Vision-based laser cutting wall-to-wall loss prevention method

Abstract

The invention relates to the technical field of laser cutting, in particular to a laser cutting wall-to-wall loss prevention method based on vision. According to the technical scheme, the method comprises the following steps that a component and an aluminum framework are fixed, a to-be-machined pattern is set, laser machining parameters are initialized, seed parameters of the half-cutting process are determined, first-time laser machining is executed based on the initialized parameters, laser emission is stopped after set machining times are completed, and a cutting channel is photographed through a high-power vision camera; and processing the shot image by adopting an image processing algorithm, extracting cutting channel information, and judging whether cutting is completed or not according to an image analysis result. According to the method, a visual algorithm is combined with dynamic management and control, the problem of wall damage prevention in the laser cutting process is effectively solved on the premise of not depending on expensive equipment, meanwhile, the cutting precision and the machining efficiency are considered, and the method is suitable for scenes such as micro devices with high requirements for machining precision and damage prevention.

Description

Technical Field

[0001] The present invention relates to the technical field of laser cutting, and in particular to a vision-based laser cutting wall loss prevention method. Background Art

[0002] With the rapid development of 3C and semiconductor industries, the size and volume of devices continue to shrink, and the requirements for the accuracy of wall damage prevention during laser cutting are becoming increasingly stringent. The current protection methods for wall damage prevention in the industry have obvious limitations:

[0003] The passive injury prevention method of adding fillers is limited by the trend of miniaturization of device size and can no longer meet the injury prevention needs in a small space;

[0004] Installing an acoustic and photoelectric sensor (such as a spectral device) on the bottom for light leakage detection is not only expensive, but also requires a lot of supporting verification work and has poor adaptability.

[0005] The solution of using coaxial vision to perceive the cutting completion status and light leakage lacks a dynamic control mechanism. When the cutting path is only partially completed, the subsequent operation strategy cannot be clearly defined, which can easily lead to damage to the wall.

[0006] Therefore, there is an urgent need for a laser cutting wall loss prevention technology that can dynamically adjust the laser trajectory and power, intelligently predict the detection timing, and take into account processing efficiency, so as to solve the defects of existing solutions in micro-device processing. Summary of the Invention

[0007] The purpose of the present invention is to address the problem of wall damage prevention during laser cutting in the background technology and to propose a vision-based laser cutting wall loss prevention method.

[0008] The technical solution of the present invention is a vision-based laser cutting wall loss prevention method, comprising the following steps:

[0009] S1. Fix the components and aluminum frame and set the graphics to be processed;

[0010] S2, initialize laser processing parameters and determine seed parameters of half-cut process;

[0011] S3, performing the first laser processing based on the initialization parameters;

[0012] S4, after completing the set number of processing times, the laser stops emitting light;

[0013] S5. Take pictures of the cutting path using a high-magnification visual camera;

[0014] S6. Using an image processing algorithm to process the captured image and extract cutting path information;

[0015] S7, judging whether the cutting is completed according to the image analysis result;

[0016] S8. If cutting is not completed, dynamically adjust the cutting trajectory and laser power;

[0017] S9, perform aging control to optimize processing efficiency;

[0018] S10, saving the optimized parameters as the initialization parameters for the next round;

[0019] S11. After cutting is completed, the material is fed through the automated system.

[0020] Optionally, fixing the components and the aluminum frame and setting the graphics to be processed in step S1 specifically includes: placing the components and the aluminum frame in corresponding carriers and fixing them, and setting the style and size parameters of the graphics to be processed in the canvas.

[0021] Optionally, in step S2, the laser processing parameters are initialized, and the seed parameters of the half-cutting process are determined, specifically including: the process personnel debug the process parameters according to the cutting trajectory, and the process parameters include laser power, speed, focus position, and number of processing times to ensure that the aluminum skeleton has no cut parts and no damage under the process parameters. The process parameters are used as seed parameters for subsequent algorithm iterative optimization.

