Laser power nonlinear control method and system based on fuzzy neural network

Abstract

This invention proposes a laser power nonlinear control method and system based on fuzzy neural networks, relating to the field of laser control technology. The method includes: acquiring power control parameters of a fiber laser and constructing a power input feature vector based on the power control parameters; performing digital conversion and preprocessing on the power input feature vector to obtain control input parameters; fuzzifying the control input parameters to obtain membership data, and constructing a laser power nonlinear control model based on a fuzzy neural network according to a preset fuzzy rule base and the membership data; inputting the control input parameters into the laser power nonlinear control model, performing fuzzy inference and fuzzy neural network forward calculation to obtain a laser drive control quantity for adjusting the fiber laser, and using the laser drive control quantity to regulate the power of the fiber laser.

Description

Technical Field

[0001] This invention relates to the field of laser control technology, and in particular to a nonlinear control method and system for laser power based on a fuzzy neural network. Background Technology

[0002] Fiber lasers are widely used in precision machining, additive manufacturing, and other fields due to their advantages such as high beam quality, high energy density, and high efficiency. However, in practical applications, the output power of fiber lasers is affected by various factors, including pump source characteristics, fiber gain medium temperature, power supply fluctuations, and changes in cooling conditions. This results in strong nonlinear and time-varying characteristics, and the laser is easily disturbed by external disturbances, leading to problems such as slow dynamic response, large steady-state error, and insufficient robustness.

[0003] Chinese patent CN119129341A discloses a laser power adaptive control method for laser-assisted processing, including the following steps: S1, obtaining data through thermal modeling and simulation experiments; S2, designing a model predictive control algorithm; S3, actuator control software; controlling the laser's output power through laser control software, using the predicted output power of the model predictive control algorithm as input to the control software, thereby dynamically updating the laser's output power in real time; S4, designing and detecting sensors. However, the above scheme relies on a pre-established thermal finite element model in laser power control applications, making it difficult to reflect the nonlinear and time-varying characteristics of the laser itself in a timely manner, resulting in insufficient accuracy in power regulation of the laser under complex and multi-disturbance conditions. Therefore, it is essential to provide a nonlinear laser power control method and system based on fuzzy neural networks to improve the power regulation accuracy of fiber lasers. Summary of the Invention

[0004] In view of this, the present invention proposes a nonlinear control method and system for laser power based on fuzzy neural networks.

[0005] This invention provides a nonlinear control method for laser power based on a fuzzy neural network, the method comprising: The power control parameters of the fiber laser are acquired, and a power input feature vector is constructed based on the power control parameters; The power input feature vector is digitally converted and preprocessed to obtain control input parameters; The control input parameters are fuzzified to obtain membership data, and a nonlinear control model for laser power based on a fuzzy neural network is constructed according to a preset fuzzy rule base and the membership data. The control input parameters are input into the nonlinear control model of laser power, and fuzzy inference and fuzzy neural network forward calculation are performed to obtain the laser drive control quantity for adjusting the fiber laser, and the power of the fiber laser is regulated by the laser drive control quantity.

[0006] Based on the above technical solutions, preferably, the acquisition of power control parameters of the fiber laser and the construction of a power input feature vector based on the power control parameters specifically include: The target power setting value of the fiber laser is generated based on the target laser output power and the allowable power fluctuation range of the fiber laser. The current laser output power signal and auxiliary status sampling value of the fiber laser are collected, and the power signal is converted from analog to digital to generate real-time power sampling value. The auxiliary status sampling value includes ambient temperature, cooling water temperature, driving current and power supply voltage. Based on the target power setting value and the real-time power sampling value, the power error and error change rate of the fiber laser at the current moment are calculated, and the power error, the error change rate, and the auxiliary state sampling value are combined to form a power input feature vector.

[0007] Based on the above technical solutions, the preferred step of digitally converting and preprocessing the power input feature vector to obtain control input parameters specifically includes: The power input feature vector is filtered and denoised to generate filtered and optimized input features; The filtered and optimized input features are normalized with a preset range to obtain normalized control input parameters; Based on the input domain of the fuzzy neural network, the normalized control input parameters are interval mapped to generate control input parameters adapted to the fuzzy neural network.

[0008] More preferably, the step of fuzzifying the control input parameters to obtain membership data, and constructing a nonlinear control model for laser power based on a preset fuzzy rule base and the membership data, specifically includes: The control input parameters are partitioned into a domain and normalized to extract the power error, power error rate of change, and optional state parameters. Based on the power error, power error rate of change, and optional state parameters, the input variable set of the fuzzy neural network is constructed. Based on the value range of each input variable in the input variable set, configure corresponding fuzzy linguistic variables and membership functions, and perform fuzzification processing on the input variables to obtain the membership degree data of each input variable on different fuzzy subsets; Based on the preset fuzzy rule library, multiple fuzzy control rules are constructed and rule antecedent parameters and rule consequent parameters are generated, using power error and power error change rate as rule antecedents and laser drive control quantity as rule consequents. At the same time, a fuzzy neural network is constructed using the preset fuzzy rule library. The membership function center parameter, width parameter, and rule consequent weights in the fuzzy neural network are initialized to obtain an initial neural network weight parameter set. Based on the membership data, the rule antecedent parameters, the rule consequent parameters, and the initial neural network weight parameter set, a nonlinear control model for laser power is constructed. The nonlinear control model for laser power includes an input layer, a fuzzification layer, a rule layer, a normalization layer, and an output layer.

[0009] More preferably, the method further includes: The actual output power value of the fiber laser after adjustment based on the laser drive control quantity is obtained, and the membership function parameters and connection weight values ​​in the laser power nonlinear control model are updated according to the power deviation value between the actual output power value and the target power setting value of the fiber laser. During the multiple updates of membership function parameters and connection weight values, the power dynamic response performance index, steady-state error index, and robustness index of the fiber laser are comprehensively evaluated to obtain the power evaluation result corresponding to the fiber laser. Based on the power evaluation result, the control parameters and learning rate of the laser power nonlinear control model are co-optimized.

[0010] More preferably, updating the membership function parameters and connection weight values ​​in the laser power nonlinear control model specifically includes: The power deviation value of the fiber laser is obtained based on the target power setting value and the actual output power value. Using the power deviation value as a learning error signal, the membership function parameters and connection weight values ​​in the laser power nonlinear control model are adaptively updated using the gradient descent method to obtain the updated laser power nonlinear control model.

[0011] More preferably, obtaining the power evaluation result corresponding to the fiber laser specifically includes: During the process of updating the membership function parameters and connection weight values ​​in the nonlinear control model of laser power multiple times, the output power response data of the fiber laser is recorded in multiple control cycles. Based on the output power response data, the power dynamic response performance index, steady-state error index, and robustness index of the fiber laser are calculated respectively, and the power dynamic response performance index, the steady-state error index, and the robustness index are comprehensively evaluated to obtain the power evaluation result corresponding to the fiber laser.

