Laser cutting control method and system for battery casing

Abstract

This invention relates to the field of cutting control technology, specifically to a laser cutting control method and system for battery casings. The method includes: acquiring the operating data of the equipment during the cutting process of the battery casing within a set time period; based on the operating data within the set time period, using an improved particle swarm optimization algorithm for iterative optimization to obtain the command speed and command power for the next moment, thereby controlling the laser cutting of the battery casing; that is, the solution of this invention can adjust servo parameters in real time according to the dynamic changes of the cutting trajectory, effectively eliminating slag and ripple phenomena at sharp corners, and achieving a dual improvement in processing efficiency and cutting accuracy.

Description

Technical Field

[0001] This invention relates to the field of cutting control technology. More specifically, this invention relates to a laser cutting control method and system for battery casings. Background Technology

[0002] Power battery casings (such as square aluminum casings) are typically made of aluminum alloy, which has an extremely high thermal conductivity. During the cutting process, in order to ensure the sealing of subsequent welding and prevent internal short circuits, there are extremely stringent "zero-defect" requirements for burrs and slag on the cut surfaces.

[0003] Laser cutting is a process that uses a high-power-density laser beam to irradiate the material to be cut, rapidly heating the material to its vaporization temperature. As the beam moves relative to the material, a kerf is formed, thus completing the process.

[0004] Laser cutting plays an irreplaceable role in the field of power batteries due to its advantages such as high cutting precision, small heat-affected zone, and good processing flexibility. However, the laser cutting process is a nonlinear device involving strong coupling of multiple physical fields such as optics, mechanics, electronics, and heat. The cutting quality is affected by a variety of process parameters, including laser power, cutting speed, focal point position, and auxiliary gas pressure. Traditional open-loop control or empirical trial-and-error methods are difficult to adapt to the dynamic processing requirements of different materials and thicknesses, which can easily lead to rough cuts, slag buildup, or ablation.

[0005] Therefore, controlling the laser cutting process of the battery casing to maintain a high-precision dynamic response under high-speed processing is key to improving the performance of high-end manufacturing equipment.

[0006] In existing technologies, Particle Swarm Optimization (PSO) has been widely adopted in laser cutting control equipment. This algorithm simulates the collective intelligent behavior of bird flocks hunting, using individual and global extrema to guide particles in the solution space. It has advantages such as simple principle, ease of implementation, and fewer parameters requiring adjustment. However, in the laser cutting of battery casings, especially when cutting complex trajectories such as sharp corners and arcs, the equipment has extremely high requirements for real-time servo response and energy tracking. Furthermore, the objective function exhibits complex characteristics with multiple peaks. If PSO is directly used to optimize the control strategy, the particle population diversity is lost too quickly in the later stages of iteration, easily leading the algorithm into local optima. Consequently, the optimized control parameters cannot guarantee the contour tracking accuracy of the battery casing during the variable trajectory processing. Summary of the Invention

[0007] The purpose of this invention is to propose a laser cutting control method and system for battery casings, in order to solve the problem that when the particle swarm optimization algorithm is directly used to optimize the control strategy in the prior art, it is easy for the algorithm to get stuck in a local optimum, and the contour tracking accuracy of the battery casing during the variable trajectory processing is not guaranteed; to this end, this invention provides solutions in the following two aspects.

[0008] In a first aspect, the present invention provides a laser cutting control method for a battery casing, comprising: The equipment acquires operating data during the cutting process of the battery casing within a set time period. The operating data includes multiple dimension sequences, including the actual position of the laser cutting, the commanded position, the actual speed, the commanded speed, the real-time power and the commanded power, and the current of the servo motor. Based on the operating data within a set time period, an improved particle swarm optimization algorithm is used for iterative optimization to obtain the command speed and command power at the next moment, so as to control the laser cutting of the battery casing. The improved particle swarm optimization algorithm includes an adaptive inertia weight, which is positively correlated with the difference between the minimum inertia weight, the maximum inertia weight, and the minimum inertia weight, as well as the product of the energy oscillation index and the number of iterations. The energy oscillation index characterizes the degree of oscillation in the laser cutting process.

[0009] Optionally, the adaptive inertia weight is: ; in, This represents the adaptive inertia weight during the k-th iteration at time t. and These are the minimum inertia weight and the maximum inertia weight in the particle swarm optimization algorithm, respectively. Let k be a hyperbolic cosine function, and k be the number of iterations. Let be the energy oscillation index at time t.