[0022] Optionally, in step S5, photographing the cutting path with a high-magnification visual camera specifically includes: using a coaxial or paraxial high-magnification visual camera to photograph the cutting path with auxiliary light to ensure that the cutting path is clearly visible and can capture the cutting path area that has been cut in advance due to inconsistent material height.

[0023] Optionally, in step S6, using an image processing algorithm to process the captured image and extract cutting lane information specifically includes:

[0024] Intercept the cutting path area ROI to improve processing speed;

[0025] Binarize the image to enhance contrast;

[0026] Perform blob analysis to remove high noise points and extract cutting path information;

[0027] The center cutting line is fitted based on the cutting lane information on both sides. If there is a completed cutting area, it is removed by fitting.

[0028] Optionally, determining whether the cutting is completed according to the image analysis result in step S7 specifically includes:

[0029] If the image algorithm detects that the cutting path is completely transparent, it is determined that the cutting is completed;

[0030] If there is a non-transparent area in the cutting path, it is determined that the cutting is not completed.

[0031] Optionally, the dynamic adjustment of the cutting trajectory and laser power in step S8 specifically includes:

[0032] The cutting trajectory is dynamically adjusted, and matrix transformation is performed based on the trajectory array and spatial coordinate system information output by the image algorithm to generate canvas trajectory information to update the uncut trajectory;

[0033] The laser power is dynamically adjusted, and the laser power output is reduced by 10% to 90% based on the image analysis results. For aluminum parts with a thickness of 0.07 to 0.09 mm, the power needs to be reduced after the initial cutting is completed to a depth of 0.04 to 0.06 mm to avoid overcutting and damaging the carbon steel underneath.

[0034] Optionally, performing aging control in step S9 to optimize processing efficiency specifically includes:

[0035] Set a processing times threshold. If the current parameters fail to complete lossless cutting within the threshold times, the cutting parameters for the next product will be self-optimized.

[0036] If the cutting power is not completed after the second cutting power is reduced by 60% and the third cutting power is reduced by 80%, it is adjusted to reduce by 40% and 80% for the second cutting power and to achieve lossless cutting within 3 times.

[0037] Optionally, saving the optimized parameters as the initialization parameters for the next round in step S10 specifically includes: saving the parameters optimized in the previous cycle, and as the number of cutting increases, eliminating abnormal data that deviates from the mean of the parameters of the same batch by more than 20%. For example, for the same product, if the processing is completed in 10 times and the number is within 5 times, the abnormal value appears outside the range. For example, if the processing number is 6 times, a deviation of 20% from the standard is still considered to meet the requirements, but if the processing number is 7 times, a deviation of 40% from the benchmark is considered to be abnormal data processing; if the processing number is too small, it is calculated synchronously according to this method.

[0038] Compared with the prior art, this application has at least one of the following beneficial technical effects:

[0039] This invention overcomes the limitations of passive wall protection methods such as adding fillers due to the limited space available, making it suitable for micro-device processing. It eliminates the need for expensive testing equipment such as spectroscopy, reducing equipment costs and the complexity of supporting verification work. It can dynamically adjust the laser trajectory and optimize the processing path of uncut areas based on visual algorithm analysis results, improving cutting accuracy.

[0040] This invention dynamically adjusts laser power based on the actual cutting path, avoiding overcutting and damage to the underlying material caused by excessive power, thus achieving non-destructive cutting. Through time control and iterative parameter optimization, it can also dynamically adjust processing parameters to improve cutting efficiency and reduce processing times. This invention features an intelligent cutting status determination mechanism, combined with visual inspection, to accurately determine whether the cut is complete, avoiding ineffective or insufficient processing.

[0041] The present invention combines visual algorithms with dynamic control to effectively solve the problem of wall damage prevention during laser cutting without relying on expensive equipment. It also takes into account both cutting accuracy and processing efficiency, and is suitable for scenarios such as micro devices that have high requirements for processing accuracy and damage prevention. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 The figure is a flow chart of a vision-based method for preventing wall loss in laser cutting. DETAILED DESCRIPTION

[0043] The following describes the embodiments of the present invention through specific examples. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments. The details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the following embodiments and features in the embodiments can be combined with each other unless they conflict.

