A laser galvanometer intelligent control system and method based on big data

Abstract

This invention provides a big data-based intelligent control system and method for a laser galvanometer, belonging to the field of intelligent control systems. The method comprises: S1. Data acquisition; S2. Data preprocessing; S3. Feature extraction; S4. Model training; S5. Real-time prediction; S6. Control instruction generation; S7. Control instruction execution; and S8. Feedback and optimization. This invention enables high-precision closed-loop control of the laser galvanometer's motion, effectively reducing the deviation between the actual motion state and the target motion state, improving laser processing accuracy, and meeting the requirements of applications with extremely high precision requirements, such as fine processing of electronic chips.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent control systems, and in particular to a laser galvanometer intelligent control system and method based on big data. Background Art

[0002] In modern industrial production, laser processing technology has been widely used in many fields, such as 3D printing, electronic manufacturing, and automotive parts processing, due to its advantages of high precision, high efficiency, and non-contact. As a key executive component in the laser processing system, the motion control accuracy and speed of the laser galvanometer directly affect the quality and efficiency of laser processing. Traditional laser galvanometer control methods are mostly based on empirical models or simple control algorithms, which are difficult to cope with complex and changing working environments and the growing demand for high-precision processing. Under complex working conditions, laser galvanometers are easily affected by environmental factors such as temperature and humidity, and as the complexity of processing tasks increases, traditional methods are unable to adjust the motion state of the galvanometer in real time to ensure processing accuracy.

[0003] The rapid development of emerging technologies such as big data and artificial intelligence has provided new opportunities for the innovation of laser galvanometer control technology. By introducing big data technology, it is possible to collect, analyze, and process the massive amounts of data generated during the operation of the laser galvanometer, and to explore the patterns hidden in the data. This allows for accurate prediction and intelligent control of the laser galvanometer's motion state, thus addressing the shortcomings of traditional control methods and meeting the industrial production requirements for higher quality and higher efficiency in laser processing. Summary of the Invention

[0004] The present invention provides a laser galvanometer intelligent control system and method based on big data to achieve high-precision intelligent control.

[0005] To achieve the above object, the present invention adopts the following technical solutions:

[0006] A laser galvanometer intelligent control method based on big data, comprising:

[0007] S1. Data acquisition: Collect real-time position data, motion speed data, acceleration data, and environmental data of the laser galvanometer, package and timestamp the data, and obtain raw position data, raw velocity data, raw acceleration data, and raw environmental data;

[0008] S2. Data preprocessing: Abnormal data removal and normalization are performed on the raw position data, raw velocity data, raw acceleration data, and raw environmental data to obtain valid position data, valid velocity data, valid acceleration data, and valid environmental data, which are then integrated to obtain the preprocessed data set.

[0009] S3. Feature extraction: Extract the laser galvanometer position change trend characteristics, velocity change rate characteristics, acceleration change trend characteristics, and related characteristics of the impact of environmental factors on the laser galvanometer performance from the preprocessed data set to form a feature vector set; process the feature vector set using the single pendulum wave technique to obtain an enhanced feature vector set;

[0010] S4. Model training: Divide the enhanced feature vector set into a training set and a test set; use the training set to train the long short-term memory network model and the random forest model respectively; perform a weighted fusion of the trained long short-term memory network model and the random forest model to obtain a laser galvanometer motion prediction model;

[0011] S5. Real-time prediction; inputting the real-time enhanced feature vector set into the laser galvanometer motion prediction model to obtain the predicted laser galvanometer position coordinates, velocity, and acceleration;

[0012] S6. Control instruction generation: Based on the predicted laser galvanometer position coordinates, velocity, and acceleration and the preset target motion state, the position deviation, velocity deviation, and acceleration deviation between the two are calculated; using the proportional-integral-differential control algorithm, the control instructions for position, velocity, and acceleration are calculated based on these deviations to form a control instruction set;

[0013] S7. Control instruction execution: transmitting the control instruction set to the laser galvanometer control system, adjusting the laser galvanometer motor drive parameters, and achieving real-time control of the laser galvanometer motion state;

[0014] S8. Feedback and optimization; after the laser galvanometer executes the control instruction, its actual motion state data is collected again, and the above steps S1-S7 are repeated; the newly collected data is compared and analyzed with the predicted data and the control target data. If the deviation between the actual motion state and the target motion state exceeds the preset error range, the parameters in the proportional-integral-differential control algorithm and the parameters of the laser galvanometer motion prediction model are readjusted to optimize the model.

[0015] In this specification, in S1, a fixed sampling frequency is set to periodically collect the position data, velocity data, acceleration data and environmental data of the laser galvanometer.

[0016] In this specification, in S2, the basis for judging abnormal data is: the position coordinates, speed, acceleration or environmental data of the laser galvanometer exceeds the preset reasonable range.

[0017] In this specification, in S3, the position change trend feature is obtained by calculating the difference between the position coordinates of adjacent time points; the speed change rate feature is obtained by calculating the ratio of the difference between the speeds of adjacent time points to the time interval; and the acceleration change trend feature is obtained by calculating the difference between the accelerations of adjacent time points.

[0018] In this specification, in S3, the single pendulum wave technology simulates the periodic motion law of a simple pendulum to perform fluctuation analysis and enhancement on the data in the feature vector set, thereby mining the potential periodic characteristics of the data.

[0019] In this specification, in S4, the mean square error is used as the loss function during the training of the long short-term memory network model, and the model parameters are adjusted through the back propagation algorithm.

