Laser cutting parameter adjusting system based on machine learning

Abstract

The invention discloses a laser cutting parameter adjusting system based on machine learning, and relates to the technical field of laser cutting. According to the method, multiple laser cutting parameters in the laser cutting process are collected, a short-term state index and a long-term state index are calculated based on the cutting parameters, and a laser cutting instant index and a laser cutting long-term index are generated respectively; based on the laser cutting instant index and the laser cutting long-term index, optimized laser cutting parameters are generated, and working parameters of the laser cutting machine are adjusted in real time; the adjustment result and the related data are sent back to the data storage module for updating the historical database; a real-time distribution curve is generated and compared with a target distribution curve, deviation is calculated, and a parameter optimization target curve is generated; and according to the optimized sub-targets, laser cutting parameters are adjusted in real time, and the stability and high efficiency of the laser cutting process are ensured. According to the system, accurate control over the laser cutting path is achieved through multi-layer and non-uniform control unit division and time-space correlation model construction.

Description

Technical Field

[0001] The present invention belongs to the technical field of laser cutting, and in particular relates to a laser cutting parameter adjustment system based on machine learning. Background Art

[0002] Laser cutting is a high-precision and high-efficiency material processing technology, which is widely used in cutting a variety of materials such as metals, plastics, and ceramics. Laser cutting achieves cutting by focusing a high-energy laser beam on the surface of the material, causing the material to melt, evaporate, or burn rapidly. The key parameters of laser cutting include laser power, cutting speed, auxiliary gas type and pressure, nozzle diameter, focus position, etc.

[0003] Although laser cutting technology is quite mature, it still faces some challenges in practical applications: difficulty in parameter optimization, insufficient real-time monitoring, lack of long-term performance evaluation, and imperfect data management.

[0004] At present, some laser cutting systems on the market have certain automation functions, but they still have the following limitations:

[0005] Static parameter settings: Most systems rely on preset fixed parameters and cannot be dynamically adjusted according to actual conditions.

[0006] Single indicator monitoring: Existing monitoring systems usually only focus on one or two key parameters and lack comprehensive multi-dimensional monitoring.

[0007] Data silos: Data collection and storage are scattered, lacking a unified management and analysis platform.

[0008] Frequent manual intervention: Frequent manual intervention is required to adjust parameters, which increases the complexity of operation and the probability of error. Summary of the invention

[0009] In order to overcome the shortcomings and deficiencies of the above-mentioned prior art, the present invention adopts the following technical solutions:

[0010] A laser cutting parameter adjustment system based on machine learning, comprising a data acquisition module, a data storage module, a first indicator module, a second indicator module, a parameter optimization module and a real-time adjustment module;

[0011] Data acquisition module: collects parameters during the laser cutting process and sends the data to the data storage module. The parameters include laser power, cutting speed, auxiliary gas type and pressure, nozzle diameter, focus position, cutting quality, material type and thickness;

[0012] Data storage module: stores the data collected by the data acquisition module and is configured with a historical database. The storage content includes laser cutting parameters, cutting quality data and historical data. The historical data includes historical feature vectors, timestamps and cutting quality evaluations.

[0013] The first indicator module: calculates the short-term state indicators of laser cutting based on cutting parameters, generates the instantaneous index of laser cutting, and sends the results to the parameter optimization module;

[0014] The second indicator module: calculates the long-term state indicators of laser cutting based on the cutting parameters, generates the long-term index of laser cutting, and sends the results to the parameter optimization module;

[0015] Parameter optimization module: Generates optimized laser cutting parameters based on the immediate index and long-term index, and sends the optimized parameters to the real-time adjustment module;

[0016] Real-time adjustment module: According to the optimized laser cutting parameters, the working parameters of the laser cutting machine are adjusted in real time, and the adjustment results and related data are sent back to the data storage module for updating the historical database.

[0017] Preferably, the short-term status indicators include the number of laser power fluctuations, the number of cutting speed fluctuations, the number of auxiliary gas pressure fluctuations, the number of nozzle diameter abnormalities, the number of focus position abnormalities and the number of cutting quality abnormalities.