[0010] Optionally, the energy oscillation index is: ; in, This represents the energy oscillation index at time t. This represents the hysteresis distortion index at time t. and Let represent the values ​​of the j-th elements in the differential following error sequence and the differential current sequence at time t, respectively. The preset hyperparameters are used; the differential follower error sequence and differential current sequence are obtained by performing first-order differential processing on the corresponding follower error sequence and current data within the sliding window, respectively; the follower error sequence is composed of the Euclidean distance between the actual position and the command position of the laser cutting at each moment within the sliding window; the sliding window is a window of multiple moments before any given moment, taken from a set time period, starting from any given moment.

[0011] The aforementioned energy oscillation index combines the dynamic relationship between current and error, effectively quantifying the degree of micro-oscillation and providing a quantitative basis for evaluating the quality of the cut surface.

[0012] Optionally, the hysteresis distortion index is: ; in, The hysteresis distortion index is represented at time t, N is the length of the sliding window, and i is the index of the time within the sliding window. This represents the value of the i-th element in the following error sequence at time t. This represents the value of the i-th element in the current sequence at time t. This represents the value of the i-th element in the command velocity sequence at time t. These are the preset hyperparameters.

[0013] The aforementioned hysteresis distortion index can accurately identify the nonlinear region of mechanical transmission under the influence of backlash and static friction, overcoming the shortcomings of traditional methods that are insensitive to specific working conditions.

[0014] Optionally, the particle swarm optimization algorithm further includes a fitness function, which is positively correlated with the product of the instruction velocity and instruction power at the next time step to be optimized, and negatively correlated with the standard deviation of the following error within the sliding window corresponding to the previous time step.

[0015] Optionally, the fitness function is calculated as follows: ; in, This represents the fitness function to be optimized at time t+1. This represents the instruction speed at time t+1, which is to be optimized. This represents the instruction power at time t+1, which is to be optimized. This represents the standard deviation of the following error within the sliding window corresponding to time t. These are the preset hyperparameters.

[0016] Optionally, the method further includes a preprocessing step for the operational data, the preprocessing including: The current sequence formed by the current of the servo motor is filtered using a wavelet threshold denoising method. The actual position sequence and the instruction position sequence, respectively, are synchronized and aligned based on the collection timestamp.

[0017] Optionally, the commanded position indicates the target position that the device needs to reach; the commanded power indicates the target power data of the device during cutting; and the commanded speed indicates the target speed of the device during cutting.

[0018] Optionally, it also includes normalizing all dimensional sequences.

[0019] In the second aspect, the laser cutting control device for the battery casing includes: processor; The memory stores computer instructions for a laser cutting control method for a battery casing, which, when executed by the processor, cause the device to perform the aforementioned laser cutting control method for the battery casing.

[0020] The beneficial effects of this invention are as follows: This invention improves upon the fixed weights of the traditional particle swarm optimization algorithm by combining the operating data of the cutting equipment during the laser cutting process of the battery casing. It directly maps physical oscillation characteristics to the search strategy of the algorithm, solving the technical problem that the traditional particle swarm algorithm is prone to premature convergence and getting trapped in local optima in complex multi-peak problems. It obtains adaptive weights and can adjust servo parameters in real time according to the dynamic changes of the cutting trajectory during the optimization process. This effectively eliminates slag and ripple phenomena at sharp corners, achieving a dual improvement in processing efficiency and cutting accuracy. It meets the stringent requirements of high-end laser processing for high dynamic performance and significantly improves the control performance of laser cutting. Attached Figure Description

[0021] Figure 1 The flowchart illustrating the steps of the laser cutting control method for a battery casing in this embodiment is shown schematically. Figure 2 A line graph illustrating the variation of adaptive inertia weights with the number of iterations is shown. Figure 3 This diagram illustrates the iterative convergence of the traditional particle swarm optimization algorithm during the iterative process. Figure 4 The diagram illustrates the iterative convergence of the improved particle swarm optimization algorithm during the iterative process.

[0022] Figure 5 The schematic diagram illustrates the structural block diagram of the laser cutting control system for the battery casing in this embodiment. Detailed Implementation

[0023] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0024] This embodiment uses a high-power fiber laser cutting device for battery casing cutting as an example to describe in detail the laser cutting control method for battery casings of the present invention. This device is equipped with a high-performance servo drive and a real-time bus controller.

[0025] Specifically, such as Figure 1 As shown, the laser cutting control method for the battery casing in this embodiment includes the following steps: Step S1: Obtain the operating data of the equipment during the set time period in the process of cutting the battery casing.