[0012] A second aspect of this application provides a laser power nonlinear control system based on a fuzzy neural network. The laser power nonlinear control system includes a data acquisition module, a model building module, and a power regulation module, wherein... The data acquisition module is used to acquire the power control parameters of the fiber laser and construct a power input feature vector based on the power control parameters; The model building module is used to perform digital conversion and preprocessing on the power input feature vector to obtain control input parameters, perform fuzzification on the control input parameters to obtain membership data, and construct a nonlinear control model for laser power based on a preset fuzzy rule library and the membership data. The power regulation module is used to input the control input parameters into the laser power nonlinear control model, perform fuzzy inference and fuzzy neural network forward calculation to obtain the laser drive control quantity for adjusting the fiber laser, and use the laser drive control quantity to regulate the power of the fiber laser.

[0013] A third aspect of this application provides an electronic device including a processor, a memory, a user interface, and a network interface, wherein the memory is used to store instructions, the user interface and the network interface are used to communicate with other devices, and the processor is used to execute the instructions stored in the memory.

[0014] A fourth aspect of this application provides a non-transitory computer-readable storage medium having a computer program stored thereon, the computer program being executed by a processor to implement the steps of a laser power nonlinear control method based on a fuzzy neural network.

[0015] The laser power nonlinear control method and system based on fuzzy neural networks provided by this invention have the following advantages over existing technologies: (1) By constructing the power control parameters as feature vectors and performing digital preprocessing, combined with the nonlinear mapping capability of fuzzy neural networks, the nonlinear relationship between the power input and output of fiber lasers can be accurately characterized, significantly improving the power regulation accuracy of fiber lasers and reducing steady-state error. By combining fuzzy rules with the learning capability of neural networks, the nonlinearity, saturation characteristics and time-varying parameters of fiber lasers under different working states can be adaptively compensated, making the control quantity more consistent with the actual dynamic characteristics, improving the consistency and stability of power control. Furthermore, by using fuzzy inference and forward calculation to output the laser drive control quantity in real time, the system response time can be shortened, the overshoot and oscillation in the power regulation process can be reduced, the dynamic tracking performance can be improved, and the overall stability of the system can be enhanced. At the same time, fuzzy neural networks can adaptively adjust to external disturbances, load changes and environmental fluctuations, maintaining good control performance and robustness under conditions of parameter uncertainty and operating condition changes.

[0016] (2) By dividing the domain of the control input parameters and normalizing the mapping, and extracting the power error, the rate of change of power error and the optional state parameters, a more targeted set of input variables is constructed, which makes the fuzzy neural network more accurate in characterizing the nonlinear characteristics and dynamic process of laser power, improves the accuracy of the output of the control model, and uniformly initializes the membership function center parameter, width parameter and regular consequent weight, forming an initial set of neural network weight parameters, which provides a good starting point for subsequent online / offline learning, which is conducive to improving the convergence speed and stability of network training, and enables the model to adaptively adjust the control law according to the working conditions. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 A flowchart illustrating the nonlinear laser power control method based on a fuzzy neural network provided by this invention; Figure 2 A schematic diagram of the structure of the laser power nonlinear control system provided by the present invention; Figure 3 This is a schematic diagram of the structure of the electronic device provided by the present invention.

[0019] Explanation of reference numerals in the attached figures: 1. Laser power nonlinear control system; 11. Data acquisition module; 12. Model building module; 13. Power regulation module; 2. Electronic equipment; 21. Processor; 22. Communication bus; 23. User interface; 24. Network interface; 25. Memory. Detailed Implementation

[0020] The technical solutions of the present invention will be clearly and completely described below with reference to the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0021] This invention discloses a nonlinear control method for laser power based on a fuzzy neural network, with reference to... Figure 1 The steps of this method include S1 to S4.

[0022] Step S1: Collect the power control parameters of the fiber laser and construct a power input feature vector based on the power control parameters.

[0023] The power control parameters used to adjust the output power of the fiber laser can be obtained through an external controller connected to the fiber laser. These parameters may include the pump power supply's drive current setting, duty cycle setting, modulation frequency setting, operating mode setting, and parameters related to the fiber laser's operating status, such as ambient temperature and cooling water temperature. The control module collects these power control parameters within a preset sampling period and outputs them as digital signals. The external control system can be an industrial computer, a programmable logic controller (PLC), a numerical control system, or other host control devices with process parameter setting and process control functions.

[0024] The collected power control parameters are combined according to a preset parameter order to construct a power input feature vector characterizing the current power control state of the fiber laser. For example, the pump current setpoint, duty cycle setpoint, modulation frequency setpoint, and ambient temperature are denoted as follows: , , as well as Then the power input feature vector can be constructed as follows: .

[0025] Step S2 involves digitally converting and preprocessing the power input feature vector to obtain control input parameters.

[0026] Step S3: Fuzzyenize the control input parameters to obtain membership data, and construct a nonlinear control model for laser power based on a fuzzy neural network according to the preset fuzzy rule base and membership data.

[0027] Step S4: Input the control input parameters into the nonlinear control model of laser power, perform fuzzy inference and fuzzy neural network forward calculation to obtain the laser drive control quantity for adjusting the fiber laser, and use the laser drive control quantity to regulate the power of the fiber laser.

[0028] Specifically, laser drive control quantity Within each control cycle, the system is updated in real time based on the latest acquired power error, error rate of change, and selectable state parameters. By continuously inputting control input parameters into the laser power nonlinear control model, fuzzy inference and fuzzy neural network forward calculation are performed iteratively, and the laser drive control quantity is output and applied accordingly. Under complex operating conditions such as changes in ambient temperature, fluctuations in cooling conditions, power supply voltage disturbances, and load changes, the output power of the fiber laser can still achieve rapid tracking and stable adjustment of the desired setpoint, thereby improving the dynamic response performance and robustness of the laser power control system.

[0029] In this embodiment, by constructing power control parameters as feature vectors and performing digital preprocessing, combined with the nonlinear mapping capability of fuzzy neural networks, the nonlinear relationship between the power input and output of the fiber laser can be accurately characterized, significantly improving the power regulation accuracy of the fiber laser and reducing steady-state error. By combining fuzzy rules with the learning capability of neural networks, adaptive compensation can be performed on the nonlinearity, saturation characteristics, and time-varying parameters of the fiber laser under different operating states, making the control quantity more consistent with the actual dynamic characteristics, improving the consistency and stability of power control. Furthermore, by using fuzzy inference and forward calculation to output the laser drive control quantity in real time, the system response time can be shortened, overshoot and oscillation during power regulation can be reduced, dynamic tracking performance can be improved, and the overall system stability can be enhanced. At the same time, the fuzzy neural network can adaptively adjust to external disturbances, load changes, and environmental fluctuations, maintaining good control performance and robustness even under conditions of parameter uncertainty and changing operating conditions.