[0020] In this specification, in S4, the random forest model is composed of multiple decision trees. When constructing each decision tree, training samples are selected by random sampling with replacement, and some features are randomly selected for node splitting.

[0021] In this specification, in S4, when the long short-term memory network model and the random forest model are weightedly fused, the weighting coefficient is determined by adjusting the parameters on the validation set.

[0022] In this specification, in S8, a particle swarm optimization algorithm or a genetic algorithm is used to adjust the parameters in the proportional-integral-differential control algorithm and the parameters of the laser galvanometer motion prediction model.

[0023] A laser galvanometer intelligent control system based on big data, using any of the above-mentioned laser galvanometer intelligent control methods based on big data, the laser galvanometer intelligent control system based on big data comprising:

[0024] The data acquisition module is used to collect the real-time position data, motion speed data, acceleration data, and environmental data of the laser galvanometer, and perform data packaging and time stamping processing respectively to obtain the original position data, original speed data, original acceleration data, and original environmental data;

[0025] The data preprocessing module is used to delete abnormal data and normalize the original position data, original velocity data, original acceleration data and original environment data to obtain valid position data, valid velocity data, valid acceleration data and valid environment data, and integrate them to obtain the preprocessed data set;

[0026] The feature extraction module is used to extract the laser galvanometer position change trend characteristics, velocity change rate characteristics, acceleration change trend characteristics, and related characteristics of the impact of environmental factors on the laser galvanometer performance from the preprocessed data set to form a feature vector set; the feature vector set is processed using the single pendulum wave technology to obtain an enhanced feature vector set;

[0027] The model training module is used to divide the enhanced feature vector set into a training set and a test set; use the training set to train the long short-term memory network model and the random forest model respectively; and perform a weighted fusion of the trained long short-term memory network model and the random forest model to obtain a laser galvanometer motion prediction model;

[0028] A real-time prediction module is used to input the real-time enhanced feature vector set into the laser galvanometer motion prediction model to obtain the predicted laser galvanometer position coordinates, velocity and acceleration;

[0029] The control instruction generation module is used to calculate the position deviation, velocity deviation, and acceleration deviation between the predicted laser galvanometer position coordinates, velocity, and acceleration and the preset target motion state; adopts the proportional-integral-differential control algorithm to calculate the control instructions for position, velocity, and acceleration based on the above deviations to form a control instruction set;

[0030] The control instruction execution module is used to transmit the control instruction set to the control system of the laser galvanometer, adjust the motor drive parameters of the laser galvanometer, and realize real-time control of the motion state of the laser galvanometer;

[0031] The feedback and optimization module is used to collect the actual motion state data of the laser galvanometer again after it executes the control instruction, and compare and analyze the newly collected data with the predicted data and the control target data. If the deviation between the actual motion state and the target motion state exceeds the preset error range, the parameters in the proportional-integral-differential control algorithm and the parameters of the laser galvanometer motion prediction model are readjusted to optimize the model.

[0032] In summary, the present invention has at least the following beneficial effects:

[0033] High-Precision Control: By collecting, preprocessing, and deeply analyzing multi-source data, including the laser galvanometer's position, velocity, acceleration, and environmental data, combined with advanced feature extraction methods and integrated learning models, the system can accurately predict the laser galvanometer's future motion state. By comparing it with the preset target motion state and generating control instructions using a PID control algorithm, high-precision closed-loop control of the laser galvanometer's motion is achieved, effectively reducing the deviation between the actual motion state and the target motion state, improving laser processing accuracy, and meeting extremely high-precision application scenarios such as the fine processing of electronic chips.

[0034] Efficient Real-Time Response: The entire control process, from data acquisition to control command execution, utilizes highly efficient algorithms and process designs, enabling rapid processing of real-time data and timely adjustments to the motion of the laser galvanometer. The system can rapidly respond to dynamic changes in processing tasks. For example, in 3D printing, where model structures are complex and ever-changing, this intelligent control method can rapidly adjust the galvanometer motion, ensuring the accuracy of the laser scanning path and significantly improving processing efficiency.

[0035] Improved Stability: The integrated LSTM model and random forest model enhance the model's learning and generalization capabilities for complex data, making prediction results more stable and reliable. Furthermore, by regularly optimizing model parameters and PID control parameters, as well as evaluating and optimizing the performance of the decision tree within the random forest, we can effectively address the impact of factors such as changes in the working environment and equipment aging on the stability of the laser galvanometer motion control, ensuring the continued stability of the laser processing process and reducing the rate of defective products caused by equipment instability.

[0036] Enhanced Adaptability: The system features a feedback and optimization mechanism that automatically adjusts control parameters based on the deviation between the laser galvanometer's actual motion and the target state, providing excellent adaptability to diverse working environments and processing tasks. Whether operating in harsh environments such as high temperature and high humidity, or processing objects of varying materials and shapes, the system ensures optimal motion control of the laser galvanometer through continuous optimization of control strategies, broadening the application scope of laser processing technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. 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 any creative work.

[0038] Figure 1 This is a flow chart of the big data-based laser galvanometer intelligent control method involved in the present invention.

[0039] Figure 2 Schematic diagram of the feature extraction and enhancement process involved in the present invention.

[0040] Figure 3 Schematic diagram of the model training and prediction process involved in the present invention.

[0041] Figure 4 Schematic diagram of the big data-based laser galvanometer intelligent control system involved in the present invention. DETAILED DESCRIPTION

[0042] Hereinafter, only certain exemplary embodiments are briefly described. As will be appreciated by those skilled in the art, the described embodiments may be modified in various ways without departing from the spirit or scope of the embodiments of the present invention. Therefore, the drawings and description are to be regarded as illustrative in nature and not restrictive.