[0018] Preferably, the instantaneous laser cutting index is generated as follows:

[0019] Read the laser power, count the number of laser power fluctuations each time per unit time, as the number of laser power fluctuations; read the cutting speed, count the number of cutting speed fluctuations each time per unit time, as the number of cutting speed fluctuations; read the auxiliary gas pressure, count the number of auxiliary gas pressure fluctuations each time per unit time, as the number of auxiliary gas pressure fluctuations; read the nozzle diameter, count the number of nozzle diameter anomalies each time per unit time, as the number of nozzle diameter anomalies; read the focal position, count the number of focal position anomalies each time per unit time, as the number of focal position anomalies; read the cutting quality, count the number of cutting quality anomalies each time per unit time, as the number of cutting quality anomalies;

[0020] The instant index calculation formula is as follows: Where H represents the instantaneous laser cutting index; I i represents the i-th short-term state indicator; b i is the offset coefficient; ω i is the weight coefficient.

[0021] Preferably, the long-term status indicators include average laser power, average cutting speed, average auxiliary gas pressure, average nozzle diameter, average focus position and average cutting quality.

[0022] Preferably, the laser cutting long-term index is generated as follows:

[0023] The long-term state index is calculated once every statistical period of time T. The method for calculating the average laser power is as follows: Among them, A i represents the average laser power within a statistical period T; P(t) represents the laser power within time t;

[0024] The long-term index calculation formula is as follows: Where L represents the long-term index of laser cutting; L i represents the i-th long-term state indicator; R i is the reference threshold; ω i is the weight coefficient.

[0025] Preferably, the optimization method of the parameter optimization module is as follows:

[0026] S11, obtaining real-time status data of the laser cutting machine, generating a real-time distribution curve according to the real-time status data, receiving a preset target distribution curve, and calculating a deviation between the real-time distribution curve and the target distribution curve; the target distribution curve includes a target laser power distribution curve, a target cutting speed distribution curve, and a target cutting quality distribution curve;

[0027] Calculating the deviation between the target distribution curve and the real-time distribution curve includes:

[0028] The target laser power distribution curve is compared with the real-time laser power distribution curve point by point, and the laser power deviation of the laser power value at the corresponding position is calculated; the target cutting speed distribution curve is compared with the real-time cutting speed distribution curve point by point, and the cutting speed deviation of the cutting speed value at the corresponding position is calculated; the target cutting quality distribution curve is compared with the real-time cutting quality distribution curve point by point, and the cutting quality deviation of the cutting quality value at the corresponding position is calculated;

[0029] S12, calculating a deviation threshold according to the deviation between the target distribution curve and the real-time distribution curve, and when the deviation exceeds the deviation threshold, triggering an over-limit alarm in time, and continuing to generate a parameter optimization target curve;

[0030] The deviation threshold is calculated based on the deviation between the target distribution curve and the real-time distribution curve. The calculation method is as follows:

[0031] Where D represents the deviation threshold; D i Represents the deviation value of the i-th sampling point; Represents the average value of the deviation; σ D represents the standard deviation of the deviation value; N represents the number of sampling points;

[0032] S13, dividing the laser cutting path into a plurality of multi-level, non-uniform control units, each control unit corresponding to a group of intelligent actuators; generating a personalized parameter optimization sub-goal for each control unit according to the parameter optimization target curve;

[0033] S14. Construct a spatial correlation matrix using the spatial position information of the sensor, construct a temporal correlation matrix based on the time series of the data, and construct a spatiotemporal correlation model based on the spatial correlation matrix and the temporal correlation matrix.

[0034] Preferably, the adjustment method of the real-time adjustment module is as follows:

[0035] S21, obtaining sub-goals: the intelligent actuator obtains the parameter optimization sub-goals of the control unit in which it is located, including the laser power sub-goal, the cutting speed sub-goal and the auxiliary gas pressure sub-goal;

[0036] S22, status feedback: the intelligent actuator obtains the status feedback data of the control unit in real time, including laser power feedback, cutting speed feedback and auxiliary gas pressure feedback;

[0037] S23, deviation calculation: calculate the deviation between the laser power feedback and the laser power sub-target, if the deviation exceeds the power threshold, adjust the laser power; calculate the deviation between the cutting speed feedback and the cutting speed sub-target, if the deviation exceeds the speed threshold, adjust the cutting speed; calculate the deviation between the auxiliary gas pressure feedback and the auxiliary gas pressure sub-target, if the deviation exceeds the pressure threshold, adjust the auxiliary gas pressure;