[0030] Based on the above embodiments, as an optional embodiment, step S1: acquiring the power control parameters of the fiber laser and constructing a power input feature vector based on the power control parameters may further include the following steps: Step S11: Generate the target power setting value of the fiber laser based on the target laser output power and the allowable power fluctuation range of the fiber laser.

[0031] First, based on the processing parameters to be executed or the system application requirements, the target laser output power and the corresponding allowable power fluctuation range of the fiber laser are determined in advance. The target laser output power can be set by the operator according to process conditions such as the type of material being processed, the thickness of the sheet, the processing speed, and the kerf width; this value is denoted as the target laser output power. The allowable power fluctuation range is used to characterize the permissible deviation of the actual output power from the target laser output power, and can be expressed as an absolute power deviation value. Alternatively, it can be set as a relative percentage.

[0032] Read the target laser output power The allowable power fluctuation range parameter is converted into a numerical target power setpoint for closed-loop control based on a preset power unit and numerical format. Specifically, the target power setpoint can be denoted as... And determine the corresponding power control accuracy range based on the allowable power fluctuation range, for example:

[0033] ,

[0034] in, This indicates the lower limit of the allowable fluctuation range of the fiber laser's output power. This represents the upper limit of the allowable fluctuation range of the fiber laser's output power, expressed as... or As a benchmark value for subsequent power error calculation, and This serves as the allowable fluctuation range for evaluating whether the output power meets the process requirements.

[0035] Step S12: Acquire the current laser output power signal and auxiliary status sampling value of the fiber laser, and perform analog-to-digital conversion on the power signal to generate real-time power sampling value. The auxiliary status sampling value includes ambient temperature, cooling water temperature, drive current and power supply voltage.

[0036] In this step, a power detection device, such as a power detector or photoelectric conversion sensor, is installed at the output end of the fiber laser to detect the current laser output power and output an analog voltage or current signal proportional to the output power. The output end of the power detection device is electrically connected to an external controller to sample the analog power signal within a preset sampling period. The analog power signal is then converted from analog to digital using an analog-to-digital converter to obtain the corresponding digital real-time power sample value. The external controller can then... k The real-time power sample value obtained within each control cycle is denoted as . .

[0037] Secondly, to more comprehensively reflect the operating status of the fiber laser, the external controller also collects auxiliary status sampling values ​​related to the laser's operating conditions. Specifically, ambient temperature sensors, cooling water temperature sensors, drive current detection units, and power supply voltage detection units can be set up in the fiber laser and its supporting systems to acquire status parameters such as the ambient temperature around the fiber laser, cooling water temperature, laser drive current, and power supply voltage. These auxiliary status signals can be analog or digital signals: for analog auxiliary status signals, analog-to-digital converters in the data acquisition module are used for analog-to-digital conversion to generate digital auxiliary status sampling values ​​corresponding to each status quantity; for digital signals, they can be directly read through the bus interface.

[0038] Within each control cycle, the external controller aligns and stores the real-time power sample value of the current laser output power signal with the corresponding auxiliary state sample value, forming a data set that corresponds one-to-one with the current control moment. The auxiliary state sample values ​​include at least parameters such as ambient temperature, cooling water temperature, drive current, and power supply voltage, and can be further expanded to include other state variables according to specific application requirements. These real-time power sample values ​​and auxiliary state sample values ​​serve as the foundational data for calculating power error, error change rate, and constructing the device power input feature vector.

[0039] Step S13: Calculate the power error and error rate of change of the fiber laser at the current moment based on the target power setpoint and the real-time power sampling value, and combine the power error, error rate of change, and auxiliary state sampling value to form a power input feature vector.

[0040] In the k Within each control cycle, the external controller reads the generated target power setpoint. and the obtained current real-time power sample value Based on the difference between the two values, the power error of the fiber laser at the current moment is calculated. For example, it can be defined as:

[0041]

[0042] in, Indicates the first k The power deviation within each control cycle reflects the degree of deviation of the actual output power from the target power setpoint. To characterize the trend of power error over time, the control system calculates the error change rate based on the power errors of two adjacent control cycles. Specifically, the power deviation within each control cycle can be... k Error change rate per control cycle Defined as:

[0043]

[0044] in, Indicates the first k The power deviation within one control cycle is expressed as the error change rate. This can reflect the dynamic changes in current power control. During system initial startup or when there is no historical error data, It can be initialized according to preset rules.

[0045] To obtain the power error at the current moment Error change rate After collecting auxiliary state sampling values, the control system combines the above data according to a preset feature arrangement order to form a power input feature vector. Specifically, the ambient temperature sampling values ​​within the current control cycle can be used as input. Cooling water temperature sampling value Drive current sampling value and power supply voltage sampling value Combined with power error and error change rate, an feature vector of the following form is constructed:

[0046]

[0047] Power input eigenvector Used for comprehensive characterization of fiber lasers in the first The power deviation status and its changing trend under each control cycle also reflect the environmental and electrical operating condition parameters that are closely related to power control.

[0048] Based on the above embodiments, as an optional embodiment, step S2: digitally converting and preprocessing the power input feature vector to obtain the control input parameters may further include the following steps: Step S21: Filter and denoise the power input feature vector to generate filtered and optimized input features.

[0049] In this step, the power input feature vector obtained in the previous step is read within the current control cycle. The power input feature vector includes at least the power error. Error change rate The system also includes auxiliary state sampling values ​​such as ambient temperature, cooling water temperature, drive current, and power supply voltage at corresponding times. For the raw data collected by sensors in each component, due to the superposition of measurement noise, sampling jitter, and electromagnetic interference, directly using it for control calculations will cause oscillations or overshoot in the output power regulation process. Therefore, it is necessary to perform filtering and noise reduction processing.

[0050] Based on the dynamic characteristics and response speed requirements of different feature quantities, a preset digital filtering algorithm is selected to filter each component of the power input feature vector. Specifically, one or more digital filtering methods can be used, such as moving average filtering, finite impulse response (FIR) low-pass filtering, infinite impulse response (IIR) low-pass filtering, or exponential weighted average filtering.

[0051] For example, for slowly changing variables such as ambient temperature and cooling water temperature, a moving average filter with a large window length can be used to effectively smooth random noise; for rapidly changing power errors... and error change rate If necessary, a low-pass filter with a suitable cutoff frequency can be used to suppress high-frequency noise while preserving the necessary dynamic response. The filter coefficients, window length, and cutoff frequency of the above filtering algorithm can be pre-tuned according to actual operating conditions during system calibration or debugging.

