Improved spgd method for laser coherent combining system

Abstract

The application belongs to the field of laser technology and application, and discloses an improved SPGD algorithm for a laser coherent synthesis system, which is used for solving the problem that the traditional SPGD algorithm cannot meet the bandwidth requirement of large-scale laser beam phase control. The operation mode of the algorithm is as follows: setting an initial phase control voltage and an initial gain coefficient; calculating the random disturbance voltage amplitude of this iteration; applying positive and negative random disturbance voltages satisfying Bernoulli distribution; according to the evaluation function value, dividing two cases, and performing corresponding segmented value on the gain coefficient; calculating the phase control voltage of each iteration; obtaining the updated system performance evaluation function value, and judging whether it converges or not, and if not, repeating the above steps until the algorithm converges. The improved SPGD algorithm can ensure the stability of the convergence process while improving the convergence speed of the laser beam phase tending to be consistent, and can be extended to coherent synthesis systems with different beam numbers without modifying any parameters, and is simple to operate and has strong universality.

Description

Technical fields:

[0001] This invention relates to adaptive optics technology, and more specifically to an improved SPGD algorithm for laser coherent combining systems. Background technology:

[0002] High-energy fiber lasers are widely used in industrial manufacturing, medical, and military fields. However, limited by the power output of a single fiber laser, coherent combining of multiple laser beams is currently the main technical approach to obtain high-power and high-quality laser output. The key to laser coherent combining is achieving precise phase control of each laser beam. Main techniques include heterodyne methods, multi-jitter methods, stochastic parallel gradient algorithms, and machine learning algorithms that have emerged in recent years. Among these, the SPGD algorithm, which combines artificial neural network technology and stochastic approximation theory, is the most mature. This algorithm estimates the correct iteration direction by the change and perturbation of the evaluation function, applies a voltage to the phase controller of each laser beam to control the phase change, and the evaluation function gradually approaches its maximum value during iteration, causing the phases of each laser beam to gradually converge. The SPGD algorithm has a simple principle, control logic, and system structure, and can achieve rapid convergence without precise measurement of the phase of each laser beam. It is currently widely used in laser coherent combining systems. However, as the number of laser paths increases, the number of iterations required for the SPGD algorithm to converge will increase exponentially, and the corresponding phase control bandwidth will decrease rapidly, making it difficult to meet the real-time requirements of large-scale laser coherent synthesis systems. Summary of the Invention:

[0003] The purpose of this invention is to overcome the shortcomings of the prior art and provide an improved SPGD algorithm for laser coherent combining systems. The improved algorithm can improve the convergence speed while ensuring the stability of the convergence process, and can be extended to coherent systems with different beam numbers without modifying any parameters. It is simpler to operate and more versatile.

[0004] The technical solution of this invention to solve the above-mentioned technical problems is: an improved SPGD algorithm for laser coherent combining systems, comprising the following steps:

[0005] (S1) Using the power received by the photodetector at the receiving end in the bucket as the evaluation function of the system performance, set the initial phase control voltage and the initial gain coefficient γ0, and measure the initial evaluation function value J. (0) ;

[0006] (S2) Update the random perturbation voltage amplitude required for this iteration based on the current evaluation function value:

[0007]

[0008] (S3) Apply a random perturbation voltage following a Bernoulli distribution to the phase modulator in the coherent combining system, and obtain the positive and negative evaluation function values ​​and the difference between them:

[0009] The steps for applying the random perturbation voltage during the nth iteration are as follows:

[0010] (S3-1) Apply a positive random disturbance voltage u (n) +δu (n) The positive evaluation function J is obtained. + (n) ;

[0011] (S3-2) Apply a negative random disturbance voltage u (n) -δu (n) The positive evaluation function J is obtained. - (n) ;

[0012] (S3-3) Calculate the difference δJ between the positive and negative evaluation functions. (n) =J + (n) -J - (n) ;

[0013] (S4) Update the gain coefficient required for this iteration based on S2: Determine max(J) + (n) J - (n) ) and J (n) The relationship between max(J) + (n) J - (n) )≤J (n) Then let γ = γ0, if max(J + (n) J - (n) )>J (n) Then let

[0014] (S5) Calculate the phase control voltage u based on the parameters obtained in the above steps. (n+1) =u (n) +γδJ (n) δu (n) It is then applied to a phase modulator to adjust the phase of the laser beam;

[0015] (S6) The updated system performance evaluation function value J is measured in the photodetector at the receiving end. (n+1) Then determine whether it converges. If it converges, end the algorithm; otherwise, repeat the above steps.

[0016] Preferably, in step (S2), a variable random disturbance voltage amplitude is used, and the functional relationship between the amplitude and the system performance evaluation function value is determined. The range of variation of the random disturbance voltage amplitude is set to a fixed range, so that no debugging or modification is required when applied to coherent synthesis systems with different numbers of channels.

