Parameter self-adaption method and system for laser cutting

Abstract

The invention relates to the technical field of laser cutting, in particular to a parameter self-adaption method and system for laser cutting, and the method comprises the steps: constructing a solution space containing laser power and cutting speed, and initializing a plurality of nectar sources of an artificial bee colony algorithm; the follow-up error in the laser cutting process is collected in real time, the risk index of the current working condition is calculated according to the Rayleigh length, the size of the follow-up error and the change rate, and then the search step length and the fitness weight are determined; generating a new solution in the neighborhood of the nectar source by using the search step size, and calculating a fitness function value of the new solution according to the fitness weight; and the nectar sources are updated according to a greedy selection strategy, the nectar source with the highest fitness function value is selected as the optimal control parameter after iteration is completed, and the laser power and the cutting speed are adjusted according to the optimal control parameter. According to the technical scheme, the stability of a cutting energy field can be maintained, and the quality defects of slag adhering, incomplete cutting and the like are effectively overcome.

Description

Technical Field

[0001] This application relates to the field of laser cutting technology, and in particular to a parameter adaptive method and system for laser cutting. Background Technology

[0002] With the rapid development of intelligent manufacturing technology, fiber laser cutting has been widely used in metal material processing, automobile manufacturing, and aerospace. As a key link in ensuring the quality of workpiece forming, how to ensure the stability of laser energy's effect on the material under complex working conditions such as uneven surface flatness and continuously increasing cutting speed, thereby avoiding processing defects such as slag buildup and incomplete cutting, has become a technical problem that the industry urgently needs to solve.

[0003] Currently, patent application CN118657171A discloses a laser cutting parameter adaptive generation system based on geometric features and a process library. It relies on a general process database pre-set by the equipment manufacturer. Before processing, it calls a set of fixed reference parameters from the database according to the material type and thickness of the sheet to be processed, or applies uniform process parameters to the entire cutting path through CAM software, thereby realizing the control of the laser cutting equipment.

[0004] However, during high-speed cutting, the motor system that controls the height will have a delay in action due to the inertia of the mechanical structure itself. This makes it impossible for the height of the laser nozzle to be precisely locked at the set value at all times. Due to the untimely follow-up, a follow-up error will occur. This slight height deviation will cause the laser focus to deviate from the surface of the board, resulting in an unexpected change in the area of ​​the laser spot acting on the board, that is, defocusing. This leads to an unstable cutting energy field and makes it impossible to guarantee the quality of laser cutting. Summary of the Invention

[0005] To address the technical problem of unstable cutting energy fields that prevent the quality of laser cutting, this application provides a parameter adaptive method and system for laser cutting. This method and system can maintain the stability of the cutting energy field, effectively solve quality defects such as slag buildup and incomplete cutting, and achieve high-precision adaptive control of the laser cutting process.

[0006] In a first aspect, this application provides a parameter adaptive method for laser cutting. The adaptive method includes: constructing a solution space containing laser power and cutting speed, and initializing multiple honey sources using an artificial bee colony algorithm within the solution space; real-time acquisition of the follow-up error during the laser cutting process, and calculating a risk index for the current operating condition based on the Rayleigh length and the magnitude and rate of change of the follow-up error; determining a search step size and fitness weight, wherein the search step size is negatively correlated with the risk index, and the fitness weight is positively correlated with the risk index; generating a new solution in the neighborhood of the honey source using the search step size, calculating the fitness function value of the new solution based on the fitness weight, wherein the fitness function value characterizes the ability of the new solution to maintain energy density under the current operating condition; updating the honey source according to a greedy selection strategy, and selecting the honey source with the highest fitness function value as the optimal control parameter after iteration, and adjusting the laser power and cutting speed according to the optimal control parameter.

[0007] The search step size and fitness weight are dynamically adjusted based on the risk index. When the risk is high, the step size is reduced to avoid blind and large adjustments that may cause oscillations. At the same time, the weight is increased to strengthen the penalty for energy density deviation. The ability to maintain constant energy density is used as the fitness function to iteratively update each honey source, ensuring that the final selected control parameters can effectively compensate for follow-up errors and maintain the stability of cutting quality.