[0043] The disclosure below provides many different embodiments or examples for implementing different structures of the embodiments of the present invention. In order to simplify the disclosure of the embodiments of the present invention, the components and configurations of specific examples are described below. Of course, these are merely examples and are not intended to limit the embodiments of the present invention. In addition, the embodiments of the present invention may repeat reference numerals and / or reference letters in different examples. Such repetition is for the purpose of simplicity and clarity and does not in itself indicate the relationship between the various embodiments and / or configurations discussed.

[0044] The embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0045] like Figure 1 As shown, this embodiment provides a laser galvanometer intelligent control method based on big data, including:

[0046] S1. Data acquisition: Collect real-time position data, motion speed data, acceleration data, and environmental data of the laser galvanometer, package and timestamp the data, and obtain raw position data, raw velocity data, raw acceleration data, and raw environmental data;

[0047] S2. Data preprocessing: Abnormal data removal and normalization are performed on the raw position data, raw velocity data, raw acceleration data, and raw environmental data to obtain valid position data, valid velocity data, valid acceleration data, and valid environmental data, which are then integrated to obtain the preprocessed data set.

[0048] S3. Feature extraction: Extract the laser galvanometer position change trend characteristics, velocity change rate characteristics, acceleration change trend characteristics, and related characteristics of the impact of environmental factors on the laser galvanometer performance from the preprocessed data set to form a feature vector set; process the feature vector set using the single pendulum wave technique to obtain an enhanced feature vector set;

[0049] S4. Model training: Divide the enhanced feature vector set into a training set and a test set; use the training set to train the long short-term memory network model and the random forest model respectively; perform a weighted fusion of the trained long short-term memory network model and the random forest model to obtain a laser galvanometer motion prediction model;

[0050] S5. Real-time prediction; inputting the real-time enhanced feature vector set into the laser galvanometer motion prediction model to obtain the predicted laser galvanometer position coordinates, velocity, and acceleration;

[0051] S6. Control instruction generation: Based on the predicted laser galvanometer position coordinates, velocity, and acceleration and the preset target motion state, the position deviation, velocity deviation, and acceleration deviation between the two are calculated; using the proportional-integral-differential control algorithm, the control instructions for position, velocity, and acceleration are calculated based on these deviations to form a control instruction set;

[0052] S7. Control instruction execution: transmitting the control instruction set to the laser galvanometer control system, adjusting the laser galvanometer motor drive parameters, and achieving real-time control of the laser galvanometer motion state;

[0053] S8. Feedback and optimization; after the laser galvanometer executes the control instruction, its actual motion state data is collected again, and the above steps S1-S7 are repeated; the newly collected data is compared and analyzed with the predicted data and the control target data. If the deviation between the actual motion state and the target motion state exceeds the preset error range, the parameters in the proportional-integral-differential control algorithm and the parameters of the laser galvanometer motion prediction model are readjusted to optimize the model.

[0054] In some embodiments, in S1, a fixed sampling frequency is set to periodically collect position data, velocity data, acceleration data, and environmental data of the laser galvanometer.

[0055] In some embodiments, in S2, the abnormal data is determined based on whether the position coordinates, velocity, acceleration or environmental data of the laser galvanometer exceed a preset reasonable range.

[0056] In some embodiments, in S3, the position change trend feature is obtained by calculating the difference between the position coordinates of adjacent time points; the speed change rate feature is obtained by calculating the ratio of the difference between the speeds of adjacent time points to the time interval; and the acceleration change trend feature is obtained by calculating the difference between the accelerations of adjacent time points.

[0057] In some embodiments, in S3, the single pendulum wave technology simulates the periodic motion of a simple pendulum to perform fluctuation analysis and enhancement on the data in the feature vector set, thereby mining the potential periodic characteristics of the data.

[0058] In some embodiments, in S4, mean square error is used as a loss function during the training of the long short-term memory network model, and the model parameters are adjusted through a back propagation algorithm.

[0059] In some embodiments, in S4, the random forest model is composed of multiple decision trees, and each decision tree is constructed by using random sampling with replacement to select training samples and randomly selecting some features for node splitting.

[0060] In some embodiments, in S4, when the long short-term memory network model and the random forest model are weightedly fused, the weighting coefficient is determined by adjusting the parameters on the validation set.

[0061] In some embodiments, in S8, a particle swarm optimization algorithm or a genetic algorithm is used to adjust the parameters in the proportional-integral-differential control algorithm and the parameters of the laser galvanometer motion prediction model.

[0062] In some embodiments, data collection:

[0063] In the working environment of the laser galvanometer, a variety of sensors are used for data collection. Among them, the position sensor is used to collect the real-time position data of the laser galvanometer, the speed sensor collects the movement speed data of the laser galvanometer, the acceleration sensor collects the acceleration data of the laser galvanometer, and the environmental sensor collects the temperature, humidity and other environmental data in the working environment. Set a fixed sampling frequency (Unit: Hertz, Hz), collect the above data periodically. Mark each set of collected data with the collection time (Unit: seconds, s), composing the original data set: ;

[0064] in Indicates the The three-dimensional position coordinates of the laser galvanometer during the acquisition, Indicates the The movement speed of the laser galvanometer during the acquisition, Indicates the The acceleration of the laser galvanometer during the acquisition, Indicates the Environmental data at the time of collection, Indicates the The time of collection, Indicates the total number of data collections.