[0052] During the filtering operation, the control module combines the power input feature vector components of each control cycle with historical sampled data to calculate the corresponding filtered output value. For the first... In each control cycle, the filtered power error, error rate of change, and each auxiliary state variable can be denoted as follows: After completing the filtering operations for all components, the filtering results are recombine in the same order as the original power input feature vector to form the optimized input feature vector:

[0053]

[0054] in, Indicates the first k Power error after filtering within each control cycle Indicates the first k Rate of change of filtered error within each control cycle Indicates the first k Filtered ambient temperature sample values ​​within each control cycle Indicates the first k The filtered cooling water temperature sample value within each control cycle Indicates the first k The filtered drive current sample value within each control cycle Indicates the first k The filtered power supply voltage sample value within each control cycle.

[0055] Step S22: Normalize the filtered and optimized input features with a preset range to obtain normalized control input parameters.

[0056] In this step, the input feature vector is optimized for filtering. For each component, during the system design and calibration phase, a corresponding lower and upper limit value is preset based on its physical meaning and typical operating range.

[0057] Based on the aforementioned preset range, normalization operations are performed on each component of the filtered and optimized input feature vector, preferably using linear normalization. For any filtered feature... Its normalized value It can be represented as:

[0058]

[0059] This allows for mapping from the original physical quantity interval to the [0,1] interval. For applications requiring a symmetrical interval, it can be further mapped to the [-1,1] interval, as shown below:

[0060]

[0061] in, and These are the upper and lower limits of the preset range for the input feature vector used for filtering optimization. For filter feature values ​​that occasionally exceed the preset range, they can be handled according to the saturation principle, limiting them to the corresponding limits. Within the range, to avoid outliers from adversely affecting subsequent control calculations.

[0062] The normalized control input parameters after normalization can then be denoted as:

[0063] in, Indicates the first k Normalized power error within each control cycle Indicates the first k Normalized error change rate within each control cycle Indicates the first k Normalized ambient temperature sample values ​​within each control cycle Indicates the first k Normalized cooling water temperature sample values ​​within each control cycle Indicates the first k Normalized drive current sample value within each control cycle Indicates the first k Normalized power supply voltage sample value within each control cycle.

[0064] Step S23: Based on the input domain of the fuzzy neural network, perform interval mapping on the normalized control input parameters to generate control input parameters adapted to the fuzzy neural network.

[0065] In the design phase of a fuzzy neural network, the corresponding input universe of discourse is pre-defined for each input variable. For example, the universe of discourse for the power error input variable is set to... The universe of discourse for the error rate of change input variable is set to The domains of discourse for auxiliary state variables such as ambient temperature, cooling water temperature, drive current, and power supply voltage are set to their respective numerical ranges.

[0066] Based on the correspondence between the normalization interval and the universe of discourse of each input variable, an interval mapping operation is performed on each normalized control input parameter. Taking a normalization interval of [0,1] as an example, for any normalized input parameter... The control input parameters that are mapped to the input domain [a,b] of the fuzzy neural network x fnn It can be obtained through a linear mapping relationship:

[0067]

[0068] in, This represents the lower bound of the input universe of discourse for a fuzzy neural network. This represents the upper limit of the input domain of the fuzzy neural network. When the normalization interval is [-1, 1], it can be linearly converted to the [0, 1] interval, or it can be directly mapped to the domain of the corresponding input variable according to the corresponding linear proportional relationship. For different input variables, the control module uses interval mapping parameters that match their corresponding domains to achieve a one-to-one numerical conversion.

[0069] Based on the above embodiments, as an optional embodiment, step S3, which involves fuzzifying the control input parameters to obtain membership data and constructing a nonlinear control model for laser power based on a preset fuzzy rule base and membership data, may further include the following steps: Step S31: Perform universe partitioning and normalization mapping on the control input parameters to extract the power error, power error rate of change and optional state parameters from the control input parameters, and construct the input variable set of the fuzzy neural network based on the power error, power error rate of change and optional state parameters.

[0070] In this step, the current control cycle Below, the control input parameter vector has been adapted to the input domain of the fuzzy neural network. , denoted as:

[0071] in, Indicates the first k The power error in the control input parameters within each control cycle Indicates the first k Rate of change of error in control input parameters within each control cycle Indicates the first k The ambient temperature sample value is controlled in the input parameters of each control cycle. Indicates the first k The sampled value of cooling water temperature in the control input parameters within each control cycle. Indicates the first k The drive current sample value in the control input parameters within each control cycle. Indicates the first k The power supply voltage sample value in the control input parameters within each control cycle. , , as well as All are optional state parameters.

[0072] In the design and calibration phase of the fuzzy neural network, the upper and lower limits of its working universe of discourse are determined for power error, power error rate of change, and each selectable state parameter, and this universe of discourse is further linearly mapped to a unified internal standard universe of discourse (e.g., the [-1,1] interval). For the control input parameter vector... any control input component The variables mapped to the internal standard domain It can be represented as:

[0073]

[0074] in, Indicates control input components The lower bound within the presupposed domain, Indicates control input components The upper limit within the predefined domain, for , The optional state parameters are then mapped as described above to obtain the corresponding internal standard input quantities. , and several state inputs Based on the input quantities after partitioning and normalizing the universe of discourse, a set of input variables for the fuzzy neural network is constructed, including power error, power error rate of change, and... p The optional state parameters are uniformly represented as follows:

[0075]

[0076] in, Indicates the first k The power error in the set of input variables of the fuzzy neural network within one control cycle. Indicates the first kThe rate of change of error in the set of input variables of the fuzzy neural network within one control cycle. Indicates the first k The first of the input variables in the fuzzy neural network during the control cycle. p Optional state parameters.

[0077] Step S32: Configure corresponding fuzzy linguistic variables and membership functions according to the value range of each input variable in the input variable set, and perform fuzzification processing on the input variables to obtain the membership degree data of each input variable on different fuzzy subsets.

[0078] In this step, during the fuzzy neural network design and calibration stage, based on the distribution characteristics of each input variable in the internal standard universe of discourse and the control accuracy requirements, a corresponding set of fuzzy linguistic variables and membership functions are pre-configured for each input variable. The power error in the input variable set is then used as the basis for this process. and power error change rate For example, its fuzzy linguistic variable set can be set as follows:

[0079]

[0080]

[0081] in, Indicates power error A set of fuzzy linguistic variables Indicates the rate of change of power error The set of fuzzy linguistic variables, NB, NM, NS, ZO, PS, PM, and PB, represent fuzzy subsets such as "large negative, medium negative, small negative, zero, small positive, medium positive, and large positive," respectively. For each optional state parameter... Based on its physical meaning and range of change, its fuzzy linguistic variable set can be set as follows:

[0082]

[0083] Where L, M, and H represent fuzzy subsets such as "low", "medium", and "high", respectively.