[0008] Preferably, before calculating the risk index of the current operating condition, the adaptive method further includes: taking the ratio of the absolute value of the real-time follow-up error to the maximum error threshold as the normalized follow-up error; dividing the absolute value of the difference between the follow-up errors at two adjacent sampling times by the product of the sampling period and the maximum speed fluctuation threshold to obtain the normalized error change rate; calculating the square of the ratio of the real-time follow-up error to the Rayleigh length of the laser beam, and taking the minimum value between the square and 1 as the normalized optical sensitivity.

[0009] Preferably, calculating the risk index of the current operating condition includes: calculating a weighted sum of normalized error, normalized jitter value and normalized optical sensitivity based on preset coefficients; the risk index is positively correlated with the weighted sum.

[0010] The risk index is derived by combining the combined error, jitter value, and optical sensitivity. This risk index can sensitively reflect the degree of deterioration of the current processing status, providing a precise quantitative basis for the dynamic adjustment of the subsequent search step size and fitness weight.

[0011] Preferably, the method for determining the search step size includes: calculating a damping factor that is negatively correlated with the risk index; and using the product of the base step size and the damping factor as the search step size, wherein the base step size includes the base step size of the laser power and the cutting speed, and the base step size of the laser power is the product of the range of laser power values ​​and a random number within a preset interval.

[0012] When the risk index is high, the search step size is forcibly compressed, forcing the artificial bee colony algorithm to make fine adjustments or maintain the status quo near the current solution. This effectively prevents control oscillations caused by excessive parameter adjustments under harsh conditions and ensures the stability of the control process.

[0013] Preferably, the method for determining fitness weights includes: calculating the power function value of the risk index, wherein the exponent of the power function is greater than or equal to 1; using the product of the power function value and the weight amplification gain as the weight adjustment amount, and the sum of the weight adjustment amount and the base weight as the fitness weight.

[0014] When the risk is low, the fitness weight remains constant, while when the risk increases, the fitness weight is significantly increased, thereby amplifying the proportion of energy density deviation in fitness evaluation. This makes the artificial bee colony algorithm pay more attention to the magnitude of error under dangerous conditions and prioritize the selection of parameter combinations that can accurately compensate for the error.

[0015] Preferably, calculating the fitness function value of the new solution based on the fitness weight includes: calculating the actual energy density based on the laser power and cutting speed in the new solution; calculating the absolute value of the deviation between the actual energy density and the target energy density; and using the product of the fitness weight and the absolute value of the deviation as the dynamic deviation, wherein the fitness function value of the new solution is negatively correlated with the dynamic deviation.

[0016] Preferably, the calculation of the actual energy density includes: estimating the actual spot area based on the current servo error; and in the new solution, dividing the laser power by the product of the cutting speed and the actual spot area to obtain the actual energy density.

[0017] Preferably, updating the nectar source according to the greedy selection strategy includes: in the follower bee search phase of the artificial bee colony algorithm, calculating the selection probability based on the fitness function value of each nectar source, and selecting the target nectar source using a roulette wheel strategy; constructing a guiding vector, wherein the guiding vector is the difference between the position of the globally optimal nectar source and the position of the target nectar source; and in the neighborhood of the target nectar source, superimposing the search step size and the guiding vector to generate a new solution, and performing greedy selection on the new solution.

[0018] By introducing a guiding vector consisting of the difference between the global optimal nectar source location and the target nectar source location during the follower bee search phase, the optimal information in the population is used to guide the search direction. This prevents the follower bees from blindly searching randomly, but instead directs them toward the currently known optimal solution region, thereby significantly accelerating the convergence speed of the algorithm and improving the response efficiency of real-time control.

[0019] Preferably, updating nectar sources based on a greedy selection strategy further includes: during the reconnaissance bee reset phase of the artificial bee colony algorithm, counting the number of iterations in which each nectar source has not been updated consecutively; in response to any nectar source having more than a preset threshold number of iterations, determining that the nectar source is trapped in a local optimum; calculating the local gradients of the nectar source trapped in a local optimum in the dimensions of laser power and cutting speed; retaining the parameter values ​​corresponding to dimensions where the local gradient is less than the gradient threshold; randomly resetting the parameter values ​​corresponding to dimensions where the local gradient is not less than the gradient threshold in the solution space; and combining the retained parameter values ​​with the reset parameter values ​​to form an alternative nectar source.