[0038] Specifically, the deviation between the laser power feedback and the laser power sub-target is calculated. If the deviation exceeds the threshold, the intelligent actuator locally adjusts the laser power according to the following rule: q = k p (P t -P f ), where q represents the adjusted laser power; k p is the laser power control gain coefficient; P t is the laser power sub-target; P f is the laser power feedback;

[0039] Calculate the deviation between the cutting speed feedback and the cutting speed sub-target. If the deviation exceeds the threshold, the intelligent actuator adjusts the cutting speed according to the following rule: v = k v (V t -V f ), where v represents the adjusted cutting speed; k v is the cutting speed control gain coefficient; V tis the cutting speed sub-goal; V f It is the cutting speed feedback;

[0040] Calculate the deviation between the auxiliary gas pressure feedback and the auxiliary gas pressure sub-target. If the deviation exceeds the threshold, the intelligent actuator adjusts the auxiliary gas pressure according to the following rule: p = k g (G t -G f ), where p represents the adjusted auxiliary gas pressure; k g is the auxiliary gas pressure control gain coefficient; G t is the auxiliary gas pressure sub-target; G f is the auxiliary gas pressure feedback;

[0041] S24, saturation detection: The intelligent actuator determines whether the parameters of its own output have reached saturation. If saturated, it sends out an alarm signal and requests to reallocate the parameters to optimize the sub-goals.

[0042] Preferably, obtaining real-time status data of the laser cutting machine includes:

[0043] S111, arranging multiple sensors along the laser cutting path, acquiring data such as laser power, cutting speed, auxiliary gas pressure, etc. on the laser cutting path in real time, and generating a real-time laser power distribution curve reflecting the actual laser power distribution state;

[0044] S112. Arrange sensors at key locations of the laser cutting path to obtain cutting quality data in real time and generate a real-time cutting quality distribution curve that reflects the actual cutting quality distribution state.

[0045] Preferably, dividing the laser cutting path into a plurality of multi-level, non-uniform control units comprises:

[0046] Identify the process stages of the laser cutting path and divide the laser cutting path non-uniformly based on the process stages; for the starting section of cutting, use fine granularity to divide the control unit; for the middle section of cutting, use medium granularity to divide the control unit; for the ending section of cutting, use coarse granularity to divide the control unit;

[0047] In each control unit, multiple control layers are divided along different directions according to the characteristics of the laser cutting path;

[0048] Methods for identifying the process stages of a laser cutting path include:

[0049] Based on the data collected by the sensors on the laser cutting path, a spatiotemporal correlation model is constructed, and the process stage of the laser cutting path is identified according to the spatiotemporal correlation model;

[0050] A three-dimensional combination code of process stage number + control unit number + control layer number is used to establish a globally unique control unit identification;

[0051] The construction methods of spatiotemporal correlation models include:

[0052] The collected data is represented as a three-dimensional tensor, where x and y represent the spatial coordinates of the sensor on the laser cutting path, and t represents the timestamp of the sampling.

[0053] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are:

[0054] 1. The present invention calculates short-term status indicators based on the collected data and generates an instant laser cutting index, which provides an instant health assessment of the laser cutting process, helps operators quickly identify and solve instantaneous problems, and ensures the stability and consistency of the cutting process; the present invention calculates long-term status indicators based on the collected data and generates a long-term laser cutting index, which provides a long-term performance assessment of the laser cutting process, helps operators identify and optimize long-standing problems, extends the service life of equipment, and improves overall production efficiency.

[0055] 2. The present invention generates optimized laser cutting parameters based on the immediate index and the long-term index. Through the machine learning algorithm, the intelligent optimization of the parameters is realized, the cutting quality and production efficiency are improved, and the frequency and complexity of manual intervention are reduced. According to the optimized laser cutting parameters, the working parameters of the laser cutting machine are adjusted in real time, and the adjustment results and related data are sent back to the data storage module for updating the historical database, forming a closed-loop control system to ensure that the laser cutting machine always works according to the optimal parameters, thereby improving the stability and reliability of the system.

[0056] 3. The present invention divides the laser cutting path into a plurality of multi-level, non-uniform control units, each of which corresponds to a group of intelligent actuators. Refined control is performed according to the characteristics of the laser cutting path, thereby achieving precise control of different process stages, improving cutting quality and production efficiency, and performing particularly well in complex cutting tasks.