[0084] Configure a specific membership function for each of the above fuzzy subsets. Triangular, trapezoidal, or Gaussian membership functions can be used to cover the internal standard universe of discourse [-1, 1]. Taking the first... m Input variables and its first j a fuzzy subset For example, when using the triangular membership function, its membership function can be expressed as:

[0085]

[0086] in, Representing fuzzy subsets The left-endpoint parameter in the internal standard universe of discourse, Representing fuzzy subsets The center point parameter in the internal standard universe of discourse, Representing fuzzy subsets The right-endpoint parameter in the internal standard universe of discourse, and satisfies .

[0087] Step S33: Based on the preset fuzzy rule library, multiple fuzzy control rules are constructed using power error and power error change rate as rule antecedents and laser drive control quantity as rule consequents, and rule antecedent parameters and rule consequent parameters are generated. At the same time, a fuzzy neural network is constructed using the preset fuzzy rule library.

[0088] In this step, a fuzzy control rule base is constructed based on a combination of fuzzy subsets of power error and fuzzy subsets of the rate of change of power error. Let the power error be divided into... A fuzzy subset, the power error change rate is divided into _ ... A fuzzy subset can then be used to construct a combination of multiple two-dimensional rules, the general form of which is the first fuzzy subset. r A fuzzy control rule can be expressed as:

[0089] like for ,and for And optional state parameters for Then the laser drive control quantity for .

[0090] in, Indicates the first r The fuzzy linguistic value corresponding to the power error in the rule, and ; Indicates the first r The fuzzy linguistic value corresponding to the power error change rate in the rule, and ; Indicates the first r The fuzzy language values ​​corresponding to the optional state parameters in the rule, and ; This represents the output control quantity description corresponding to the consequent of the rule, obtained by traversing... and All combinations can form These fuzzy control rules constitute a preset fuzzy rule library.

[0091] Based on typical operating condition data and control performance indicators, the aforementioned rule antecedent parameters and rule consequent parameters are initially tuned to ensure the speed and stability of laser power adjustment within common operating ranges. In implementations requiring adaptive adjustment, the rule antecedent parameters and rule consequent parameters can also be used as trainable weights of a fuzzy neural network, updated through a learning algorithm during online operation, thereby further improving control performance.

[0092] Step S34: Initialize the membership function center parameter, width parameter, and rule consequent weights in the fuzzy neural network to obtain the initial neural network weight parameter set.

[0093] In this step, for each input variable in the input layer Within its corresponding internal standard domain, based on the number of fuzzy subsets preset by the input variable. N m Determine the initial center parameters and width parameters of each fuzzy subset. An equally spaced center distribution can be used to divide the fuzzy subset into... m The first input variable j The central parameter of a fuzzy subset The value is set to be uniformly distributed along the internal standard universe of discourse, so that each fuzzy subset sequentially covers and overlaps with each other on the universe of discourse; at the same time, the width parameter is... Alternatively, the equivalent width can be set to a value proportional to the distance between adjacent centers to ensure a moderate overlap between adjacent membership functions, thereby achieving continuous fuzzy coverage of the input space. The aforementioned center and width parameters can be stored in the corresponding nodes of the fuzzification layer, serving as the initial antecedent parameters called when that node calculates its membership degree.

[0094] The membership function center parameter, width parameter, and rule consequent weights mentioned above are collectively denoted as the initial neural network weight parameter set. .in, Includes the set of central parameters of all fuzzy subsets of each input variable. Width parameter set and the set of consequent parameters corresponding to all rules. In subsequent offline training or online adaptive running phases, it is possible to... Based on this, the parameters are iteratively updated using gradient descent, least squares, or other optimization algorithms, so that the fuzzy neural network gradually approximates the nonlinear characteristics of the laser power control object, thereby achieving precise adjustment of the laser drive control quantity.

[0095] Step S35: Based on membership data, rule antecedent parameters, rule consequent parameters, and the initial neural network weight parameter set, construct a nonlinear control model for laser power. The nonlinear control model for laser power includes an input layer, a fuzzification layer, a rule layer, a normalization layer, and an output layer.

[0096] In this step, during the fuzzy neural network design and calibration phase, a fuzzy control rule base R is pre-constructed based on empirical knowledge of the laser power control mechanism and historical test data. Assume the number of input variables is... .

[0097] The input layer receives the set of input variables from the fuzzy neural network:

[0098] In the fuzzification layer, each input variable from the input layer is fuzzified according to the pre-configured set of fuzzy linguistic variables and their membership functions. Specifically, for the first... m Input variables Each fuzzy subset in its corresponding set of fuzzy linguistic variables Through the corresponding membership functions Calculate its membership degree:

[0099]

[0100] in, Indicates the relationship with the first m Input variables The corresponding membership function.

[0101] Each node in the fuzzification layer corresponds one-to-one with a certain fuzzy subset of a certain input variable, and its output is the membership degree of that fuzzy subset under the current input value, thus forming a membership degree dataset for all input variables within the network.

[0102] In the rule layer, based on a preset fuzzy rule library, using the fuzzy values ​​of power error and the rate of change of power error as primary antecedents, and optionally introducing fuzzy values ​​of state parameters as auxiliary antecedents, a rule node is configured for each preset fuzzy control rule. Let the number of preset rules be... R The rule layer contains R Rule node. r The general form of a rule can be expressed as:

[0103] Two of the required input variables are power error and the rate of change of power error, and the rest... Each of these is an optional state parameter. Therefore, the r-th fuzzy control rule can be expressed as:

[0104] like for ,and for And optional state parameters for Then the laser drive control quantity for In the rule layer, the first r Each rule node receives the membership degree from the corresponding subset of the fuzzy layer and performs an AND operation on it to calculate the activation strength of the rule within the current control cycle. . No. Activation intensity under each control cycle It can be represented as:

[0105]

[0106] in, Indicates the first r Rule number 1 m Fuzzy subsets corresponding to each input variable No. r Rule number 1 m The membership function corresponding to the fuzzy subset of each input variable.

[0107] In the normalization layer, the activation strength of each rule output from the rule layer is normalized to obtain the relative weight of each rule at the current time, thereby avoiding numerical instability caused by differences in the number of rules and activation strength. For the ... r The normalized weights of the rule can be expressed as:

[0108]

[0109] in, Indicates the first Under the control cycle, the first r The normalized weights of the rules, Indicates the first s The fuzzy rule in the first The activation intensity under each control cycle, each node of the normalization layer corresponds one-to-one with a rule, and its output satisfy This provides standardized weights for the weighted synthesis of the output layer.