[0020] In the scout bee phase, a gradient-based selective reset mechanism is introduced to calculate the local gradient of the honey source trapped in a local optimum in each dimension. The parameter values ​​of converged dimensions with smaller gradients are retained, while the parameter values ​​of non-converged dimensions with larger gradients are randomly reset. This not only preserves the effective information that has been searched and avoids algorithm degradation, but also provides the necessary perturbation to escape local traps, effectively improving the ability to escape local optima.

[0021] In a second aspect, this application also provides a parameter adaptive system for laser cutting, including a processor and a memory, wherein the memory stores computer program instructions that, when executed by the processor, implement a parameter adaptive method for laser cutting according to the first aspect of this application.

[0022] The technical solution of this application has the following beneficial technical effects: By initializing the artificial bee colony algorithm within the solution space of laser power and cutting speed, and collecting the follower error in real time to calculate the risk index reflecting the degree of working condition deterioration, the search step size and fitness weight of the artificial bee colony algorithm are dynamically adjusted based on the risk index. New solutions are generated in the neighborhood of the nectar source using the adjusted parameters. Finally, through a series of improvement measures such as greedy selection, improved follower bee-guided search, and scout bee parameter reset, iterative optimization is carried out to select the parameter combination with the highest fitness for closed-loop control of laser power and cutting speed. When the working condition deteriorates, the algorithm can prevent control oscillation by compressing the search step size and strengthen error constraints by increasing the fitness weight. At the same time, it can accelerate convergence by using the guiding vector and gradient reset to escape local optima. This enables rapid and accurate compensation for the change in spot area caused by Z-axis follower lag, and maintains the stability of the cutting energy field. This effectively solves quality defects such as slag buildup and incomplete cutting, and realizes high-precision adaptive control of the laser cutting process. Attached Figure Description

[0023] Figure 1 This is a flowchart of a parameter adaptation method for laser cutting according to an embodiment of this application.

[0024] Figure 2This is a comparison diagram of the convergence process of the improved artificial bee colony algorithm and the traditional artificial bee colony algorithm according to the embodiments of this application.

[0025] Figure 3 This is a comparison diagram of the effect of cutting energy field stability according to the embodiments of this application.

[0026] Figure 4 This is a structural block diagram of a parameter adaptive system for laser cutting according to an embodiment of this application. Detailed Implementation

[0027] According to a first aspect of this application, this application provides a parameter adaptive method for laser cutting. Figure 1 This is a flowchart of a parameter adaptation method for laser cutting according to an embodiment of this application. To facilitate understanding of the technical solution of this application, the Artificial Bee Colony (ABC) algorithm involved in this application is first briefly described. The ABC algorithm is a swarm intelligence optimization algorithm that simulates the nectar-collecting behavior of a bee colony. In this algorithm, the location of the nectar source represents a potential solution to the optimization problem, and the amount of nectar in the nectar source corresponds to the fitness of the potential solution, i.e., the quality of the potential solution. The bee colony is divided into three roles: hired bees, follower bees, and scout bees. Hired bees are responsible for searching within the neighborhood of known nectar sources and calculating the fitness of new solutions. Follower bees, based on information shared by hired bees, select nectar sources with a certain probability for further neighborhood searches. Scout bees are responsible for randomly searching for new nectar sources to replace abandoned nectar sources when a nectar source remains unimproved after multiple searches, i.e., it is trapped in a local optimum. This application uses the process parameters involved in laser cutting as nectar sources in the ABC algorithm, and achieves adaptive optimization of the optimal cutting parameters through the cooperative search of bees in each role. Figure 1 As shown, the parameter adaptation method for laser cutting includes steps S101 to S106, which are described in detail below.

[0028] S101, construct a solution space containing laser power and cutting speed, and initialize multiple honey sources for the artificial bee colony algorithm within the solution space.