[0057] 4. The present invention constructs a spatiotemporal correlation model through data collected by sensors, identifies the process stages of the laser cutting path, and generates a globally unique control unit identifier, providing accurate modeling and identification of the laser cutting path, helping to optimize the planning and control of the cutting path and improve cutting efficiency and quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0059] Figure 1 A module diagram of a laser cutting parameter adjustment system based on machine learning of the present invention is shown;

[0060] Figure 2 The optimization flow chart of the parameter optimization module of the present invention is shown;

[0061] Figure 3 The adjustment flow chart of the real-time adjustment module of the present invention is shown. DETAILED DESCRIPTION

[0062] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0063] In addition, the described features, structures or characteristics may be combined in one or more example embodiments in any suitable manner. In the following description, many specific details are provided to provide a full understanding of the example embodiments of the present disclosure. However, those skilled in the art will appreciate that the technical solutions of the present disclosure may be practiced while omitting one or more of the specific details, or other methods, components, steps, etc. may be adopted. In other cases, well-known structures, methods, implementations or operations are not shown or described in detail to avoid obscuring various aspects of the present disclosure.

[0064] Embodiment 1:

[0065] See also Figure 1 As shown, the laser cutting parameter adjustment system based on machine learning of this embodiment includes a data acquisition module, a data storage module, a first indicator module, a second indicator module, a parameter optimization module and a real-time adjustment module.

[0066] Data acquisition module: measures various parameters of the laser cutting process in real time and sends the data to the data storage module. Parameters include but are not limited to laser power, cutting speed, auxiliary gas type and pressure, nozzle diameter, focus position, cutting quality (such as edge roughness, cutting accuracy), material type and thickness.

[0067] Data storage module: stores the data collected by the data acquisition module and is configured with a historical database. The storage content includes laser cutting parameters, cutting quality data and historical data (including historical feature vectors, timestamps and cutting quality evaluation).

[0068] The first indicator module: calculates the short-term status indicators of laser cutting based on the cutting parameters, generates the laser cutting instant index, and sends the results to the parameter optimization module.

[0069] The second indicator module: calculates the long-term status indicators of laser cutting based on the cutting parameters, generates a long-term index of laser cutting, and sends the results to the parameter optimization module.

[0070] Parameter optimization module: Generates optimized laser cutting parameters based on the immediate index and long-term index, and sends the optimized parameters to the real-time adjustment module.

[0071] Real-time adjustment module: According to the optimized laser cutting parameters, the working parameters of the laser cutting machine are adjusted in real time, and the adjustment results and related data are sent back to the data storage module for updating the historical database.

[0072] The first indicator module calculates the short-term state indicators of laser cutting based on the collected data and generates the laser cutting instant index. The short-term state indicators include but are not limited to the number of laser power fluctuations, the number of cutting speed fluctuations, the number of auxiliary gas pressure fluctuations, the number of nozzle diameter abnormalities, the number of focus position abnormalities and the number of cutting quality abnormalities.

[0073] The calculation method is as follows:

[0074] The laser power is read, and the number of laser power fluctuations per unit time is counted as the number of laser power fluctuations.

[0075] The cutting speed is read, and the number of cutting speed fluctuations per unit time is counted as the number of cutting speed fluctuations.

[0076] The auxiliary gas pressure is read, and the number of auxiliary gas pressure fluctuations per unit time is counted as the number of auxiliary gas pressure fluctuations.

[0077] The nozzle diameter is read, and the number of nozzle diameter anomalies per unit time is counted as the number of nozzle diameter anomalies.

[0078] The focus position is read, and the number of focus position anomalies per unit time is counted as the number of focus position anomalies.

[0079] The cutting quality is read, and the number of abnormal cutting quality occurrences per unit time is counted as the number of abnormal cutting quality occurrences.

[0080] The instant index calculation formula is as follows:

[0081]

[0082] Where H represents the instantaneous laser cutting index; I i represents the i-th short-term state indicator; b i is the offset coefficient; ω i is the weight coefficient.

[0083] The second indicator module calculates the long-term status indicators of laser cutting based on the collected data and generates a long-term index for laser cutting, wherein the long-term status indicators include but are not limited to average laser power, average cutting speed, average auxiliary gas pressure, average nozzle diameter, average focal position and average cutting quality.