[0044] Example

[0045] like Figure 1 As shown, a vision-based laser cutting wall loss prevention method of the present invention includes the following steps:

[0046] Step S1: Place the components and a 0.08mm thick aluminum frame in the corresponding carrier and secure them. Set the pattern to be processed on the canvas. The carrier secures the components and aluminum frame to a stable position during processing, preventing deviations in the cutting path due to displacement. The pre-set processing pattern on the canvas provides a precise path reference for subsequent laser processing, ensuring the accuracy of the initial processing.

[0047] Step S2: Initialize the laser processing parameters. For the corresponding cutting trajectory, the process technician adjusts the process parameters, such as laser power, speed, laser focus position, and number of processing times. This ensures that the skeleton is not cut through at all under the current cutting parameters, ensuring no damage. Subsequent algorithm iterations optimize these seed parameters. Pre-adjusted seed parameters ensure that the initial cut does not damage the skeleton, providing a safe and reliable foundation for subsequent algorithm iterations and avoiding the risk of wall damage caused by inappropriate initial parameters.

[0048] Step S3: The first laser process begins, and the aluminum frame is cut based on the initialized parameters. Starting with the verified seed parameters, the initial cut is completed while ensuring wall safety, providing an actual cutting sample for subsequent visual inspection and parameter adjustment.

[0049] Step S4: After the processing times are completed, the laser stops emitting light. This stops emitting light according to the preset number of times to avoid power accumulation damage caused by meaningless continuous processing and to provide a stable laser-free environment for visual inspection.

[0050] Step S5: Use a coaxial or paraxial high-magnification camera to photograph the cutting path to ensure clear visibility. Due to material height inconsistencies, some cutting paths may be completed prematurely. The camera imaging can clearly capture the completed areas with auxiliary lighting. The high-magnification camera combined with auxiliary lighting can clearly capture the details of the cutting path, especially the areas that are prematurely completed due to material height inconsistencies. This provides high-quality data for subsequent image analysis and ensures inspection accuracy.

[0051] Step S6: Input the captured image. The image processing algorithm can adopt Halcon or OpenCV.

[0052] Cutting path area ROI interception to improve algorithm processing speed.

[0053] Binarize the corresponding images to highlight the contrast.

[0054] Blob analysis removes corresponding high noise and extracts cutting path information.

[0055] Cutting path information processing: a. Fit the center cutting line (i.e., the laser cutting trajectory) based on the cutting path information on both sides. b. If a portion of the cutting has been completed, remove the corresponding area through fitting. This provides an accurate basis for trajectory adjustment and avoids repeated processing of completed parts.

[0056] Step S7: Determine whether the cutting is completed:

[0057] a. Through the above image algorithm processing, detect whether the cutting path is transparent. If the cutting work has been completed, the judgment result is yes.

[0058] b. If there are still opaque areas on the cutting path, the result is judged as negative. Objective judgment based on images replaces manual experience, accurately identifying uncut areas, avoiding under-processing or over-cutting due to misjudgment, and improving the stability of cutting quality.

[0059] Step S8: a. Dynamic adjustment of cutting trajectory: Matrix transformation is performed based on the trajectory array and spatial coordinate system information output by the image algorithm to become canvas trajectory information, completing the dynamic transformation of the uncut trajectory. Dynamic adjustment of the trajectory ensures that the laser only acts on the uncut area, reducing invalid paths and improving processing efficiency. b. Dynamic adjustment of laser power: Based on the results of image algorithm processing, the laser power output is reduced by 60%. Reducing power can avoid overcutting (for example, after a 0.08mm aluminum part has been cut by 0.05mm, low power can prevent damage to the carbon steel below), fundamentally solving the problem of wall damage prevention.