[0029] In one embodiment, the solution space refers to the set of values ​​for each process parameter in the laser cutting process. In this application, the process parameters include laser power and cutting speed. A "honey source" refers to a combination of parameters representing a potential solution in the artificial bee colony algorithm; one honey source corresponds to one laser cutting control scheme. Based on the material, thickness, and machine tool performance of the sheet metal to be processed, the feasible domains of laser power and cutting speed are determined, thereby defining the solution space, which is a two-dimensional space composed of laser power and cutting speed.

[0030] For example, for a 10mm thick carbon steel plate, the laser power is set to a range of 2000 watts to 3000 watts, and the cutting speed is set to a range of 1.2 meters per minute to 1.8 meters per minute. Within this solution space, N sets of process parameter combinations are generated using a uniformly random distribution, and each set of parameter combinations is an initial honey source.

[0031] In this way, the solution space was determined and multiple honey sources were initialized within the solution space, providing a data foundation for the adaptive determination of subsequent parameters.

[0032] S102, collects the follow-up error during the laser cutting process in real time, and calculates the risk index of the current working condition based on the Rayleigh length and the magnitude and rate of change of the follow-up error.

[0033] In one embodiment, the follow-up error refers to the deviation between the actual distance between the laser cutting head nozzle and the surface of the material and the set ideal distance. Rayleigh length refers to the axial distance along the propagation direction when the spot radius increases to the square root of 2 times the waist spot radius; it is a key optical parameter for measuring the focusing performance of a laser beam. The risk index is used to characterize the degree of threat posed by the current processing state to the cutting quality; the higher the risk index, the more likely the current processing state is to lead to substandard cutting quality.

[0034] Before calculating the risk index of the current operating condition, the collected data needs to be normalized to eliminate the influence of different physical dimensions. Specifically, before calculating the risk index of the current operating condition, the adaptive method further includes: taking the ratio of the absolute value of the real-time follow-up error to the maximum error threshold as the normalized follow-up error; dividing the absolute value of the difference between the follow-up errors at two adjacent sampling times by the product of the sampling period and the maximum speed fluctuation threshold to obtain the normalized error change rate; calculating the square of the ratio of the real-time follow-up error to the Rayleigh length of the laser beam, and taking the minimum value between the square and 1 as the normalized optical sensitivity.

[0035] Among them, the normalized follower error Satisfying the relation: ;in, This represents the real-time tracking error at the current sampling moment. The maximum error threshold is defined as ; the normalized follow-up error reflects the deviation between the height of the laser cutting head nozzle and the ideal height at the current sampling moment.

[0036] Normalized error rate of change Satisfying the relation: ;in, This represents the tracking error from the previous sampling time. The sampling period is The maximum speed fluctuation threshold of the Z-axis servo system is defined as ; the normalized error rate of change characterizes the severity of Z-axis mechanical vibration.

[0037] Normalized optical sensitivity Satisfying the relation: ;in, This represents the real-time tracking error at the current sampling moment. The Rayleigh length is used; normalized optical sensitivity characterizes the degree to which the current defocusing amount affects the change in spot area.

[0038] After normalization is completed, the risk index for the current operating condition is calculated by: calculating a weighted sum of normalization error, normalized jitter value and normalized optical sensitivity based on preset coefficients; the risk index is positively correlated with the weighted sum.

[0039] Specifically, risk index Satisfying the relation: ; in, , , These are the weighting coefficients for normalized error, normalized error rate of change, and normalized optical sensitivity, respectively, and they satisfy the following conditions: This ensures that the risk index ranges from 0 to 1. As each normalized indicator increases, the weighted sum increases, leading to an increase in the risk index, thus enabling the risk index to accurately reflect the degree of deterioration in the working conditions.

[0040] In this way, the risk index can adaptively identify the safe zone and the dangerous zone in the processing, providing a quantitative basis for subsequent algorithm strategy adjustments.

[0041] S103, determine the search step size and fitness weight, wherein the search step size is negatively correlated with the risk index and the fitness weight is positively correlated with the risk index.