[0084] The calculation method is as follows:

[0085] The long-term status indicators are calculated once every statistical period of T.

[0086] For example, the average laser power is calculated as follows:

[0087]

[0088] Among them, A i represents the average laser power within a statistical period; P(t) represents the laser power within time t.

[0089] The long-term index calculation formula is as follows:

[0090]

[0091] Where L represents the long-term index of laser cutting; L i represents the i-th long-term state indicator; R i is the reference threshold; ω i is the weight coefficient.

[0092] See also Figure 2 As shown, the optimization method of the parameter optimization module is as follows:

[0093] S11, acquiring real-time status data of the laser cutting machine, generating a real-time distribution curve according to the real-time status data, receiving a preset target distribution curve, and calculating a deviation between the real-time distribution curve and the target distribution curve.

[0094] The target distribution curves include a target laser power distribution curve, a target cutting speed distribution curve and a target cutting quality distribution curve.

[0095] The step of obtaining the real-time status data of the laser cutting machine includes:

[0096] S111. Arrange multiple sensors along the laser cutting path to obtain data such as laser power, cutting speed, auxiliary gas pressure, etc. on the laser cutting path in real time, and generate a real-time laser power distribution curve reflecting the actual laser power distribution state.

[0097] S112. Arrange sensors at key locations of the laser cutting path to obtain cutting quality data in real time and generate a real-time cutting quality distribution curve that reflects the actual cutting quality distribution state.

[0098] The calculation of the deviation between the target distribution curve and the real-time distribution curve includes:

[0099] The target laser power distribution curve is compared point by point with the real-time laser power distribution curve, and the laser power deviation of the laser power value at the corresponding position is calculated.

[0100] The target cutting speed distribution curve is compared with the real-time cutting speed distribution curve point by point, and the cutting speed deviation of the cutting speed value at the corresponding position is calculated.

[0101] The target cutting quality distribution curve is compared point by point with the real-time cutting quality distribution curve, and the cutting quality deviation of the cutting quality value at the corresponding position is calculated.

[0102] A parameter optimization target curve is generated according to the deviation between the target distribution curve and the real-time distribution curve. The method for generating the parameter optimization target curve includes:

[0103] S12. Calculate a deviation threshold according to the deviation between the target distribution curve and the real-time distribution curve. When the deviation exceeds the deviation threshold, trigger an over-limit alarm in time and continue to generate a parameter optimization target curve.

[0104] The deviation threshold is calculated based on the deviation between the target distribution curve and the real-time distribution curve, and the calculation method is as follows:

[0105]

[0106] Where D represents the deviation threshold; D i Represents the deviation value of the i-th sampling point; Represents the average value of the deviation; σ D represents the standard deviation of the deviation value; N represents the number of sampling points.

[0107] S13, dividing the laser cutting path into a plurality of multi-level, non-uniform control units, each control unit corresponding to a group of intelligent actuators; generating personalized parameter optimization sub-goals for each control unit according to the parameter optimization target curve.

[0108] The method of dividing the laser cutting path into a plurality of multi-level, non-uniform control units includes:

[0109] The process stages of the laser cutting path are identified, and the laser cutting path is non-uniformly divided based on the process stages.

[0110] For example, for the cutting start segment, the control unit is divided into fine granularity; for the cutting middle segment, the control unit is divided into medium granularity; and for the cutting end segment, the control unit is divided into coarse granularity.

[0111] In each control unit, multiple control layers are divided along different directions according to the characteristics of the laser cutting path.

[0112] The method for identifying a process stage of a laser cutting path comprises:

[0113] According to the data collected by the sensors on the laser cutting path, a spatiotemporal correlation model is constructed, and the process stages of the laser cutting path are identified according to the spatiotemporal correlation model.

[0114] A three-dimensional combined coding of process stage number + control unit number + control layer number is used to establish a globally unique control unit identification.

[0115] The method for constructing the spatiotemporal correlation model includes:

[0116] The collected data is represented as a three-dimensional tensor, where x and y represent the spatial coordinates of the sensor on the laser cutting path, and t represents the timestamp of the sampling.

[0117] S14. Construct a spatial correlation matrix using the spatial position information of the sensor, construct a temporal correlation matrix based on the time series of the data, and construct a spatiotemporal correlation model based on the spatial correlation matrix and the temporal correlation matrix.