[0026] In controlling the laser cutting of the battery casing, it is first necessary to acquire key data reflecting the dynamic performance and processing status of the machine tool in real time.

[0027] Specifically, the CNC equipment and servo drive of the laser cutting machine are used to collect operating data in real time within a set time period at a set sampling frequency.

[0028] The operational data within the set time period in this embodiment includes a series of multiple dimensions, including the actual position during cutting, the commanded position, the actual speed, the commanded speed, the real-time power, the commanded power, and the servo motor current. The commanded position represents the target position the device needs to reach, the commanded speed represents the device's set feed speed (target speed), and the commanded power represents the device's set output energy (target power).

[0029] The sampling frequency mentioned above is 100Hz, but it can be determined according to the actual situation.

[0030] Because the collected operational data contains high-frequency electromagnetic noise and mechanical vibration interference, in order to eliminate the impact of high-frequency electromagnetic noise and mechanical vibration on the accuracy of the operational data, this embodiment preprocesses the collected raw data as follows: Filtering: For the current sequence composed of current data from the servo motor, a wavelet thresholding denoising method is used. Specifically, the db4 wavelet basis function is selected to perform a 5-level decomposition of the current signal, a soft thresholding function is used to process the high-frequency coefficients, and finally the signal is reconstructed to retain the low-frequency characteristics reflecting load changes in the current signal and remove high-frequency noise.

[0031] Time alignment: Due to the different delays in sensor and bus transmission, the data collected in various dimensions are not synchronized. It is also necessary to synchronize and align the actual position sequence and the command position sequence according to the collection timestamp.

[0032] Normalization: Perform maximum and minimum value normalization on all dimensional sequences to map the data to... Intervals are used to eliminate differences in dimensions.

[0033] Step S2: Based on the running data within a set time period, an improved particle swarm optimization algorithm is used for iterative optimization to obtain the instruction speed and instruction power at the next moment.

[0034] In this embodiment, in order to solve the problem that the traditional particle swarm optimization (PSO) algorithm is prone to getting trapped in local optima under complex laser cutting conditions, an improved PSO algorithm including adaptive inertia weights and fitness functions is constructed.

[0035] The specific optimization process includes the following sub-steps: Step S21: Calculate the hysteresis distortion index.

[0036] During laser cutting of battery casings, especially when performing right-angle turns, sharp-angle cutting, or high-speed circular interpolation, significant abrupt changes occur due to backlash and sudden changes in static friction in the mechanical transmission chain (such as gears, racks, and ball screws). This physical characteristic manifests as follows: when the feed speed approaches zero or reverses direction, although the servo motor outputs a large current, the actuator fails to respond promptly due to the clamping force of static friction, resulting in a non-linear hysteresis relationship between the following error and the command speed. This characteristic directly reflects a sharp decrease in the dynamic stiffness of the equipment under specific operating conditions. If the algorithm fails to perceive this characteristic and continues to optimize parameters according to the logic of the conventional linear region, it will lead to overcutting or rounding at sharp corners. Therefore, this application constructs a hysteresis distortion index to reflect this hysteresis characteristic.

[0037] Specifically, the process of obtaining the lag distortion index is as follows: First, obtain a sliding window that includes the current time from the set time period.

[0038] The sliding window is based on the current time. Starting from this point, extract the data up to that time. A window of sampling points. In this embodiment, N is 60.

[0039] Secondly, the sequence of command speed and servo motor current data at each moment within the sliding window is recorded as the command speed sequence and current sequence. The Euclidean distance between the actual cutting position and the command position is used as the following error, and the sequence of following errors at each moment within the sliding window is recorded as the following error sequence.

[0040] Then, based on the follow-up error sequence, current sequence, and command speed sequence, a hysteresis distortion index is constructed.

[0041] Among them, the lag distortion index for: ; in, express The time-lag distortion index; This is the index of the time within the sliding window; express The following error sequence corresponding to time 1 One following error; express The current sequence corresponding to time 1 One current value; express The first instruction velocity sequence corresponding to time moment Speed ​​per instruction; To prevent preset hyperparameters with a denominator of zero.

[0042] The above preset hyperparameters .

[0043] From the above formula, it can be seen that when the cutting speed is low ( Smaller (i.e., close to zero or reversed) and exhibits a larger following error ( ) and larger current ( When the value is 0, it indicates that the device is in the low-response, nonlinear friction-dominated region. At this time, the numerator increases significantly and the denominator is small. The value will increase significantly, thus enabling the hysteresis characteristics of mechanical transmission to be more acutely detected.