[0110] In the output layer, based on a rule-based consequent structure, the laser-driven control quantity is used as the network output. A T-S type consequent form is adopted, representing the consequent output of each rule as a linear combination of the input variables, i.e., the... r The consequent output of a rule can be represented as:

[0111]

[0112] in, Indicates the first The first control cycle r The consequent output of the rule, , ... All are the first r The output layer performs a weighted sum of the consequent parameters of all rules to obtain the laser drive control quantity for the current control cycle.

[0113] in, Indicates the first Laser drive control quantity under each control cycle.

[0114] In this embodiment, by partitioning and normalizing the domain of the control input parameters, and extracting power error, power error rate of change, and optional state parameters, a more targeted set of input variables is constructed. This allows the fuzzy neural network to more accurately characterize the nonlinear characteristics and dynamic processes of laser power, improving the accuracy of the control model output. Based on the value range of each input variable, corresponding fuzzy linguistic variables and membership functions are configured. Multiple fuzzy control rules are constructed using power error and power error rate of change as rule antecedents and laser-driven control quantities as rule consequents. This intuitively reflects the control logic, giving the control strategy strong engineering interpretability and providing reasonable control quantities for different operating conditions. By uniformly initializing the membership function center parameters, width parameters, and rule consequent weights, an initial neural network weight parameter set is formed, providing a good starting point for subsequent online / offline learning. This improves the convergence speed and stability of network training, enabling the model to adaptively adjust the control law according to changes in operating conditions.

[0115] In this embodiment, steps S5 and S6 are also included.

[0116] Step S5: Obtain the actual output power value of the fiber laser after adjustment based on the laser drive control quantity, and update the membership function parameters and connection weight values ​​in the laser power nonlinear control model according to the power deviation value between the actual output power value and the target power setting value of the fiber laser.

[0117] In this step, the power deviation value of the fiber laser is obtained based on the target power setpoint and the actual output power value. Using the power deviation value as a learning error signal, the membership function parameters and connection weight values ​​in the laser power nonlinear control model are adaptively updated using the gradient descent method to obtain the updated laser power nonlinear control model.

[0118] Specifically, in the current control cycle Internally, firstly based on the laser drive control quantity After completing the power adjustment of the fiber laser, the actual output power sample value of the fiber laser is acquired in real time and recorded as follows: Simultaneously, the target power setpoint of the fiber laser within this control cycle is read from the upper-level control system and recorded as follows. Based on this, the power deviation value for the current control cycle is defined. , can be represented as:

[0119]

[0120] in, The nonlinear control model of laser power is represented in the first... Power deviation value under each control cycle.

[0121] With power deviation value As a learning error signal, an instantaneous performance index function is constructed:

[0122] in, Indicates the first Instantaneous performance index function under each control cycle.

[0123] The adjustable parameters in the nonlinear control model of laser power are updated online based on the gradient descent method. The nonlinear control model of laser power is implemented by the fuzzy neural network in step S33, and its first... The set of neural network weight parameters under each control cycle is denoted as . This includes the center parameter and width parameter of the membership function, as well as the connection weight parameters of each rule consequent. For any parameter to be updated... Its update law can be expressed as:

[0124]

[0125] in, Represents the first value in the set of neural network weight parameters. i The learning rate of each parameter This indicates the sensitivity of the laser drive control quantity to this parameter, which can be analytically differentiated based on the forward computation structure of the fuzzy neural network and the chain rule, or derived in advance according to a conventional form.

[0126] Furthermore, the learning rate can be adjusted based on the system's stability and convergence requirements. Differential quantization settings or adaptive adjustments can be made, such as selecting different upper limits for learning rates for different types of parameters, or appropriately increasing the learning rate when the error is large and gradually decreasing the learning rate when the error tends to converge, to avoid oscillations caused by overly rapid updates and slow convergence caused by overly slow updates. Furthermore, when necessary, boundary constraints can be applied to the updated membership function centers and width parameters to ensure that they remain within the preset internal standard universe of discourse and meet basic monotonicity and coverage requirements.

[0127] Step S6: During the multiple updates of membership function parameters and connection weight values, the power dynamic response performance index, steady-state error index, and robustness index of the fiber laser are comprehensively evaluated to obtain the power evaluation result corresponding to the fiber laser. Based on the power evaluation result, the control parameters and learning rate of the laser power nonlinear control model are co-optimized.

[0128] In this step, during the process of updating the membership function parameters and connection weight values ​​in the nonlinear control model of laser power multiple times, the output power response data of the fiber laser is recorded in multiple control cycles. Based on the output power response data, the power dynamic response performance index, steady-state error index, and robustness index of the fiber laser are calculated respectively. The power dynamic response performance index, steady-state error index, and robustness index are comprehensively evaluated to obtain the power evaluation results corresponding to the fiber laser.

[0129] Specifically, during the multiple updates of the membership function parameters and connection weights in the nonlinear control model of laser power, the control module continuously records the output power response data of the fiber laser and the corresponding expected power setpoint within multiple control cycles before, during, and after each parameter update, forming a power response dataset for performance evaluation. Based on the power response dataset, the following performance indicators are calculated:

[0130] Power dynamic response performance metrics: These metrics include, but are not limited to, rise time, settling time, overshoot, peak time, and a comprehensive response speed index. For example, under a step power given condition, rise time is defined as the time required for the output power to reach a certain percentage (e.g., 10%–90%) of the setpoint from its initial value; overshoot is the percentage deviation of the maximum peak output power from the setpoint; and settling time is the time required for the output power to enter and remain within a specified error band (e.g., ±2% or ±5%) near the setpoint. The control module can calculate a normalized dynamic performance score based on these sub-metrics to reflect the dynamic response quality of the fiber laser under the current control parameters and learning rate configuration.

[0131] Steady-state error index: After the power output enters a steady state, the deviation between the sampled output power values ​​and the desired power setpoint over several consecutive control cycles is statistically analyzed to obtain the steady-state error index. This index may include steady-state average error, mean absolute value of steady-state error, and steady-state mean square error, etc.

[0132] Robustness Indicators: To evaluate the stability and adaptability of the control system under external disturbances, changes in operating conditions, and model uncertainties, robustness indicators are formed by statistically analyzing the response data of the fiber laser under various typical disturbance scenarios. Disturbance scenarios can include step changes in ambient temperature, fluctuations in cooling water temperature, power supply voltage disturbances, and rapid load changes. Under these scenarios, the recovery time of output power, the peak instantaneous deviation caused by the disturbance, the change in steady-state error after the disturbance, and the degree of fluctuation in control input are recorded, and a robustness evaluation function can be constructed based on these data.