[0118] See also Figure 3 As shown, the adjustment method of the real-time adjustment module is as follows:

[0119] S21. Obtain sub-goals: The intelligent actuator obtains the parameter optimization sub-goals of the control unit in which it is located, including the laser power sub-goal, the cutting speed sub-goal and the auxiliary gas pressure sub-goal.

[0120] S22, status feedback: The intelligent actuator obtains the status feedback data of the control unit in real time, including laser power feedback, cutting speed feedback and auxiliary gas pressure feedback.

[0121] S23, deviation calculation: calculate the deviation between the laser power feedback and the laser power sub-target. If the deviation exceeds the power threshold, adjust the laser power; calculate the deviation between the cutting speed feedback and the cutting speed sub-target. If the deviation exceeds the speed threshold, adjust the cutting speed; calculate the deviation between the auxiliary gas pressure feedback and the auxiliary gas pressure sub-target. If the deviation exceeds the pressure threshold, adjust the auxiliary gas pressure.

[0122] Specifically, the deviation between the laser power feedback and the laser power sub-target is calculated. If the deviation exceeds the threshold, the intelligent actuator locally adjusts the laser power. The adjustment law is:

[0123] q=k p (P t -P f );

[0124] Where, q represents the adjusted laser power; k p is the laser power control gain coefficient; P t is the laser power sub-target; P f is the laser power feedback.

[0125] Calculate the deviation between the cutting speed feedback and the cutting speed sub-target. If the deviation exceeds the threshold, the intelligent actuator adjusts the cutting speed. The adjustment law is:

[0126] v = k v (V t -V f );

[0127] Where, v represents the adjusted cutting speed; k v is the cutting speed control gain coefficient; V t is the cutting speed sub-goal; V f It is the cutting speed feedback.

[0128] The deviation between the auxiliary gas pressure feedback and the auxiliary gas pressure sub-target is calculated. If the deviation exceeds the threshold, the intelligent actuator adjusts the auxiliary gas pressure. The adjustment law is:

[0129] p=k g (G t -G f );

[0130] Where, p represents the adjusted auxiliary gas pressure; k g is the auxiliary gas pressure control gain coefficient; G t is the auxiliary gas pressure sub-target; G f is the auxiliary gas pressure feedback.

[0131] S24, saturation detection: The intelligent actuator determines whether the parameters of its own output have reached saturation. If saturated, it sends out an alarm signal and requests to reallocate the parameters to optimize the sub-goals.

[0132] The beneficial effects of this embodiment are as follows: through real-time monitoring and intelligent optimization, the system can timely detect and correct cutting problems to ensure optimal quality and efficiency. Automated parameter adjustment reduces manual intervention and improves reliability. Long-term performance monitoring extends equipment life, and multi-level control unit division improves cutting accuracy and efficiency. The historical database supports data analysis and continuous optimization, and closed-loop control improves system stability and reliability and reduces failures.

[0133] The weight coefficients of the present invention are used to measure the degree of influence of different factors or variables on a certain result or decision. The definition of weight coefficient refers to the value assigned to each factor when comparing and evaluating multiple factors to reflect its importance or priority. These weight coefficients can be determined according to specific circumstances and needs, and are usually jointly formulated and confirmed by professionals or relevant stakeholders. By reasonably setting the weight coefficients, it can help the program or system to make decisions or predictions more accurately.

[0134] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

[0135] The preferred embodiments of the present invention disclosed above are only used to help explain the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to only specific implementation methods. Obviously, many modifications and changes can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present invention, so that those skilled in the art can understand and use the present invention well. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. Laser cutting parameter adjustment system based on machine learning, characterized in that: The system includes a data acquisition module, a data storage module, a first indicator module, a second indicator module, a parameter optimization module and a real-time adjustment module; Data acquisition module: collects parameters during the laser cutting process and sends the data to the data storage module. The parameters include laser power, cutting speed, auxiliary gas type and pressure, nozzle diameter, focus position, cutting quality, material type and thickness; Data storage module: stores the data collected by the data acquisition module and is configured with a historical database. The storage content includes laser cutting parameters, cutting quality data and historical data. The historical data includes historical feature vectors, timestamps and cutting quality evaluations. The first indicator module: calculates the short-term state indicators of laser cutting based on cutting parameters, generates the instantaneous index of laser cutting, and sends the results to the parameter optimization module; The second indicator module: calculates the long-term state indicators of laser cutting based on the cutting parameters, generates the laser cutting long-term index, and sends the results to the parameter optimization module; Parameter optimization module: Generates optimized laser cutting parameters based on the immediate index and long-term index, and sends the optimized parameters to the real-time adjustment module; Real-time adjustment module: According to the optimized laser cutting parameters, the working parameters of the laser cutting machine are adjusted in real time, and the adjustment results and related data are sent back to the data storage module for updating the historical database.