[0042] In one embodiment, the search step size determines the range of movement when exploring new solutions in the solution space. Fitness weights are used to adjust the penalty for the error term in the fitness function value; a larger penalty indicates that the artificial bee colony algorithm is more inclined to find the solution with the smallest error term. To prevent the algorithm from blindly and drastically adjusting parameters under unstable operating conditions, leading to fluctuations in cutting quality, and to strengthen the error constraint under dangerous operating conditions, a risk index is used to dynamically adjust the search step size and fitness weights.

[0043] The method for determining the search step size includes: calculating a damping factor that is negatively correlated with the risk index; and using the product of the base step size and the damping factor as the search step size, wherein the base step size includes the base step size of the laser power and the cutting speed, and the base step size of the laser power is the product of the range of laser power values ​​and a random number within a preset interval.

[0044] Specifically, search step size Satisfying the relation: ; in, The base step size is related to the range of laser power or cutting speed, and can adapt to the differences in magnitude of various parameters. The preset damping coefficient, This is the risk index. When the risk index is high, the damping factor... The rapid decay to near 0 reduces the search step size, allowing the artificial bee colony algorithm to be fine-tuned or maintain the status quo, avoiding significant parameter adjustments when the risk index is high.

[0045] For example, the base step size is a two-dimensional vector, including the base step size of the laser power and the cutting speed; similarly, the search step size... This also includes the search step size for laser power and cutting speed; the preset range is [-0.5, +0.5], and the random number within the preset range is 0.1; for laser power, the laser power value range is set to 2000 watts to 3000 watts, then the basic step size for laser power is... For the cutting speed, the range is set from 1.2 meters per minute to 1.8 meters per minute, then the basic step size of the cutting speed is... meters per minute.

[0046] The method for determining fitness weights includes: calculating the power function value of the risk index, wherein the exponent of the power function is greater than or equal to 1; using the product of the power function value and the weight amplification gain as the weight adjustment amount; and using the sum of the weight adjustment amount and the base weight as the fitness weight.

[0047] Specifically, if the base weight is set to 1, then the fitness weight... Satisfying the relation: .in, To amplify the gain by weighting, The exponent of the power function is and satisfies Understandably, when the risk index is low, the fitness weight is close to the basic weight of 1, and the artificial bee colony algorithm remains normal; when the risk index increases, the fitness weight grows exponentially, forcing the artificial bee colony algorithm to reduce its tolerance for energy density deviations at dangerous moments and prioritize retaining those solutions that can accurately compensate for errors.

[0048] In this way, by adaptively adjusting the search step size and fitness weights through the risk index, the artificial bee colony algorithm can improve its resistance to oscillations and its sensitivity to errors under harsh conditions with a high risk index.

[0049] S104: Generate a new solution in the neighborhood of the honey source using the search step size, and calculate the fitness function value of the new solution based on the fitness weight. The fitness function value characterizes the ability of the new solution to maintain energy density under the current operating conditions.

[0050] In one embodiment, this step corresponds to the hired bee stage in the artificial bee colony algorithm. After determining the search step size, the hired bees use the search step size to perform a local search within the neighborhood of the current nectar source, generating a new solution that includes the new laser power and the new cutting speed. Subsequently, the fitness function value of the new solution is calculated to evaluate the quality of the new solution.

[0051] The calculation of the fitness function value of the new solution based on the fitness weight includes: calculating the actual energy density based on the laser power and cutting speed in the new solution; calculating the absolute value of the deviation between the actual energy density and the target energy density; and taking the product of the fitness weight and the absolute value of the deviation as the dynamic deviation. The fitness function value of the new solution is negatively correlated with the dynamic deviation.

[0052] Furthermore, the calculation of the actual energy density includes: estimating the actual spot area based on the current servo error; and in the new solution, dividing the laser power by the product of the cutting speed and the actual spot area to obtain the actual energy density.