[0044] Step S22: Calculate the energy oscillation index.

[0045] If a nonlinear distortion occurs during the cutting process, laser cutting often enters a brief energy release phase. This is because excessive control energy has accumulated in the previous integration stage. When the mechanical components overcome static friction and begin to move, energy overshoot can easily occur, causing high-frequency vibration of the cutting head at the microscale. This characteristic is manifested in the data as: in the hysteresis distortion index... Shortly after reaching its peak value, the servo motor's current and position error (following error) exhibit phase reversal or disordered high-frequency fluctuations. This means that energy is not effectively converted into cutting feed but is dissipated as mechanical vibration. This oscillation directly affects the roughness of the cut surface. If the algorithm parameters result in insufficient damping of the equipment, this oscillation will persist. Therefore, it is necessary to construct an energy oscillation index based on the hysteresis distortion index to characterize the failure of energy coupling and the degree of oscillation.

[0046] Specifically, the process of obtaining the energy oscillation index is as follows: First, the following error sequence and the current sequence are subjected to first-order difference processing to obtain the differential following error sequence and differential current sequence, respectively.

[0047] Secondly, based on the differential follower error sequence and the differential current sequence, the energy oscillation index at the current moment is calculated.

[0048] The formula for the energy oscillation index is: ; in, express Energy oscillation index at any given moment; and They represent The differential follower error sequence and differential current sequence corresponding to time 1 Differential follower error, differential current, To prevent the default hyperparameter from having a denominator of zero, the value is set to 1.

[0049] Since oscillations are often a follow-up effect of nonlinear distortion, the hysteresis distortion index is used as a weight. Furthermore, when equipment oscillates, the current and error often exhibit inverse or chaotic changes, making... Typically, a negative value is calculated such that... Smaller, close to -1, corresponding to the calculated The magnitude is relatively large, resulting in a relatively large energy oscillation index in the final calculation.

[0050] The larger the energy oscillation index, the less damped the equipment is under the current control parameters, resulting in low energy coupling efficiency. This allows for more sensitive detection of minute mechanical chatter and reduces surface ripples during subsequent optimization control.

[0051] Step S23: Calculate the adaptive inertia weights.

[0052] In particle swarm optimization (PSO), the inertia weight determines the degree to which particles maintain their previous velocity. Specifically, when using PSO to control the laser cutting of a battery casing at time t, if the calculated energy oscillation index is large, it indicates that the current combination of control parameters cannot effectively suppress the oscillations caused by mechanical nonlinearity. If the PSO algorithm maintains a small inertia weight at this point, the particles will cluster around their current position for a fine-tuned search, eventually converging to a local optimum. However, since the current control parameter combination cannot effectively suppress the oscillations, the resulting local optimum not only fails to eliminate the oscillations but may also exacerbate the instability of the equipment due to parameter fine-tuning. Therefore, when significant physical oscillations occur, the PSO algorithm should promptly exit the current solution space, corresponding to a higher inertia weight in the algorithm.

[0053] Specifically, this embodiment uses the energy oscillation index to dynamically adjust the inertia weight, as shown in the following formula: ; in, express At this moment Adaptive inertia weights during the next iteration; and These are the minimum inertia weight and the maximum inertia weight, respectively. It is a hyperbolic cosine function; This represents the current iteration number.

[0054] In this embodiment, the minimum inertia weight Preferred settings The maximum inertia weight Preferred settings .

[0055] The above formula establishes a direct mapping between physical oscillation characteristics and algorithmic exploration capabilities. The formula utilizes the reciprocal of the hyperbolic cosine function to construct an S-shaped variation curve. When... A smaller value indicates weak oscillations and good control, or a smaller number of iterations indicates that the algorithm has just begun iterating. When the term is close to 1, the result inside the parentheses approaches 0. Approaching The algorithm tends to explore locally, achieving fine convergence; when A significant increase indicates severe oscillations, and the algorithm is prone to getting trapped in local optima. The term increases sharply, and the result inside the parentheses approaches 1. Approaching rapidly This allows the algorithm to automatically increase the inertia weight when it detects physical oscillations, forcing particles to perform a large-scale global search to escape the region of poor solutions. This ensures that parameter adjustments are always directed towards eliminating oscillations and improving stability.

[0056] like Figure 2 As shown, different energy oscillation indices correspond to different adaptive inertia weights, and have different adaptive inertia weights at different iteration numbers.