2. The laser cutting parameter adjustment system based on machine learning according to claim 1, characterized in that: The short-term status indicators include the number of laser power fluctuations, the number of cutting speed fluctuations, the number of auxiliary gas pressure fluctuations, the number of nozzle diameter abnormalities, the number of focus position abnormalities and the number of cutting quality abnormalities.

3. The laser cutting parameter adjustment system based on machine learning according to claim 1, characterized in that: The specific details of generating the instant laser cutting index are as follows: Read the laser power, count the number of laser power fluctuations each time per unit time, as the number of laser power fluctuations; read the cutting speed, count the number of cutting speed fluctuations each time per unit time, as the number of cutting speed fluctuations; read the auxiliary gas pressure, count the number of auxiliary gas pressure fluctuations each time per unit time, as the number of auxiliary gas pressure fluctuations; read the nozzle diameter, count the number of nozzle diameter abnormalities each time per unit time, as the number of nozzle diameter abnormalities; Read the focus position, count the number of focus position abnormalities each time within a unit time, and use it as the number of focus position abnormalities; read the cutting quality, count the number of cutting quality abnormalities each time within a unit time, and use it as the number of cutting quality abnormalities; The instant index calculation formula is as follows: Where H represents the instantaneous laser cutting index; I i represents the i-th short-term state indicator; b i is the offset coefficient; ω i is the weight coefficient.

4. The laser cutting parameter adjustment system based on machine learning according to claim 1, characterized in that: The long-term status indicators include average laser power, average cutting speed, average auxiliary gas pressure, average nozzle diameter, average focus position and average cutting quality.

5. The laser cutting parameter adjustment system based on machine learning according to claim 1, characterized in that: The specific details of generating the laser cutting long-term index are as follows: The long-term state index is calculated once every statistical period of time T. The method for calculating the average laser power is as follows: Among them, A i represents the average laser power within a statistical period T; P(t) represents the laser power within time t; The long-term index calculation formula is as follows: Where L represents the long-term index of laser cutting; L i represents the i-th long-term state indicator; R i is the reference threshold; ω i is the weight coefficient.

6. The laser cutting parameter adjustment system based on machine learning according to claim 1, characterized in that: The optimization method of the parameter optimization module is as follows: S11, acquiring real-time status data of the laser cutting machine, generating a real-time distribution curve according to the real-time status data, receiving a preset target distribution curve, and calculating a deviation between the real-time distribution curve and the target distribution curve; the target distribution curve includes a target laser power distribution curve, a target cutting speed distribution curve, and a target cutting quality distribution curve; Calculating the deviation between the target distribution curve and the real-time distribution curve includes: The target laser power distribution curve is compared with the real-time laser power distribution curve point by point, and the laser power deviation of the laser power value at the corresponding position is calculated; the target cutting speed distribution curve is compared with the real-time cutting speed distribution curve point by point, and the cutting speed deviation of the cutting speed value at the corresponding position is calculated; the target cutting quality distribution curve is compared with the real-time cutting quality distribution curve point by point, and the cutting quality deviation of the cutting quality value at the corresponding position is calculated; S12, calculating a deviation threshold according to the deviation between the target distribution curve and the real-time distribution curve, and when the deviation exceeds the deviation threshold, triggering an over-limit alarm in time, and continuing to generate a parameter optimization target curve; The deviation threshold is calculated based on the deviation between the target distribution curve and the real-time distribution curve, and the calculation method is as follows: Where D represents the deviation threshold; D i Represents the deviation value of the i-th sampling point; Represents the average value of the deviation; σ D represents the standard deviation of the deviation value; N represents the number of sampling points; S13, dividing the laser cutting path into a plurality of multi-level, non-uniform control units, each control unit corresponding to a group of intelligent actuators; generating a personalized parameter optimization sub-goal for each control unit according to the parameter optimization target curve; S14. Construct a spatial correlation matrix using the spatial position information of the sensor, construct a temporal correlation matrix based on the time series of the data, and construct a spatiotemporal correlation model based on the spatial correlation matrix and the temporal correlation matrix.