[0053] As a specific implementation method, the laser beam can be regarded as a fundamental Gaussian beam, and the follower error can be regarded as the defocusing amount of the laser focus relative to the surface of the plate; based on the Gaussian beam propagation theory, the actual spot area Satisfying the relation: ; in, The minimum spot area at the ideal focal plane is determined by the inherent parameters of the laser and the focal length of the focusing lens, and is a preset constant in this embodiment; Let be the Rayleigh length of the laser beam; This represents the real-time tracking error at the current sampling moment. When the tracking error... When the value is 0, the actual spot area equals the ideal spot area, and the energy density is the highest; when the servo error is 0, the actual spot area is equal to the ideal spot area. When the absolute value of increases, the actual spot area increases non-linearly in a quadratic manner, resulting in a decrease in the energy density acting on the surface of the plate.

[0054] Specifically, actual energy density Satisfying the relation: ;in, The laser power in the new solution, The cutting speed in the new solution. The actual spot area is estimated based on the current servo error. It should be noted that the actual energy density referred to in this application refers to the laser energy absorbed per unit volume of material, in order to comprehensively characterize the coupling effect of power, velocity, and spot area.

[0055] fitness function value Satisfying the relation: ; in, For actual energy density, For the target energy density, For fitness weights, These are preset coefficients used to prevent the denominator from being zero. The dimensions of the preset coefficients are the same as... The dimensions should remain consistent. In high-risk operating conditions with a high risk index, Increasing the energy density deviation amplifies its proportion in the denominator of the fitness function, causing even small energy deviations to cause the fitness function value to drop rapidly, thus selecting the parameter combination that best maintains constant energy.

[0056] The target energy density is related to the material and thickness of the plate to be processed. The target energy density is the ideal energy density for cutting the plate to be processed corresponding to the material and thickness, and can be preset by those skilled in the art.

[0057] In this way, the changes in light spot caused by the motion error are incorporated into the calculation of the actual energy density. Through a dynamically weighted fitness function, the fitness function value is ensured to match the risk index at the current sampling time, guiding the search direction toward convergence towards the target of constant energy density.

[0058] S105, update the honey source according to the greedy selection strategy, and select the honey source with the highest fitness function value as the optimal control parameter after the iteration is completed.

[0059] In one embodiment, updating the nectar source according to the greedy selection strategy includes: in the follower bee search phase of the artificial bee colony algorithm, calculating the selection probability based on the fitness function value of each nectar source, and selecting the target nectar source using a roulette wheel strategy; constructing a guiding vector, wherein the guiding vector is the difference between the position of the globally optimal nectar source and the position of the target nectar source; and in the neighborhood of the target nectar source, superimposing the search step size and the guiding vector to generate a new solution, and performing greedy selection on the new solution.

[0060] Specifically, generate new solutions The process satisfies the following relation: ; in, The selected target nectar source location, This represents the globally optimal nectar source location within the current population. For the search step size, This is the guiding coefficient.

[0061] Understandably, Using the guiding vector, the optimal nectar source location in the population is used to guide the search direction, so that the following bees no longer blindly search randomly, but move towards the currently known optimal solution, thus significantly accelerating the convergence speed.

[0062] In this embodiment, updating the nectar source according to the greedy selection strategy further includes: during the reconnaissance bee reset phase of the artificial bee colony algorithm, counting the number of iterations in which each nectar source has not been updated consecutively; in response to any nectar source having more than a preset threshold number of iterations, determining that the nectar source is trapped in a local optimum; calculating the local gradient of the nectar source trapped in a local optimum in the dimensions of laser power and cutting speed; retaining the parameter values ​​corresponding to the dimensions where the local gradient is less than the gradient threshold; randomly resetting the parameter values ​​corresponding to the dimensions where the local gradient is not less than the gradient threshold in the solution space; and combining the retained parameter values ​​with the reset parameter values ​​to form an alternative nectar source.

[0063] Specifically, for honey sources that are determined to be trapped in a local optimum... Calculate the fitness function of each laser power. and cutting speed The absolute value of the partial derivative is used as the local gradient. and The pre-set gradient threshold is denoted as... ,like The laser power in the original honey source will be retained. Otherwise, a new laser power is randomly generated in the power solution space; similarly, if Then the cutting speed of the original honey source is preserved. Otherwise, a new cutting velocity is randomly generated in the velocity solution space. The gradient threshold is set to 0.5.