[0065] The process of data collection is as follows: according to the set sampling frequency , the position sensor, velocity sensor, acceleration sensor and environmental sensor periodically acquire the corresponding data in chronological order, package it in a fixed data format, and add the acquisition timestamp After that, it is stored in the temporary data storage area to form the original data set .

[0066] In some embodiments, data preprocessing:

[0067] The original data set The data is transmitted to the data preprocessing module. First, the data is cleaned to remove abnormal data. The basis for judging abnormal data is: if the position coordinates of the laser galvanometer are The value of any dimension exceeds the preset reasonable range ,speed Out of range , acceleration Out of range , or environmental data If the data exceeds the normal environmental parameter range (such as temperature or humidity exceeds the preset reasonable range), the data set is determined to be abnormal data and deleted. Then, the cleaned data is normalized. For the position coordinate data, the formula is used: ;

[0068] 、 、 The data is the normalized three-dimensional coordinates.

[0069] For speed data and acceleration data , using the formula: ; ;

[0070] is the normalized speed data, is the normalized acceleration data.

[0071] For environmental data , according to its specific parameter type, the corresponding normalization formula is used to process it and obtain the normalized data set .

[0072] The process of data preprocessing is: first, the original data set The data is input into the data cleaning algorithm module, which determines and deletes abnormal data according to the preset reasonable range to obtain the data after preliminary processing. The data after preliminary processing is then input into the normalization algorithm module, which calculates the normalization formula according to different data types to obtain the normalized data set. .

[0073] In some embodiments, as Figure 2 As shown, feature extraction:

[0074] From the normalized dataset For location data, extract the location change trend features by calculating the difference between the location coordinates of adjacent time points, such as , , ; For speed data, extract the speed change rate feature, and the calculation method is ; For acceleration data, extract acceleration change trend characteristics, such as At the same time, combined with environmental data, extract relevant features of the impact of environmental factors on the performance of the laser galvanometer, such as the real-time value of the ambient temperature, the temperature change rate (the amount of temperature change per unit time), the temperature change trend (such as whether the temperature rises, falls, or remains stable), the temperature fluctuation amplitude (the difference between the maximum and minimum temperature values ​​over a period of time), and other features, such as the real-time value of the ambient humidity, the humidity change rate, the influence coefficient of humidity on electronic component parameters (such as resistance and capacitance values) (predetermine the changes in electronic component parameters under different humidity levels through experiments, and calculate the influence coefficient), and the predicted value of the degree of corrosion of mechanical parts caused by humidity (based on the empirical model of humidity and corrosion degree or the model trained with historical data for prediction). All extracted features are combined into a feature vector set .

[0075] The process of feature extraction is as follows: for the normalized data set , calculate the position, velocity, acceleration and environmental data according to the above feature extraction formula, integrate the calculated features to form a feature vector set At the same time, the single pendulum wave technology is introduced to the characteristic vector set For further processing, the pendulum wave technology simulates the periodic motion law of the pendulum, performs fluctuation analysis and enhancement on the data in the feature vector, mines the potential periodic characteristics of the data, and obtains the enhanced feature vector set. The specific process is: let the pendulum wave function be ,in is the amplitude, is the angular frequency, is the initial phase, is time. The data in the feature vector set is fused with the pendulum wave function. For example, for the feature ,calculate , thus obtaining the enhanced feature vector set .

[0076] In some embodiments, the core of the application of the pendulum wave technology in laser galvanometer control is to modulate and enhance the characteristic vector by simulating the periodic law of the simple pendulum motion. The following is the specific implementation process of the pendulum wave technology in step S3:

[0077] Mathematical principles of simple pendulum wave technology:

[0078] The pendulum wave function is based on the classical pendulum motion equation, and its standard form is: ;

[0079] in, is the amplitude, which indicates the intensity of the fluctuation (which can be dynamically adjusted according to the characteristics of the data); is the angular frequency, controlling the cycle length ( , for the cycle); For the initial phase, adjust the starting position of the waveform; is the offset used for fluctuations at non-zero centers (default =0).

[0080] Parameter adaptive determination method:

[0081] The parameters of the simple pendulum wave function need to be adaptively determined according to the motion characteristics of the laser galvanometer:

[0082] amplitude : ;

[0083] in Characterized by The standard deviation of is the scaling factor (the empirical value is 0.5-1.5).

[0084] cycle : ;

[0085] in is the average peak-to-peak time interval of the characteristic fluctuations, is the adjustment coefficient (the empirical value is 0.8-1.2).

[0086] Initial Phase : ;

[0087] in is the covariance, is the variance, which is used to align the changing trend of the data.

[0088] Specific implementation steps of feature enhancement:

[0089] Time-frequency analysis: the original eigenvector Perform short-time Fourier transform (STFT) or wavelet transform to extract the main frequency components. Identify the dominant period of the laser galvanometer motion. , as the initial value of the simple pendulum wave period.

[0090] Parameter initialization: Set the initial parameters according to the time-frequency analysis results . Use different parameter sets for environmental characteristics (such as temperature and humidity) and decouple them from motion characteristics.

[0091] Characteristic modulation: ;

[0092] in is the modulation intensity coefficient (range 0.1-0.3), Characterized by timestamp.

[0093] Adaptive adjustment: every interval Sampling points (such as =100), recalculate , , . Use a sliding window (such as window size =50) to update the parameters to ensure real-time performance.