[0133] Based on the above method, this application discloses a laser power nonlinear control system based on a fuzzy neural network, referencing... Figure 2 The laser power nonlinear control system 1 includes a data acquisition module 11, a model building module 12, and a power control module 13, wherein... The data acquisition module 11 is used to acquire the power control parameters of the fiber laser and construct a power input feature vector based on the power control parameters; The model building module 12 is used to perform digital conversion and preprocessing of the power input feature vector to obtain control input parameters, perform fuzzification processing on the control input parameters to obtain membership data, and construct a laser power nonlinear control model based on a fuzzy neural network according to the preset fuzzy rule base and membership data. The power regulation module 13 is used to input the control input parameters into the laser power nonlinear control model, perform fuzzy inference and fuzzy neural network forward calculation to obtain the laser drive control quantity for adjusting the fiber laser, and use the laser drive control quantity to regulate the power of the fiber laser.

[0134] In one example, the data acquisition module 11 is used to generate a target power setting value for the fiber laser based on the target laser output power and the allowable power fluctuation range of the fiber laser; acquire the current laser output power signal and auxiliary state sampling values ​​of the fiber laser, and perform analog-to-digital conversion on the power signal to generate real-time power sampling values, wherein the auxiliary state sampling values ​​include ambient temperature, cooling water temperature, drive current and power supply voltage; calculate the power error and error change rate of the fiber laser at the current moment based on the target power setting value and the real-time power sampling values, and combine the power error, error change rate and auxiliary state sampling values ​​to form a power input feature vector.

[0135] In one example, the model building module 12 is used to filter and denoise the power input feature vector to generate filtered and optimized input features; normalize the filtered and optimized input features with a preset range to obtain normalized control input parameters; and perform interval mapping on the normalized control input parameters according to the input domain of the fuzzy neural network to generate control input parameters adapted to the fuzzy neural network.

[0136] In one example, the model building module 12 is used to partition and normalize the control input parameters to extract the power error, power error rate of change, and optional state parameters from the control input parameters. Based on the power error, power error rate of change, and optional state parameters, it constructs the input variable set of the fuzzy neural network. According to the value range of each input variable in the input variable set, it configures corresponding fuzzy linguistic variables and membership functions, performs fuzzification processing on the input variables to obtain the membership degree data of each input variable on different fuzzy subsets. Based on a preset fuzzy rule base, it uses power error and power error rate of change as... Using the laser-driven control quantity as the rule antecedent and the laser-driven control quantity as the rule consequent, multiple fuzzy control rules are constructed, and rule antecedent parameters and rule consequent parameters are generated. At the same time, a fuzzy neural network is constructed using a preset fuzzy rule library. The membership function center parameters, width parameters, and rule consequent weights in the fuzzy neural network are initialized to obtain an initial neural network weight parameter set. Based on the membership data, rule antecedent parameters, rule consequent parameters, and the initial neural network weight parameter set, a laser power nonlinear control model is constructed. The laser power nonlinear control model includes an input layer, a fuzzification layer, a rule layer, a normalization layer, and an output layer.

[0137] In one example, the system is also configured to acquire the actual output power value of the fiber laser after adjustment based on the laser drive control quantity, and update the membership function parameters and connection weight values ​​in the laser power nonlinear control model according to the power deviation value between the actual output power value and the target power setpoint of the fiber laser; during the multiple updates of membership function parameters and connection weight values, the power dynamic response performance index, steady-state error index and robustness index of the fiber laser are comprehensively evaluated to obtain the power evaluation result corresponding to the fiber laser, and the control parameters and learning rate of the laser power nonlinear control model are co-optimized based on the power evaluation result.

[0138] In one example, the membership function parameters in the nonlinear control model of laser power are updated, specifically including: The power deviation of the fiber laser is obtained by comparing the target power setting with the actual output power. Using the power deviation value as the learning error signal, the gradient descent method is used to adaptively update the membership function parameters and connection weight values ​​in the laser power nonlinear control model to obtain the updated laser power nonlinear control model.

[0139] In one example, obtaining the power evaluation results corresponding to a fiber laser includes: During the process of updating the membership function parameters and connection weight values ​​in the nonlinear control model of laser power multiple times, the output power response data of the fiber laser in multiple control cycles were recorded. Based on the output power response data, the power dynamic response performance index, steady-state error index, and robustness index of the fiber laser are calculated respectively. The power dynamic response performance index, steady-state error index, and robustness index are comprehensively evaluated to obtain the power evaluation results corresponding to the fiber laser.

[0140] Please see Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Figure 3 As shown, the electronic device 2 may include: at least one processor 21, at least one network interface 24, user interface 23, memory 25, and at least one communication bus 22.

[0141] The communication bus 22 is used to enable communication between these components.

[0142] The user interface 23 may include a display screen and a camera. Optionally, the user interface 23 may also include a standard wired interface and a wireless interface.

[0143] The network interface 24 may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface).

[0144] The processor 21 may include one or more processing cores. The processor 21 connects to various parts of the server using various interfaces and lines, and performs various server functions and processes data by running or executing instructions, programs, code sets, or instruction sets stored in the memory 25, and by calling data stored in the memory 25. Optionally, the processor 21 may be implemented using at least one hardware form of Digital Signal Processing (DSP), Field-Programmable Gate Array (FPGA), or Programmable Logic Array (PLA). The processor 21 may integrate one or a combination of several of the following: Central Processing Unit (CPU), Graphics Processing Unit (GPU), and modem. The CPU primarily handles the operating system, user interface, and applications; the GPU is responsible for rendering and drawing the content required for display; and the modem handles wireless communication. It is understood that the modem may also be implemented as a separate chip without being integrated into the processor 21.

[0145] The memory 25 may include random access memory (RAM) or read-only memory. Optionally, the memory 25 may include non-transitory computer-readable storage medium. The memory 25 can be used to store instructions, programs, code, code sets, or instruction sets. The memory 25 may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for at least one function (such as touch function, sound playback function, image playback function, etc.), instructions for implementing the above-described method embodiments, etc.; the data storage area may store data involved in the above-described method embodiments, etc. Optionally, the memory 25 may also be at least one storage device located remotely from the aforementioned processor 21. Figure 3 As shown, the memory 25, which serves as a computer storage medium, may include an operating system, a network communication module, a user interface module, and an application program for a nonlinear laser power control method based on a fuzzy neural network.

[0146] exist Figure 3In the electronic device 2 shown, the user interface 23 is mainly used to provide an input interface for the user and to obtain the user input data; while the processor 21 can be used to call an application program stored in the memory 25 for a nonlinear control method of laser power based on a fuzzy neural network. When executed by one or more processors, the electronic device executes one or more methods as described in the above embodiments.

[0147] A non-transitory computer-readable storage medium stores instructions that, when executed by one or more processors, cause a computer to perform one or more methods as described in the above embodiments.