[0064] Understandably, a small gradient indicates that the dimension is changing gradually within the current region and is close to its extreme point. Retaining this parameter helps preserve the converged and valid information. Conversely, a large gradient and an algorithm stagnation indicate that the dimension is in a region of drastic change or oscillation. Randomly resetting this dimension can help the algorithm escape the current complex terrain. The selective resetting strategy based on gradient thresholds can effectively utilize historical search experience while maintaining population diversity, thus improving the algorithm's optimization efficiency.

[0065] After each complete bee colony search cycle, the algorithm determines whether it meets the iteration completion conditions. These conditions include: the current iteration count reaching a preset maximum iteration count, or the change in the fitness function value of the global optimal solution being less than a preset change value for multiple consecutive iterations. The maximum iteration count is 50, and the preset change value is 0.001.

[0066] Please see Figure 2 The figure shows a comparison of the convergence process of the improved artificial bee colony algorithm and the traditional artificial bee colony algorithm according to the embodiments of this application. As can be seen from the figure, the improved artificial bee colony algorithm of the embodiments of this application has a faster growth rate of fitness function value in the early stage of iteration due to the introduction of the guiding vector, and the fitness function value has less oscillation in the later stage due to the effect of the adaptive search step size, resulting in a faster convergence speed.

[0067] Thus, through the improved follower bee guidance mechanism and the scout bee gradient selective reset mechanism, not only is the guidance of the global optimal solution utilized to accelerate convergence, but also the gradient information is combined to achieve population regeneration, ensuring that the algorithm can find the optimal solution quickly and accurately.

[0068] S106 adjusts the laser power and cutting speed according to the optimal control parameters.

[0069] In one embodiment, after completing the iterative calculation for one control cycle, the honey source with the highest fitness function value is selected from the final population as the optimal control parameter. The laser power in the optimal control parameter is converted into an analog voltage signal and sent to the laser generator, and the cutting speed is converted into a magnification command and sent to the motion controller. The control cycle is 10 milliseconds.

[0070] Please see Figure 3The figure shows a comparison of the effects of cutting energy field stability according to the embodiments of this application; the horizontal axis represents the control time, and the vertical axis represents the cutting energy field stability, that is, the ratio of the actual energy density to the target energy density. When the ratio is 1, it indicates that the energy field is in an ideal constant energy state, that is, the ideal value of the vertical axis is 1. As can be seen from the figure, compared with the traditional artificial bee colony algorithm, the improved artificial bee colony algorithm can more effectively maintain the stability of the cutting energy field during laser cutting.

[0071] In this way, the actuator adjusts the laser output energy and machine tool movement speed in real time according to the instructions, completes the compensation for the current follow-up error, and always maintains the stability of the cutting energy field, effectively solving the quality defects such as slag and incomplete cutting caused by follow-up lag.

[0072] According to a second aspect of this application, this application also provides a parameter adaptive system for laser cutting. Figure 4 This is a structural block diagram of a parameter adaptive system for laser cutting according to an embodiment of this application. Figure 4 As shown, the system 50 includes a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement a parameter adaptation method for laser cutting according to the first aspect of this application. The system also includes other components well-known to those skilled in the art, such as a communication bus and a communication interface. Their configuration and functions are known in the art and will not be described further here.

[0073] It should be noted that, for those skilled in the art, several modifications and improvements can be made without departing from the concept of this application, and these all fall within the scope of protection of this application.

Claims

1. A method for parameter adaptation for laser cutting, characterized in that, The adaptive method includes: constructing a solution space containing laser power and cutting speed, and initializing multiple nectar sources for the artificial bee colony algorithm within the solution space; The tracking error during the laser cutting process is collected in real time, and the risk index of the current working condition is calculated based on the Rayleigh length and the magnitude and rate of change of the tracking error. The search step size and fitness weight are determined, wherein the search step size is negatively correlated with the risk index, and the fitness weight is positively correlated with the risk index; A new solution is generated in the neighborhood of the nectar source using the search step size. The fitness function value of the new solution is calculated based on the fitness weight. The fitness function value characterizes the ability of the new solution to maintain energy density under the current operating conditions. The honey source is updated according to the greedy selection strategy, and the honey source with the highest fitness function value is selected as the optimal control parameter after iteration; the laser power and cutting speed are adjusted according to the optimal control parameter.