[0057] Step S24: Construct the fitness function and perform iterative optimization.

[0058] The core of the particle swarm optimization algorithm lies in finding the solution that optimizes the fitness function. The goal of this embodiment is to maximize processing efficiency while ensuring processing accuracy. Therefore, the following fitness function is constructed. for: ; in, Indicates areas for optimization The fitness function value at time t; The particle represents the area to be optimized. The speed of instructions at any given moment; The particle represents the area to be optimized. The power of the command at any given moment; express The standard deviation of the following error within the sliding window corresponding to each time point characterizes the stability and accuracy of the machining process. To prevent the preset hyperparameter from having a denominator of zero, the value is set to 1 in this embodiment.

[0059] The above formula shows that the fitness function is directly proportional to the product of speed and power (i.e., processing efficiency) and inversely proportional to the standard deviation of the following error (i.e., processing deviation). The algorithm aims to find a set of... This allows for the output of maximum processing energy and speed while minimizing error fluctuations.

[0060] Specifically, the execution steps of the improved particle swarm optimization algorithm are as follows: (a) Initialization: Initialize the particle swarm, with a maximum number of iterations of 50, and set the particle swarm size to 50. Each particle's position vector contains two dimensions: command velocity. and command power And randomly initialize the position and velocity of the particles.

[0061] (ii) Iterative update: For the first In the next iteration, based on the calculated adaptive inertia weights for the current iteration... The state of each particle is updated using the standard PSO velocity update formula and position update formula: ; ; in, and These represent the particle's velocity and position, respectively. The learning factor (preferred in this embodiment) ), for Random numbers between For individual extreme values, Let k be the global extremum and k be the index of the iteration number.

[0062] It should be noted that in each iteration, the algorithm no longer uses fixed or linearly decreasing inertial weights, but dynamically changes them according to the actual operating conditions of the laser and the number of iterations. This improvement can automatically increase the inertial weights when the device exhibits nonlinear abrupt changes and oscillations, and automatically shrink the search range when the device is stable, thereby achieving a dynamic balance between global search capability and local convergence accuracy.

[0063] (iii) Fitness evaluation: Substitute the updated particle positions (i.e., the predicted command velocity and power) into the fitness function. Perform the calculation.

[0064] (iv) Termination condition: Determine whether the maximum number of iterations has been reached or whether the change in the fitness value of the fitness function is less than a preset threshold. If the conditions are met, output the position vector corresponding to the globally optimal particle, which is the optimized command velocity for the next time step. and command power .

[0065] like Figure 3 and Figure 4 As shown, compared with the traditional particle swarm optimization algorithm, the improved particle swarm optimization algorithm can solve the problems of premature convergence and getting trapped in local optima in complex multimodal problems.

[0066] Step S3: Use the command speed and command power at the next moment to control the laser cutting of the battery casing.

[0067] The optimized next-moment command speed and command power are sent to the motion controller and laser controller of the equipment to achieve real-time optimized control of the cutting process.

[0068] Compared to traditional particle swarm optimization algorithms, the solution in this embodiment enables the device to automatically adjust the command speed and command power according to different cutting trajectories when controlling the laser cutting of the battery casing, thereby suppressing overshoot and oscillation at corners.

[0069] The present invention introduces a hysteresis distortion index and an energy oscillation index to accurately capture the nonlinear physical characteristics of the cutting equipment under complex trajectories such as sharp corners and arcs. By using the above physical characteristics to construct an adaptive inertial weight, the particle swarm algorithm can automatically increase the search range when the oscillation intensifies and converge quickly when the processing is stable. This achieves precise control of the contour accuracy of the laser cutting of battery casings under high-speed processing, effectively avoiding overcutting, slag buildup, and ripples.

[0070] The present invention also provides a laser cutting control system for battery casings. For example... Figure 5As shown, the system includes a processor and a memory, the memory storing computer program instructions, which, when executed by the processor, implement the laser cutting control method for battery casing described above according to the present invention.

[0071] The device also includes other components well known to those skilled in the art, such as communication buses and communication interfaces, the settings and functions of which are known in the art and therefore will not be described in detail here.

[0072] In this invention, the aforementioned memory can be any tangible medium containing or storing a program that can be used or combined with an instruction execution device, apparatus, or apparatus. For example, a computer-readable storage medium can be any suitable magnetic or magneto-optical storage medium, such as Resistive Random Access Memory (RRAM), Dynamic Random Access Memory (DRAM), Static Random Access Memory (SRAM), Enhanced Dynamic Random Access Memory (EDRAM), High-Bandwidth Memory (HBM), Hybrid Memory Cube (HMC), etc., or any other medium that can be used to store desired information and can be accessed by an application, module, or both. Any such computer storage medium can be part of a device or accessible to or connected to a device. Any application or module described in this invention can be implemented by computer-readable / executable instructions stored or otherwise maintained on such a computer-readable medium.