[0148] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0149] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0150] In the several embodiments provided in this application, it should be understood that the disclosed apparatus can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the shown or discussed mutual couplings or direct couplings or communication connections may be through some service interfaces; indirect couplings or communication connections between apparatuses or units may be electrical or other forms.

[0151] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0152] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0153] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage device (CMD). Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a memory and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned memory includes various media capable of storing program code, such as USB flash drives, portable hard drives, magnetic disks, or optical disks.

[0154] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A nonlinear control method for laser power based on a fuzzy neural network, characterized in that, The method includes: The power control parameters of the fiber laser are acquired, and a power input feature vector is constructed based on the power control parameters; The power input feature vector is digitally converted and preprocessed to obtain control input parameters; The control input parameters are fuzzified to obtain membership data, and a nonlinear control model for laser power based on a fuzzy neural network is constructed according to a preset fuzzy rule base and the membership data. The control input parameters are input into the nonlinear control model of laser power, and fuzzy inference and fuzzy neural network forward calculation are performed to obtain the laser drive control quantity for adjusting the fiber laser, and the power of the fiber laser is regulated by the laser drive control quantity.

2. The nonlinear laser power control method based on fuzzy neural network as described in claim 1, characterized in that, The process of acquiring the power control parameters of the fiber laser and constructing a power input feature vector based on these parameters specifically includes: The target power setting value of the fiber laser is generated based on the target laser output power and the allowable power fluctuation range of the fiber laser. The current laser output power signal and auxiliary status sampling value of the fiber laser are collected, and the power signal is converted from analog to digital to generate real-time power sampling value. The auxiliary status sampling value includes ambient temperature, cooling water temperature, driving current and power supply voltage. Based on the target power setting value and the real-time power sampling value, the power error and error change rate of the fiber laser at the current moment are calculated, and the power error, the error change rate, and the auxiliary state sampling value are combined to form a power input feature vector.

3. The nonlinear laser power control method based on fuzzy neural network as described in claim 1, characterized in that, The digital conversion and preprocessing of the power input feature vector to obtain control input parameters specifically includes: The power input feature vector is filtered and denoised to generate filtered and optimized input features; The filtered and optimized input features are normalized with a preset range to obtain normalized control input parameters; Based on the input domain of the fuzzy neural network, the normalized control input parameters are interval mapped to generate control input parameters adapted to the fuzzy neural network.

4. The nonlinear laser power control method based on fuzzy neural network as described in claim 1, characterized in that, The step of fuzzifying the control input parameters to obtain membership data, and constructing a nonlinear laser power control model based on a fuzzy neural network according to a preset fuzzy rule base and the membership data, specifically includes: The control input parameters are partitioned into a domain and normalized to extract the power error, power error rate of change, and optional state parameters. Based on the power error, power error rate of change, and optional state parameters, the input variable set of the fuzzy neural network is constructed. Based on the value range of each input variable in the input variable set, configure corresponding fuzzy linguistic variables and membership functions, and perform fuzzification processing on the input variables to obtain the membership degree data of each input variable on different fuzzy subsets; Based on the preset fuzzy rule library, multiple fuzzy control rules are constructed and rule antecedent parameters and rule consequent parameters are generated, using power error and power error change rate as rule antecedents and laser drive control quantity as rule consequents. At the same time, a fuzzy neural network is constructed using the preset fuzzy rule library. The membership function center parameter, width parameter, and rule consequent weights in the fuzzy neural network are initialized to obtain an initial neural network weight parameter set. Based on the membership data, the rule antecedent parameters, the rule consequent parameters, and the initial neural network weight parameter set, a nonlinear control model for laser power is constructed. The nonlinear control model for laser power includes an input layer, a fuzzification layer, a rule layer, a normalization layer, and an output layer.

5. The nonlinear laser power control method based on fuzzy neural network as described in claim 4, characterized in that, The method further includes: The actual output power value of the fiber laser after adjustment based on the laser drive control quantity is obtained, and the membership function parameters and connection weight values ​​in the laser power nonlinear control model are updated according to the power deviation value between the actual output power value and the target power setting value of the fiber laser. During the multiple updates of membership function parameters and connection weight values, the power dynamic response performance index, steady-state error index, and robustness index of the fiber laser are comprehensively evaluated to obtain the power evaluation result corresponding to the fiber laser. Based on the power evaluation result, the control parameters and learning rate of the laser power nonlinear control model are co-optimized.

6. The nonlinear laser power control method based on fuzzy neural network as described in claim 5, characterized in that, The updating of the membership function parameters and connection weight values ​​in the nonlinear control model of laser power specifically includes: The power deviation value of the fiber laser is obtained based on the target power setting value and the actual output power value. Using the power deviation value as a learning error signal, the membership function parameters and connection weight values ​​in the laser power nonlinear control model are adaptively updated using the gradient descent method to obtain the updated laser power nonlinear control model.

7. The nonlinear laser power control method based on fuzzy neural network as described in claim 5, characterized in that, The acquisition of the power evaluation result corresponding to the fiber laser specifically includes: During the process of updating the membership function parameters and connection weight values ​​in the nonlinear control model of laser power multiple times, the output power response data of the fiber laser is recorded in multiple control cycles. Based on the output power response data, the power dynamic response performance index, steady-state error index, and robustness index of the fiber laser are calculated respectively, and the power dynamic response performance index, the steady-state error index, and the robustness index are comprehensively evaluated to obtain the power evaluation result corresponding to the fiber laser.

8. A nonlinear control system for laser power based on a fuzzy neural network, characterized in that, The laser power nonlinear control system is used to execute the laser power nonlinear control method based on a fuzzy neural network as described in any one of claims 1-7, wherein the laser power nonlinear control system includes a data acquisition module, a model building module, and a power regulation module, wherein... The data acquisition module is used to acquire the power control parameters of the fiber laser and construct a power input feature vector based on the power control parameters; The model building module is used to perform digital conversion and preprocessing on the power input feature vector to obtain control input parameters, perform fuzzification on the control input parameters to obtain membership data, and construct a laser power nonlinear control model based on a fuzzy neural network according to a preset fuzzy rule library and the membership data. The power regulation module is used to input the control input parameters into the laser power nonlinear control model, perform fuzzy inference and fuzzy neural network forward calculation to obtain the laser drive control quantity for adjusting the fiber laser, and use the laser drive control quantity to regulate the power of the fiber laser.

9. An electronic device, characterized in that, The device includes a processor, a memory, a user interface, and a network interface. The memory is used to store instructions, the user interface and the network interface are used to communicate with other devices, and the processor is used to execute the instructions stored in the memory to cause the electronic device to perform the method as described in any one of claims 1-7.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1-7.

Images