2. A method for adaptive parameter selection for laser cutting as claimed in claim 1, wherein, Before calculating the risk index of the current operating condition, the adaptive method further includes: The ratio of the absolute value of the real-time follow-up error to the maximum error threshold is used as the normalized follow-up error. The normalized error rate of change is obtained by dividing the absolute value of the difference between the motion error at two adjacent sampling times by the product of the sampling period and the maximum speed fluctuation threshold. Calculate the square of the ratio of the real-time servo error to the Rayleigh length of the laser beam, and take the minimum of the square and 1 as the normalized optical sensitivity.

3. A method for adaptive parameter selection for laser cutting as claimed in claim 2, wherein, The risk index for the current operating condition is calculated by: calculating a weighted sum of normalized error, normalized jitter value and normalized optical sensitivity based on preset coefficients; the risk index is positively correlated with the weighted sum.

4. A method for adaptive parameter selection for laser cutting as claimed in claim 1 wherein, Methods for determining the search step size include: calculating the damping factor that is negatively correlated with the risk index; The product of the base step size and the damping factor is used as the search step size. The base step size includes the base step size of the laser power and the cutting speed. The base step size of the laser power is the product of the range of laser power values ​​and a random number within a preset interval.

5. A method for adaptive parameter selection for laser cutting as claimed in claim 1 wherein, The method for determining fitness weights includes: calculating the power function value of the risk index, wherein the exponent of the power function is greater than or equal to 1; using the product of the power function value and the weight amplification gain as the weight adjustment amount; and using the sum of the weight adjustment amount and the base weight as the fitness weight.

6. A method for adaptive parameter selection for laser cutting as claimed in claim 1 wherein, The fitness function value of the new solution is calculated based on the fitness weights, including: The actual energy density is calculated based on the laser power and cutting speed in the new solution; the absolute value of the deviation between the actual energy density and the target energy density is calculated; the product of the fitness weight and the absolute value of the deviation is taken as the dynamic deviation, and the fitness function value of the new solution is negatively correlated with the dynamic deviation.

7. A method for parameter adaptation for laser cutting according to claim 6, characterized in that, The calculation of the actual energy density includes: estimating the actual spot area based on the current servo error; in the new solution, dividing the laser power by the product of the cutting speed and the actual spot area to obtain the actual energy density.

8. The method of claim 1, wherein, The method of updating honey sources based on the greedy selection strategy includes: In the follower bee search phase of the artificial bee colony algorithm, the selection probability is calculated based on the fitness function value of each nectar source, and the target nectar source is selected using the roulette wheel strategy. Construct a guiding vector, which is the difference between the location of the globally optimal nectar source and the location of the target nectar source; within the neighborhood of the target nectar source, superimpose the search step size and the guiding vector to generate a new solution, and perform a greedy selection on the new solution.

9. A method for parameter adaptation for laser cutting according to claim 8, characterized in that, Updating honey sources based on the greedy selection strategy also includes: In the reconnaissance bee reset phase of the artificial bee colony algorithm, the number of iterations in which each nectar source has not been updated is counted. If the number of iterations of any nectar source exceeds a preset threshold, the nectar source is determined to be trapped in a local optimum. Calculate the local gradients of the honey source trapped in local optima in the laser power and cutting speed dimensions; retain the parameter values ​​corresponding to the dimensions where the local gradient is less than the gradient threshold, and randomly reset the parameter values ​​corresponding to the dimensions where the local gradient is not less than the gradient threshold in the solution space. Combine the retained parameter values ​​with the reset parameter values ​​to form an alternative honey source.

10. A parameter adaptive system for laser cutting, characterized by, It includes a processor and a memory, the memory storing computer program instructions that, when executed by the processor, implement a parameter adaptation method for laser cutting according to any one of claims 1 to 9.

Images