynamic resistance, including: Construct the integral area of ​​the dynamic resistance curve, as shown in formula (6). （6） in, This represents the integral area of ​​the dynamic resistance curve during the welding process. Since is a function of resistance over time, , The voltage between the electrodes. For welding current, This represents the total time of the welding process. A nonlinear model is constructed to represent the relationship between the melt core diameter and the integral area of ​​the dynamic resistance curve, as shown in equation (7). （7） in, The diameter of the molten core, , and This represents the correlation coefficient of the workpiece material.

6. The intelligent welding system according to claim 2, characterized in that, An adaptive welding parameter optimizer is constructed based on the Q-learning algorithm, including: Step 1: Initialize the Q-table, define the state space, action space, and reward function. The state space includes the dynamic resistance average, plate thickness combination, and electrode remaining lifetime; the action space is for adjusting the welding current. and pressurization time The reward function is shown in formula (8). （8） in, The target melting core diameter; Step 2: Determine the current state Choose an action based on the ε-greedy strategy. That is, to select a set of adjusted welding current values ​​and pressure application time values; Step 3: Perform the action using the new parameters; measure the new average dynamic resistance after welding, and calculate the weld nugget diameter and the new reward function to determine the new state. ; Step 4: Update the Q-table using Step 2 until the Q-table converges, resulting in the trained Q-table, which is the adaptive welding parameter optimizer.

7. The intelligent welding system according to claim 1, characterized in that, Dynamic resistance control strategies include: The rate of change of resistance of dynamic resistance is calculated using the derivative method. ,like If the rate of increase exceeds the preset value, the welding current will be increased by w%; if If the average value is lower than the benchmark value within a certain period of time, welding is stopped and the electrodes are ground.

8. The intelligent welding system according to claim 1, characterized in that, Dynamic compensation control strategies include: Based on the combination of material type and plate thickness, the peak time deviation is preset. ; Will The point is set as the peak point of the dynamic resistance, and the occurrence time of the peak point is recorded. ,calculate Peak time deviation from theoretical peak time ,like The welding current is adjusted using the current adjustment formula, and the pressure time is adjusted using the pressure time adjustment formula. The current regulation formula is shown in (9). （9） in, For the adjusted welding current, To adjust the welding current before welding, This is the preset gain coefficient; The formula for adjusting the pressurization time is shown in formula (10). （10） in, The adjusted pressurization time. The pressurization time before adjustment. The gain coefficient is the preset pressurization time.

9. The intelligent welding system according to claim 6, characterized in that, The welding current and welding time are dynamically adjusted based on the dynamic resistance and weld nugget diameter using a dynamic resistance control strategy, a dynamic compensation control strategy, and an adaptive welding parameter optimizer. This includes: The initial welding current and pressurization time are obtained by an adaptive welding parameter optimizer based on the dynamic resistance average, actual weld nugget diameter, plate thickness combination, and remaining electrode life. Welding is performed using the initial welding current and pressure application time. The resistance change rate of the dynamic resistance is detected during the welding process. If the rate of increase of the resistance change rate exceeds the preset value or the average value within a certain period of time is lower than the reference value, the initial welding current and pressure application time are adjusted using the dynamic resistance control strategy to obtain the welding current and pressure application time after secondary adjustment. Welding is performed using a secondary-adjusted welding current and pressurization time. The peak delay of the dynamic resistance is detected. If the peak delay is greater than the set time deviation, a dynamic compensation control strategy is adopted to readjust the secondary-adjusted welding current and pressurization time to obtain the final welding current and pressurization time.

Images