[0017] Preferably, in step (S4), the evaluation function value obtained from the previous iteration is compared with the positive and negative evaluation function values ​​obtained after applying perturbation in the current iteration, and the result is determined according to max(J). + (n) J - (n) )>J (n) (Case 1) and max(J) + (n) J - (n) )≤J (n) (Scenario 2) The gain coefficient is segmented and assigned values ​​in two ways. In Scenario 1, the gain coefficient takes the initial value; in Scenario 2, the gain coefficient takes the reciprocal of the absolute value of the difference between the positive and negative evaluation function values, which is equivalent to directly taking J... (n+1) Updated to J + (n) With J - (n) The larger value in the range.

[0018] The advantages of this invention compared to the prior art are as follows:

[0019] 1. The improved SPGD algorithm for laser coherent combining systems in this invention employs a variable random perturbation voltage amplitude. In the early convergence phase, when the evaluation function value is small, a larger random perturbation voltage amplitude is used to increase the magnitude of a single rise in the evaluation function value and facilitate escaping local optima. In the later convergence phase, when the evaluation function value is large, a smaller random perturbation voltage amplitude is used for accurate convergence. The random perturbation voltage amplitude adaptively changes with the evaluation function value, reducing the number of iterations required for system performance evaluation function convergence while ensuring the convergence accuracy of the evaluation function, thus improving the algorithm's convergence speed and better meeting the high bandwidth requirements of large-scale coherent combining systems.

[0020] 2. The improved SPGD algorithm for laser coherent combining systems in this invention segments the gain coefficient. It compares the larger of the positive and negative evaluation function values ​​obtained from the previous iteration with those obtained after applying perturbation in the current iteration, classifying them into two different cases and assigning corresponding values ​​to the gain coefficient. This improved SPGD algorithm utilizes a simple operation method, rationally making use of the variables generated during the iteration process without increasing the time required for a single iteration. This further reduces the number of iterations required for the system performance evaluation function to converge and improves the stability of its convergence process and results.

[0021] 3. The present invention modifies the phase control voltage by segmenting the gain coefficient, which improves the compatibility of the laser coherent combining system with laser beams of different sizes. When applied to systems with other types of beams, no parameters need to be adjusted or modified, and there is no need to change the selection range of the random disturbance voltage amplitude or the functional relationship between the random disturbance voltage amplitude and the system performance evaluation function value. Only an appropriate gain coefficient needs to be selected in the initial stage. The operation is simple and highly scalable. Attached image description:

[0022] Figure 1 This is a simplified general block diagram of an embodiment of the present invention;

[0023] Figure 2 This is a flowchart of an improved SPGD algorithm for a laser coherent combining system according to the present invention;

[0024] Figure 3 A comparison chart of convergence speeds for embodiments of the present invention (Piecewise SPGD is the specific name of the improved algorithm of the present invention);

[0025] Figure 4 This is a comparison chart of the convergence stability of embodiments of the present invention. Detailed implementation method:

[0026] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be described in further detail below with reference to specific embodiments and accompanying drawings, but the implementation of this invention is not limited thereto.

[0027] Figure 1 The simplified block diagram of the laser coherent combining system shows that after passing through a beam splitter, the seed laser is divided into N sub-beams. These sub-beams are then connected to N phase controllers for phase adjustment. The adjusted beams are then focused by a beam combiner and a lens and projected onto the surface of a photodetector. The detected power in the barrel is mapped to a system performance evaluation function value and input into the improved SPGD algorithm module. Figure 2The process shown outputs a phase control voltage to the phase controller after processing. Numerical simulation experiments of the improved SPGD algorithm were conducted on the MATLAB platform. Each beam was simplified to a Gaussian beam with the same polarization fundamental mode. The number of coherent combining paths was set to 7, and a regular hexagonal array was used to obtain a high duty cycle. The initial phase of each beam followed a random distribution from 0 to 2π. The power in the bucket was used as the system evaluation function, with a value ranging from 0 to 1. The specific simulation parameter settings for the system are shown in Table 1.

[0028]

[0029] Table 1

[0030] See Figure 2 The improved SPGD algorithm of this invention for this application case includes the following steps:

[0031] (S1) Set the initial phase control voltage and initial gain coefficient γ0, and input the system performance evaluation function value into the improved SPGD algorithm module to obtain the initial evaluation function value J. (0) ;

[0032] (S2) The improved SPGD algorithm module updates the random perturbation voltage amplitude required for this iteration based on the current evaluation function value:

[0033] (S3) Apply a random perturbation voltage that follows a Bernoulli distribution to the phase modulator and obtain the positive and negative evaluation function values ​​and the difference between them:

[0034] The steps for applying a random perturbation voltage to the phase modulator during the nth iteration are as follows:

[0035] (S3-1) Apply a positive random disturbance voltage u (n) +δu (n) The positive evaluation function J is obtained. + (n) ;

[0036] (S3-2) Apply a negative random disturbance voltage u (n) -δu (n) The positive evaluation function J is obtained. - (n) ;

[0037] (S3-3) Calculate the difference δJ between the positive and negative evaluation functions. (n) =J + (n) -J - (n) ;