[0060] Step S9: Time Control. This stage primarily improves cutting efficiency. For example, if the power is reduced by 60% for the second cut and 80% for the third, but the cut still fails, the cutting parameters for the next product will be self-optimized. The power is reduced by 40% for the second cut and 80% for the third, achieving a lossless cut in three attempts. Alternatively, a two-shot cut can be completed, which requires algorithmic self-updates and optimization. Dynamically adjusting parameters based on historical processing data shortens the processing cycle while ensuring lossless cutting, significantly improving overall processing efficiency and adapting to mass production needs.

[0061] Step S10: Save the parameters optimized based on the previous cycle, and use the optimized parameters as the initialization parameters for the next round of cutting. As the number of cutting increases, the abnormal data is eliminated, and the mean of multiple parameters is used as the initialization parameters for the next round of cutting. Accumulating optimal process data through parameter iteration and taking the mean after eliminating abnormal values ​​can reduce accidental errors, make the initialization parameters of subsequent processing more accurate, reduce debugging costs, and improve process stability. As the number of cutting increases, eliminate abnormal data that deviates from the mean of the parameters of the same batch by more than 20%. For example, for the same product, if the processing is completed 5 times out of 10 times, the abnormal value appears outside the range. For example, if the processing number is 6 times, a deviation of 20% from the standard is still considered to meet the requirements, but if the processing number is 7 times, a deviation of 40% from the benchmark is considered to be abnormal data processing; if the processing number is too small, it is calculated synchronously according to this method.

[0062] Step S11: After the aluminum frame is cut, the material is fed through the automated system. Automated feeding reduces manual intervention and the risk of positional deviation caused by manual operation, while also enabling continuous production and further improving overall processing efficiency.

[0063] The present invention dynamically adjusts the laser trajectory and generates the latest trajectory based on the size and shape of the incompletely cut path analyzed by a visual algorithm. This ensures that the laser acts only on the uncut area, avoiding accidental damage to the completed cut area or the opposite wall. It also dynamically adjusts the laser power, reducing it according to the cutting situation (e.g., by 10% to 90%) to prevent damage to the opposite wall caused by light leakage due to excessive power. This method is particularly suitable for scenarios where fragile materials such as carbon steel are located underneath 0.08mm thick aluminum parts. The cutting path is photographed using a coaxial or paraxial high-magnification visual camera. Combined with image processing algorithms such as Halcon or Opencv, the cutting path information is accurately extracted through steps such as ROI interception, binarization, and Blob analysis. The center cutting line is fitted and the completed area is removed, providing an accurate basis for trajectory adjustment and reducing cutting deviations.

[0064] The algorithm incorporates time-based control, dynamically adjusting power levels based on the cutting performance of each process. This reduces the number of processing cycles (for example, from three to two) through parameter self-optimization. Dynamic trajectory adjustment also avoids reprocessing completed sections, significantly improving overall processing efficiency. This eliminates the need for expensive testing equipment such as spectral detection, reducing equipment investment costs. Extensive verification work is also unnecessary, and iterative parameter optimization reduces the frequency of manual debugging and operational complexity.

[0065] It is worth noting that by initializing seed parameters and iteratively optimizing them, the optimized parameters are saved as the initial parameters for the next round of cutting. After eliminating abnormal data, the average of multiple parameter values ​​is used as the initial value, making the processing parameters more stable and reliable, ensuring consistent cutting quality. It can also intelligently predict the timing of stopping the light and taking photos. After the processing is completed, the light is stopped and a photo is taken for detection. Combined with image algorithms, it determines whether the cutting is complete, avoiding ineffective or insufficient processing, and improving the intelligence level of the cutting process.

[0066] The above specific embodiments are merely several optional embodiments of the present invention. Based on the technical solutions of the present invention and the relevant inspirations of the above embodiments, those skilled in the art may make various alternative improvements and combinations to the above specific embodiments.