[0094] Implementation example:

[0095] Assume that the current processing position feature , whose time-frequency analysis shows that the dominant period is 100ms (i.e. = 0.1), we can calculate:

[0096] ;

[0097] ;

[0098] ;

[0099] The wave function of the simple pendulum is: ;

[0100] The enhanced features are: .

[0101] Technical advantages:

[0102] Enhanced periodic features: highlighting the periodic pattern of the laser galvanometer motion and improving the LSTM model's ability to capture timing patterns;

[0103] Noise immunity: The pendulum wave function acts as a low-pass filter to suppress high-frequency noise and retain the main motion characteristics;

[0104] Adaptability: The dynamic parameter adjustment mechanism enables the algorithm to adapt to the galvanometer motion characteristics in different working scenarios.

[0105] For environmental characteristics (such as temperature), a fixed parameter simple pendulum wave function can be used to reduce the amount of calculation.

[0106] In this way, the single pendulum wave technology can effectively enhance the periodic characteristics in the laser galvanometer motion data, and improve the prediction accuracy and control stability of the subsequent model.

[0107] It should be noted that the formula It focuses on directly multiplying the pendulum wave function with the original feature, and mining the potential periodic features of the data by simulating the periodic motion of the pendulum; On this basis, we have improved and added Item, can more flexibly control the modulation intensity, avoid the characteristics of the distortion due to over-modulation, and at the same time through the coefficient Adaptive adjustment is achieved to better enhance data features under different working conditions. From the perspective of data processing flow, the original formula serves as the basic operation to provide preliminary enhanced features for subsequent processing; the improved formula is optimized for problems that may arise in actual applications, making the feature enhancement effect more stable and reliable. For example, when the operating conditions of the laser galvanometer are complex and changeable, the improved formula can be adjusted according to different working conditions. value, better enhance the features and provide better quality data for subsequent model training.

[0108] In some embodiments, as Figure 3 As shown, the model training:

[0109] Build an ensemble model that combines a Long Short-Term Memory (LSTM) model with a Random Forest model. A Random Forest is an ensemble learning model composed of multiple decision trees. It improves model stability and prediction accuracy by voting or averaging the predictions of multiple decision trees.

[0110] The enhanced feature vector set Divide into training set and test set , the division ratio is 7:3. Using the training set The LSTM model and random forest model are trained separately.

[0111] For the LSTM model, the mean squared error (MSE) is used as the loss function during training, and its calculation formula is: ;

[0112] in, is the number of training samples, For the The true output value of the sample, For the The predicted output value of the sample in the LSTM model. The parameters of the LSTM model, such as the weight matrix, are continuously adjusted through the back propagation algorithm. and the bias vector , to minimize the loss function.

[0113] For the random forest model, during the training process, each decision tree is randomly sampled from the training set with replacement (bootstrap sampling) for training. At the same time, when splitting each node, some features are randomly selected to select the optimal splitting feature. The decision tree construction process is based on Gini impurity, which is calculated as follows: ;

[0114] in, Represents a node, represents the number of categories, Representation node Belongs to category By continuously splitting nodes, the Gini impurity is gradually reduced until the stopping condition is met (such as the number of node samples is less than the threshold, the decrease in Gini impurity is less than the threshold, etc.), completing the construction of a single decision tree. Repeat the above process to build multiple decision trees to form a random forest model.

[0115] When the loss value of the LSTM model on the training set converges to a smaller value, and the prediction error on the test set meets the preset accuracy requirements, and the prediction error of the random forest model on the test set also meets the requirements, the training is stopped. The LSTM model and the random forest model are integrated to obtain the final laser galvanometer motion prediction model. The fusion method is to perform weighted average of the prediction results of the LSTM model and the random forest model. Suppose the prediction result of the LSTM model is , the prediction result of the random forest model is , the prediction result after fusion is ,but ,in 1 is the weighting coefficient, and the optimal value is determined by adjusting the parameters on the validation set.

[0116] The process of model training is as follows: first, the enhanced feature vector set The training set is divided into two parts according to the ratio of 7:3. and test set . Then the training set Input into the LSTM model and random forest model, the LSTM model performs forward propagation calculation based on the input data to predict the output , and calculate the loss value according to the mean square error formula, and update the parameters through the back propagation algorithm and gradient descent method; the random forest model is trained according to the above method of building a decision tree. When both models meet the corresponding training requirements, their prediction results are weighted and fused on the validation set, and the weight coefficients are adjusted. 1. Obtain the final model with the smallest prediction error on the validation set .

[0117] In some embodiments, real-time predictions:

[0118] Input the real-time data after feature extraction into the trained laser galvanometer motion prediction model In the model, the future motion state of the laser galvanometer is predicted based on the rules learned from historical data, and the predicted position coordinates are output. ,speed and acceleration .

[0119] The process of real-time prediction is as follows: the latest feature vector extracted in step S3 is input into the trained model In the model Through forward propagation calculation, the predicted future motion state data of the laser galvanometer is output .

[0120] In some embodiments, the control instructions generate:

[0121] According to the predicted laser galvanometer motion state With the preset target motion state , calculate the deviation between the two. The position deviation calculation formula is: ;

[0122] The speed deviation calculation formula is: ;

[0123] The acceleration deviation calculation formula is: ;

[0124] The Proportional-Integral-Derivative (PID) control algorithm is used to calculate the control command based on the above deviation. The output formula of the PID control algorithm is: ;

[0125] in, is the control instruction at time t4, is the proportionality coefficient, is the integration coefficient, is the differential coefficient, is the deviation value at time t4, From the initial time 0 to the current time t4, the deviation The integral over time, Deviation The time derivative is the rate of change of the deviation at time t4. In this step, Corresponding respectively , and obtain the control instructions for position, velocity, and acceleration by calculating them separately , which constitutes the control instruction set .