[0038] (S4) Update the gain coefficient required for this iteration based on S2: Determine max(J) + (n) J - (n) ) and J (n) The relationship between max(J) + (n) J - (n) )≤J (n) Then let γ = γ0, if max(J + (n) J - (n) )>J (n) Then let

[0039] (S5) Calculate the phase control voltage u based on the parameters obtained in the above steps. (n+1) =u (n) +γδJ (n) δu (n) The improved SPGD algorithm module is applied to the phase modulator to adjust the phase of the sub-beam;

[0040] (S6) The updated system performance evaluation function value J is measured in the photodetector at the receiving end. (n+1) Then determine whether it converges. If it converges, the algorithm ends, and the phases of each sub-beam tend to be consistent. If it does not converge, repeat the above steps until the evaluation function tends to a maximum value.

[0041] To demonstrate the superiority of the improved SPGD algorithm of this invention for laser coherent combining systems, the improved algorithm of this invention is compared with other typical improved algorithms and traditional algorithms. Table 2 shows the names and parameters of the compared algorithms, where Piecewise SPGD is the specific name of the improved algorithm of this invention. Figure 3 The graph showing the convergence speed comparison of embodiments of the present invention displays the iterative curve of the evaluation function of a typical SPGD algorithm. It is evident that the improved algorithm of the present invention has a faster convergence speed. Figure 4 The diagram shows a comparison of the convergence stability of embodiments of the present invention. It displays the distribution of the number of iterations when the evaluation function of each algorithm converges to 0.95. It can be seen that the improved algorithm of the present invention has the most concentrated distribution of the number of iterations when it converges and has the most stable performance.

[0042]

[0043] Table 2

[0044] The above are preferred embodiments of the present invention, but the embodiments of the present invention are not limited to the above content. Any changes, modifications, substitutions, combinations, or simplifications made without departing from the spirit and principle of the present invention shall be considered equivalent substitutions and shall be included within the protection scope of the present invention.

Claims

1. An improved SPGD algorithm for laser coherent combining systems, wherein the algorithm employs variable perturbation amplitude and segments the gain coefficient; characterized in that, Includes the following steps: (S1) Using the power received by the photodetector at the receiving end in the bucket as the evaluation function of system performance, set the initial phase control voltage and initial gain coefficient. The initial evaluation function value was obtained by measurement. ; (S2) Update the random perturbation voltage amplitude required for this iteration based on the current evaluation function value: ;in This is the evaluation function value at the nth iteration; (S3) Apply a random perturbation voltage following a Bernoulli distribution to the phase modulator in the coherent combining system, and obtain the positive and negative evaluation function values ​​and the difference between them: The steps for applying the random perturbation voltage during the nth iteration are as follows: (S3-1) Apply a positive random disturbance voltage The positive evaluation function is obtained. ;in This represents the control voltage for the nth iteration; (S3-2) Apply negative random disturbance voltage The negative evaluation function is obtained. ;in This represents the control voltage for the nth iteration; (S3-3) Calculate the difference between the positive and negative evaluation functions. ; (S4) Update the gain coefficient required for this iteration based on S2: Determine and The relationship between them, if Then let ,if Then let γ represents the gain coefficient. (S5) Calculate the phase control voltage based on the parameters obtained in the above steps. It is then applied to a phase modulator to adjust the phase of the laser beam; (S6) The updated system performance evaluation function value is measured in the photodetector at the receiving end. Then determine whether it converges. If it converges, end the algorithm. If it does not converge, repeat the above steps.

2. The improved SPGD algorithm for laser coherent combining systems according to claim 1, characterized in that: In step (S2), a variable random perturbation voltage amplitude was used, and the functional relationship between the amplitude and the system performance evaluation function value was determined. The range of variation of the random perturbation voltage amplitude was set to a fixed range, so no debugging or modification is required when applied to coherent combining systems with different numbers of paths. The system performance evaluation function includes the power in the barrel, the maximum output power, the main lobe power, the synthesized beam quality factor, and the pattern contrast of the far-field synthesized spot. The fixed range includes... The function can take any range of values, and the functional relationship includes linear relationships.

3. The improved SPGD algorithm for laser coherent combining systems according to claim 1, characterized in that: In step (S4), the evaluation function value is compared with the positive and negative evaluation function values ​​obtained after applying perturbation in this iteration, and the result is determined according to case one. Case 2 The gain coefficient is segmented into two cases: in case one, the gain coefficient takes the initial value; in case two, the gain coefficient takes the reciprocal of the absolute value of the difference between the positive and negative evaluation function values, which is equivalent to directly taking the initial value. Updated to and The larger value in the range, where, based on the relationship between the valence function value of the previous iteration and the positive and negative evaluation function values ​​obtained after applying perturbation in the current iteration, the number of division cases includes, but is not limited to, two or more, and the gain coefficient values ​​include, but are not limited to, fixed values ​​and parameters containing only... and The value that the function can take.

Images