7. The laser cutting parameter adjustment system based on machine learning according to claim 1, characterized in that: The adjustment method of the real-time adjustment module is as follows: S21, obtaining sub-goals: the intelligent actuator obtains the parameter optimization sub-goals of the control unit in which it is located, including the laser power sub-goal, the cutting speed sub-goal and the auxiliary gas pressure sub-goal; S22, status feedback: the intelligent actuator obtains the status feedback data of the control unit in real time, including laser power feedback, cutting speed feedback and auxiliary gas pressure feedback; S23, deviation calculation: calculate the deviation between the laser power feedback and the laser power sub-target, if the deviation exceeds the power threshold, adjust the laser power; calculate the deviation between the cutting speed feedback and the cutting speed sub-target, if the deviation exceeds the speed threshold, adjust the cutting speed; calculate the deviation between the auxiliary gas pressure feedback and the auxiliary gas pressure sub-target, if the deviation exceeds the pressure threshold, adjust the auxiliary gas pressure; Specifically, the deviation between the laser power feedback and the laser power sub-target is calculated. If the deviation exceeds the threshold, the intelligent actuator locally adjusts the laser power according to the following rule: q = k p (P t -P f ), where q represents the adjusted laser power; k p is the laser power control gain coefficient; P t is the laser power sub-target; P f is the laser power feedback; Calculate the deviation between the cutting speed feedback and the cutting speed sub-target. If the deviation exceeds the threshold, the intelligent actuator adjusts the cutting speed according to the following rule: v = k v (V t -V f ), where v represents the adjusted cutting speed; k v is the cutting speed control gain coefficient; V t is the cutting speed sub-goal; V f It is the cutting speed feedback; Calculate the deviation between the auxiliary gas pressure feedback and the auxiliary gas pressure sub-target. If the deviation exceeds the threshold, the intelligent actuator adjusts the auxiliary gas pressure according to the following rule: p = k g (G t -G f ), where p represents the adjusted auxiliary gas pressure; k g is the auxiliary gas pressure control gain coefficient; G t is the auxiliary gas pressure sub-target; G f is the auxiliary gas pressure feedback; S24, saturation detection: The intelligent actuator determines whether the parameters of its own output have reached saturation. If saturated, it sends out an alarm signal and requests to reallocate the parameters to optimize the sub-goals.

8. The laser cutting parameter adjustment system based on machine learning according to claim 6, characterized in that: The step of obtaining the real-time status data of the laser cutting machine includes: S111, arranging multiple sensors along the laser cutting path, acquiring data such as laser power, cutting speed, auxiliary gas pressure, etc. on the laser cutting path in real time, and generating a real-time laser power distribution curve reflecting the actual laser power distribution state; S112. Arrange sensors at key locations of the laser cutting path to obtain cutting quality data in real time and generate a real-time cutting quality distribution curve that reflects the actual cutting quality distribution state.

9. The laser cutting parameter adjustment system based on machine learning according to claim 6, characterized in that: The method of dividing the laser cutting path into a plurality of multi-level, non-uniform control units includes: Identify the process stages of the laser cutting path and divide the laser cutting path non-uniformly based on the process stages; for the starting section of cutting, use fine granularity to divide the control unit; for the middle section of cutting, use medium granularity to divide the control unit; for the ending section of cutting, use coarse granularity to divide the control unit; In each control unit, multiple control layers are divided along different directions according to the characteristics of the laser cutting path; The method for identifying a process stage of a laser cutting path comprises: Based on the data collected by the sensors on the laser cutting path, a spatiotemporal correlation model is constructed, and the process stage of the laser cutting path is identified according to the spatiotemporal correlation model; Use the three-dimensional combination code of process stage number + control unit number + control layer number to establish a globally unique control unit identification; The method for constructing the spatiotemporal correlation model includes: The collected data is represented as a three-dimensional tensor, where x and y represent the spatial coordinates of the sensor on the laser cutting path, and t represents the timestamp of the sampling.

Images