[0126] The process of generating control instructions is as follows: first calculate the various deviations between the predicted motion state and the target motion state, then substitute the deviation values ​​into the PID control algorithm formula, calculate the control instructions for different motion parameters respectively, and finally integrate these control instructions into a control instruction set. .

[0127] In some embodiments, the control instructions execute:

[0128] Control Instruction Set The data is transmitted to the control system of the laser galvanometer. The control system adjusts the motor drive parameters of the laser galvanometer, such as the voltage and current of the motor, according to the received control instructions, thereby realizing real-time control of the motion state of the laser galvanometer and making the laser galvanometer operate according to the preset target motion state.

[0129] The process of executing control instructions is as follows: the control system receives the control instruction set Then, the control instructions 、 、 、 、 They are converted into corresponding motor drive parameter adjustment signals, and the motor parameters such as voltage and current are adjusted through the drive circuit to control the movement of the laser galvanometer.

[0130] In some embodiments, feedback and optimization:

[0131] After executing the control command, the laser galvanometer collects its actual motion state data again through the sensor, repeats steps S1-S7, and compares and analyzes the newly collected data with the predicted data and the control target data. If the deviation between the actual motion state and the target motion state exceeds the preset error range, the proportional coefficient in the PID control algorithm is readjusted. , integral coefficient and differential coefficients , as well as the parameters of the deep learning model, the model is optimized to improve control precision and prediction accuracy.

[0132] The feedback and optimization process involves collecting actual motion data from the laser galvanometer after executing control commands, performing data processing and calculations according to steps S1-S7, and comparing the deviation between the actual data and the target data. If the deviation exceeds the specified range, an optimization algorithm (such as particle swarm optimization or genetic algorithm) is used to adjust the PID control parameters and deep learning model parameters, and retraining and control are performed until the accuracy requirements are met. Simultaneously, the performance of the decision trees in the random forest model is regularly evaluated. If any decision trees show performance degradation, optimization is performed through retraining or pruning.

[0133] A laser galvanometer intelligent control system based on big data, using any of the above-mentioned laser galvanometer intelligent control methods based on big data, such as Figure 4 As shown, the laser galvanometer intelligent control system based on big data includes:

[0134] The data acquisition module is used to collect the real-time position data, motion speed data, acceleration data, and environmental data of the laser galvanometer, and perform data packaging and time stamping processing respectively to obtain the original position data, original speed data, original acceleration data, and original environmental data;

[0135] The data preprocessing module is used to delete abnormal data and normalize the original position data, original velocity data, original acceleration data and original environment data to obtain valid position data, valid velocity data, valid acceleration data and valid environment data, and integrate them to obtain the preprocessed data set;

[0136] The feature extraction module is used to extract the laser galvanometer position change trend characteristics, velocity change rate characteristics, acceleration change trend characteristics, and related characteristics of the impact of environmental factors on the laser galvanometer performance from the preprocessed data set to form a feature vector set; the feature vector set is processed using the single pendulum wave technology to obtain an enhanced feature vector set;

[0137] The model training module is used to divide the enhanced feature vector set into a training set and a test set; use the training set to train the long short-term memory network model and the random forest model respectively; and perform a weighted fusion of the trained long short-term memory network model and the random forest model to obtain a laser galvanometer motion prediction model;

[0138] A real-time prediction module is used to input the real-time enhanced feature vector set into the laser galvanometer motion prediction model to obtain the predicted laser galvanometer position coordinates, velocity and acceleration;

[0139] The control instruction generation module is used to calculate the position deviation, velocity deviation, and acceleration deviation between the predicted laser galvanometer position coordinates, velocity, and acceleration and the preset target motion state; adopts the proportional-integral-differential control algorithm to calculate the control instructions for position, velocity, and acceleration based on the above deviations to form a control instruction set;

[0140] The control instruction execution module is used to transmit the control instruction set to the control system of the laser galvanometer, adjust the motor drive parameters of the laser galvanometer, and realize real-time control of the motion state of the laser galvanometer;

[0141] The feedback and optimization module is used to collect the actual motion state data of the laser galvanometer again after it executes the control instruction, and compare and analyze the newly collected data with the predicted data and the control target data. If the deviation between the actual motion state and the target motion state exceeds the preset error range, the parameters in the proportional-integral-differential control algorithm and the parameters of the laser galvanometer motion prediction model are readjusted to optimize the model.

[0142] The above embodiments are intended to illustrate the present invention, not to limit the present invention. Therefore, changes in illustrative values ​​or substitutions of equivalent components should still fall within the scope of the present invention.

[0143] From the above detailed description, it will be clear to those skilled in the art that the present invention can indeed achieve the aforementioned objectives and is in compliance with the provisions of the Patent Law.

[0144] Although preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they become aware of the basic inventive concepts. Therefore, the appended claims are intended to be interpreted as covering the preferred embodiments and all changes and modifications that fall within the scope of the invention. The foregoing description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. It should be noted that any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention are intended to be included within the scope of protection of the present invention.

[0145] It should be noted that the above description of the relevant processes is for illustration and purpose only and does not limit the scope of application of this specification. For those skilled in the art, various modifications and changes can be made to the processes under the guidance of this specification. However, such modifications and changes are still within the scope of this specification.