[0073] In the description of this specification, "multiple" means at least two, such as two, three or more, etc., unless otherwise expressly and specifically defined.

[0074] While various embodiments of the invention have been shown and described in this specification, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Many modifications, alterations, and alternatives will occur to those skilled in the art without departing from the spirit and essence of the invention.

Claims

1. A method for laser cutting control of a battery case, characterized by, include: The device acquires operating data during the cutting process of the battery casing within a set time period. The operating data includes a series of multiple dimensions, including the actual position, commanded position, actual speed, commanded speed, real-time power, commanded power, and servo motor current during cutting. Based on the operating data within a set time period, an improved particle swarm optimization algorithm is used for iterative optimization to obtain the command speed and command power at the next moment, so as to control the laser cutting of the battery casing. The improved particle swarm optimization algorithm includes an adaptive inertia weight, which is positively correlated with the difference between the minimum inertia weight, the maximum inertia weight, and the minimum inertia weight, as well as the product of the energy oscillation index and the number of iterations. The energy oscillation index characterizes the degree of oscillation in the laser cutting process.

2. The laser cutting control method for a battery case according to claim 1, wherein, The adaptive inertia weight is: ； wherein, represents the adaptive inertia weight in the kth iteration process at time t, and are the minimum inertia weight and the maximum inertia weight in the particle swarm optimization algorithm, respectively, is the hyperbolic cosine function, and k is the iteration number, is the energy oscillation index at time t.

3. The laser cutting control method for a battery case according to claim 1, wherein The energy oscillation index is: ； wherein, denotes the energy fluctuation index at time t, denotes the hysteresis distortion index at time t, and respectively denote the value of the jth element in the differential follow-up error sequence and the differential current sequence corresponding to time t, is a preset hyperparameter; the differential follow-up error sequence and the differential current sequence are obtained by first-order difference processing of the corresponding follow-up error sequence and current data in the sliding window; the follow-up error sequence is composed of the Euclidean distance between the actual position and the command position of laser cutting at each time in the sliding window; the sliding window is a window of multiple time points before any time point as the starting point, which is intercepted within a set time period.

4. The laser cutting control method for a battery case according to claim 3, wherein The hysteresis distortion index is: ； wherein, represents the hysteresis distortion index at time t, N is the length of the sliding window; i is the index of time in the sliding window; represents the value of the i-th element in the following error sequence corresponding to time t, represents the value of the i-th element in the current sequence corresponding to time t, represents the value of the i-th element in the command speed sequence corresponding to time t, is a preset hyperparameter.

5. The laser cutting control method for a battery case according to claim 1, wherein The particle swarm optimization algorithm also includes a fitness function, which is positively correlated with the product of the instruction speed and instruction power at the next time step to be optimized, and negatively correlated with the standard deviation of the following error within the sliding window corresponding to the previous time step.

6. The laser cutting control method for battery casing according to claim 5, characterized in that, The fitness function is calculated as follows: ； in, This represents the fitness function to be optimized at time t+1. This represents the instruction speed at time t+1, which is to be optimized. This represents the instruction power at time t+1, which is to be optimized. This represents the standard deviation of the following error within the sliding window corresponding to time t. These are the preset hyperparameters.

7. The laser cutting control method for battery casing according to claim 1, characterized in that, It also includes a preprocessing step for the running data, the preprocessing including: The current sequence formed by the current of the servo motor is filtered using a wavelet threshold denoising method. The actual position sequence and the instruction position sequence, respectively, are synchronized and aligned based on the collection timestamp.

8. The laser cutting control method for battery casing according to claim 1, characterized in that, The commanded position indicates the target position that the equipment needs to reach; the commanded power indicates the target power data of the equipment during cutting; and the commanded speed indicates the target speed of the equipment during cutting.

9. The laser cutting control method for battery casing according to claim 7, characterized in that, It also includes normalizing all dimensional sequences.

10. A laser cutting control system for battery casings, characterized in that, include: processor; A memory storing computer instructions for a laser cutting control method for a battery casing, which, when executed by the processor, cause the device to perform the laser cutting control method for a battery casing according to any one of claims 1-9.

Images