Claims

1. A vision-based laser cutting wall loss prevention method, characterized in that: The following steps are involved: S1. Fix the components and aluminum frame and set the graphics to be processed; S2, initialize laser processing parameters and determine seed parameters of half-cut process; S3, performing the first laser processing based on the initialization parameters; S4, after completing the set number of processing times, the laser stops emitting light; S5. Take pictures of the cutting path using a high-magnification visual camera; S6. Using an image processing algorithm to process the captured image and extract cutting path information; S7, judging whether the cutting is completed according to the image analysis result; S8. If cutting is not completed, dynamically adjust the cutting trajectory and laser power; S9, perform aging control to optimize processing efficiency; S10, saving the optimized parameters as the initialization parameters for the next round; S11. After cutting is completed, the material is fed through the automated system.

2. A vision-based laser cutting wall loss prevention method according to claim 1, characterized in that: Fixing the components and the aluminum frame and setting the graphics to be processed in step S1 specifically includes: placing the components and the aluminum frame in corresponding carriers and fixing them, and setting the style and size parameters of the graphics to be processed in the canvas.

3. The vision-based laser cutting wall loss prevention method according to claim 1 is characterized in that: In step S2, the laser processing parameters are initialized, and the seed parameters of the half-cut process are determined, specifically including: the process personnel debug the process parameters according to the cutting trajectory. The process parameters include laser power, speed, focus position, and number of processing times to ensure that the aluminum skeleton has no cut parts and no damage under the process parameters. The process parameters are used as seed parameters for subsequent algorithm iterative optimization.

4. The vision-based laser cutting wall loss prevention method according to claim 1 is characterized in that: In step S5, taking pictures of the cutting path with a high-magnification visual camera specifically includes: using a coaxial or paraxial high-magnification visual camera to photograph the cutting path with auxiliary light, ensuring that the cutting path is clearly visible and can capture the cutting path area that has been cut in advance due to inconsistent material height.

5. The vision-based laser cutting wall loss prevention method according to claim 1 is characterized in that: In step S6, the image processing algorithm is used to process the captured image and extract the cutting lane information, specifically including: Intercept the cutting path area ROI to improve processing speed; Binarize the image to enhance contrast; Perform blob analysis to remove high noise points and extract cutting path information; The center cutting line is fitted based on the cutting lane information on both sides. If there is a completed cutting area, it is removed by fitting.

6. The vision-based laser cutting wall loss prevention method according to claim 1, characterized in that: Determining whether the cutting is completed according to the image analysis result in step S7 specifically includes: If the image algorithm detects that the cutting path is completely transparent, it is determined that the cutting is completed; If there is a non-transparent area in the cutting path, it is determined that the cutting is not completed.

7. The vision-based laser cutting wall loss prevention method according to claim 1 is characterized in that: The dynamic adjustment of cutting trajectory and laser power in step S8 specifically includes: The cutting trajectory is dynamically adjusted, and matrix transformation is performed based on the trajectory array and spatial coordinate system information output by the image algorithm to generate canvas trajectory information to update the uncut trajectory; The laser power is dynamically adjusted, and the laser power output is reduced by 10% to 90% based on the image analysis results. For aluminum parts with a thickness of 0.07 to 0.09 mm, the power needs to be reduced after the initial cutting is completed to a depth of 0.04 to 0.06 mm to avoid overcutting and damaging the carbon steel underneath.

8. The vision-based laser cutting wall loss prevention method according to claim 1 is characterized in that: The aging control in step S9 to optimize the processing efficiency specifically includes: Set a processing times threshold. If the current parameters fail to complete lossless cutting within the threshold times, the cutting parameters for the next product will be self-optimized. If the cutting power is not completed after the second cutting power is reduced by 60% and the third cutting power is reduced by 80%, it is adjusted to reduce by 40% and 80% for the second cutting power and to achieve lossless cutting within 3 times.

9. The vision-based laser cutting wall loss prevention method according to claim 1, characterized in that: Saving the optimized parameters as the initialization parameters for the next round in step S10 specifically includes: saving the parameters optimized in the previous cycle, and as the number of cuts increases, eliminating abnormal data that deviates from the mean of the parameters of the same batch by more than 20%.

Images