[0146] The basic concepts have been described above. It will be apparent to those skilled in the art after reading this application that the above disclosures are merely illustrative and do not constitute limitations on this application. Although not explicitly stated herein, those skilled in the art may make various modifications, improvements, and amendments to this application. Such modifications, improvements, and amendments are suggested in this application and remain within the spirit and scope of the exemplary embodiments of this application.

[0147] At the same time, this application uses specific terms to describe the embodiments of this application. For example, "one embodiment," "an embodiment," and / or "some embodiments" refer to a certain feature, structure, or characteristic related to at least one embodiment of this application. Therefore, it should be emphasized and noted that "one embodiment," "an embodiment," or "an alternative embodiment" mentioned twice or more in different places in this specification does not necessarily refer to the same embodiment. In addition, certain features, structures, or characteristics in one or more embodiments of this application may be appropriately combined.

[0148] Furthermore, those skilled in the art will appreciate that various aspects of the present application may be illustrated and described in terms of a number of patentable categories or situations, including any new and useful process, machine, product, or combination of substances, or any new and useful improvement thereof. Thus, various aspects of the present application may be implemented entirely in hardware, entirely in software (including firmware, resident software, microcode, etc.), or in a combination of hardware and software. Each of the above hardware and software may be referred to as a "unit," "module," or "system." Furthermore, various aspects of the present application may take the form of a computer program product embodied in one or more computer-readable media, with computer-readable program code embodied therein.

[0149] The computer program code required for the operation of each part of this application can be written in any one or more programming languages, including object-oriented programming languages ​​such as Java, Scala, Smalltalk, Eiffel, JADE, Emerald, C++, C#, VB.NET, Python, etc., conventional procedural programming languages ​​such as C programming language, Visual Basic, Fortran2103, Perl, COBOL2102, PHP, ABAP, dynamic programming languages ​​such as Python, Ruby and Groovy, or other programming languages. The program code can be run entirely on the user's computer, or as a standalone software package on the user's computer, or partly on the user's computer and partly on a remote computer, or entirely on a remote computer or server. In the latter case, the remote computer can be connected to the user's computer via any network form, such as a local area network (LAN) or a wide area network (WAN), or connected to an external computer (e.g., via the Internet), or in a cloud computing environment, or used as a service such as software as a service (SaaS).

[0150] In addition, unless expressly stated in the claims, the order of the processing elements and sequences described in this application, the use of alphanumeric characters, or the use of other names are not intended to limit the order of the processes and methods of this application. Although the above disclosure discusses some embodiments of the invention that are currently considered useful through various examples, it should be understood that such details are only for illustrative purposes, and the attached claims are not limited to the disclosed embodiments. On the contrary, the claims are intended to cover all modifications and equivalent combinations that are consistent with the essence and scope of the embodiments of this application. For example, although the implementation of the various components described above can be embodied in a hardware device, it can also be implemented as a pure software solution, for example, installation on an existing server or mobile device.

[0151] Similarly, it should be noted that in order to simplify the presentation of this disclosure and thereby facilitate understanding of one or more of the invention's embodiments, the foregoing descriptions of the embodiments of this disclosure sometimes combine multiple features into a single embodiment, figure, or description thereof. However, this approach should not be interpreted as reflecting an intention that the claimed subject matter requires more features than expressly recited in each claim. Rather, the subject matter of the invention may possess fewer features than the single embodiment described above.

Claims

1. A laser galvanometer intelligent control method based on big data, characterized in that: include: S1. Data acquisition: Collect real-time position data, motion speed data, acceleration data, and environmental data of the laser galvanometer, package and timestamp the data, and obtain raw position data, raw velocity data, raw acceleration data, and raw environmental data; S2. Data preprocessing; Abnormal data is deleted and normalized on the original position data, original velocity data, original acceleration data and original environment data to obtain valid position data, valid velocity data, valid acceleration data and valid environment data, and then integrated to obtain the preprocessed data set; S3. Feature extraction; Extract the laser galvanometer position change trend characteristics, velocity change rate characteristics, acceleration change trend characteristics and related characteristics of the impact of environmental factors on the laser galvanometer performance from the preprocessed data set. Composed feature vector set; The eigenvector set is processed using the single pendulum wave technique to obtain an enhanced eigenvector set; S4. Model training: Divide the enhanced feature vector set into a training set and a test set; use the training set to train the long short-term memory network model and the random forest model respectively; perform a weighted fusion of the trained long short-term memory network model and the random forest model to obtain a laser galvanometer motion prediction model; S5. Real-time prediction; inputting the real-time enhanced feature vector set into the laser galvanometer motion prediction model to obtain the predicted laser galvanometer position coordinates, velocity, and acceleration; S6. Control instruction generation: Based on the predicted laser galvanometer position coordinates, velocity, and acceleration and the preset target motion state, the position deviation, velocity deviation, and acceleration deviation between the two are calculated; using the proportional-integral-differential control algorithm, the control instructions for position, velocity, and acceleration are calculated based on these deviations to form a control instruction set; S7. Control instruction execution: transmitting the control instruction set to the laser galvanometer control system, adjusting the laser galvanometer motor drive parameters, and achieving real-time control of the laser galvanometer motion state; S8. Feedback and Optimization: After the galvanometer executes the control command, its actual motion state data is collected again, and steps S1-S7 above are repeated. The newly collected data is compared and analyzed with the predicted data and the control target data. If the deviation between the actual motion state and the target motion state exceeds the preset error range, the parameters of the proportional-integral-differential control algorithm and the parameters of the galvanometer motion prediction model are readjusted to optimize the model. In S3, the single pendulum wave technology simulates the periodic motion of a simple pendulum to perform fluctuation analysis and enhancement on the data in the feature vector set, thereby mining the potential periodic characteristics of the data; The pendulum wave function is based on the classical pendulum motion equation, and its standard form is: ; in, is the amplitude, which indicates the intensity of the fluctuation; is the angular frequency, controlling the cycle length, , For the cycle; For time; For the initial phase, adjust the starting position of the waveform; is the offset, used for fluctuations at non-zero centers; The parameters of the simple pendulum wave function are adaptively determined according to the motion characteristics of the laser galvanometer: amplitude : ； in is the feature vector set The standard deviation of is the scaling factor; cycle : ； in is the average peak-to-peak time interval of the characteristic fluctuations, is the adjustment factor; Initial Phase : ； in is the covariance, is the variance, which is used to align the changing trend of the data; Specific implementation steps of feature enhancement: Time-frequency analysis: Perform short-time Fourier transform or wavelet transform on the features in the feature vector set to extract the main frequency components and identify the dominant period of the laser galvanometer motion. , as the initial value of the simple pendulum wave period; Parameter initialization: Set the initial parameters according to the time-frequency analysis results ;Use different parameter groups for environmental features and decouple them from motion features; Characteristic modulation: ; in is the modulation intensity coefficient, Characterized by timestamp.

2. The laser galvanometer intelligent control method based on big data according to claim 1, characterized in that: In the above-mentioned S1, a fixed sampling frequency is set to periodically collect the position data, velocity data, acceleration data and environmental data of the laser galvanometer.

3. The laser galvanometer intelligent control method based on big data according to claim 1, characterized in that: In the above S2, the abnormal data is judged based on: the position coordinates, speed, acceleration or environmental data of the laser galvanometer exceeds a preset reasonable range.

4. The laser galvanometer intelligent control method based on big data according to claim 1, characterized in that: In S3, the position change trend feature is obtained by calculating the difference between the position coordinates of adjacent time points; the speed change rate feature is obtained by calculating the ratio of the difference between the speeds of adjacent time points to the time interval; and the acceleration change trend feature is obtained by calculating the difference between the accelerations of adjacent time points.

5. The laser galvanometer intelligent control method based on big data according to claim 1, characterized in that: In S4, the mean square error is used as the loss function during the training of the long short-term memory network model, and the model parameters are adjusted through the back propagation algorithm.

6. The laser galvanometer intelligent control method based on big data according to claim 1, characterized in that: In S4, the random forest model is composed of multiple decision trees. When constructing each decision tree, training samples are selected by random sampling with replacement, and some features are randomly selected for node splitting.

7. The laser galvanometer intelligent control method based on big data according to claim 1, characterized in that: In S4, when the long short-term memory network model and the random forest model are weightedly fused, the weighting coefficient is determined by adjusting the parameters on the validation set.

8. The laser galvanometer intelligent control method based on big data according to claim 1, characterized in that: In S8, a particle swarm optimization algorithm or a genetic algorithm is used to adjust the parameters in the proportional-integral-differential control algorithm and the parameters of the laser galvanometer motion prediction model.

9. A laser galvanometer intelligent control system based on big data, characterized in that: The laser galvanometer intelligent control method based on big data according to any one of claims 1 to 8 is used, and the laser galvanometer intelligent control system based on big data includes: The data acquisition module is used to collect the real-time position data, motion speed data, acceleration data, and environmental data of the laser galvanometer, and perform data packaging and time stamping processing respectively to obtain the original position data, original speed data, original acceleration data, and original environmental data; The data preprocessing module is used to delete abnormal data and normalize the original position data, original velocity data, original acceleration data and original environment data to obtain valid position data, valid velocity data, valid acceleration data and valid environment data, and integrate them to obtain the preprocessed data set; The feature extraction module is used to extract the laser galvanometer position change trend characteristics, velocity change rate characteristics, acceleration change trend characteristics, and related characteristics of the impact of environmental factors on the laser galvanometer performance from the preprocessed data set to form a feature vector set; the feature vector set is processed using the single pendulum wave technology to obtain an enhanced feature vector set; The model training module is used to divide the enhanced feature vector set into a training set and a test set; use the training set to train the long short-term memory network model and the random forest model respectively; and perform a weighted fusion of the trained long short-term memory network model and the random forest model to obtain a laser galvanometer motion prediction model; A real-time prediction module is used to input the real-time enhanced feature vector set into the laser galvanometer motion prediction model to obtain the predicted laser galvanometer position coordinates, velocity and acceleration; The control instruction generation module is used to calculate the position deviation, velocity deviation, and acceleration deviation between the predicted laser galvanometer position coordinates, velocity, and acceleration and the preset target motion state; adopts the proportional-integral-differential control algorithm to calculate the control instructions for position, velocity, and acceleration based on the above deviations to form a control instruction set; The control instruction execution module is used to transmit the control instruction set to the control system of the laser galvanometer, adjust the motor drive parameters of the laser galvanometer, and realize real-time control of the motion state of the laser galvanometer; The feedback and optimization module is used to collect the actual motion state data of the laser galvanometer again after it executes the control instruction, and compare and analyze the newly collected data with the predicted data and the control target data. If the deviation between the actual motion state and the target motion state exceeds the preset error range, the parameters in the proportional-integral-differential control algorithm and the parameters of the laser galvanometer motion prediction model are readjusted to optimize the model.

Images