Artificial diamond synthesis optimization method based on adaptive growth algorithm

Abstract

The invention relates to the technical field of artificial diamond synthesis, and discloses an artificial diamond synthesis optimization method based on a self-adaptive growth algorithm, which comprises the following stages: firstly, constructing a digital twin model and a Pareto frontier objective function, then collecting chamber data in real time through a multi-modal sensor, constructing a state vector, and finally, constructing a multi-modal model; and then planning an artificial diamond macroscopic growth route based on an ant colony algorithm to obtain high-quality growth points, then taking the high-quality growth points as input, performing fine adjustment on parameters based on a particle swarm algorithm, and outputting a final parameter set to realize dynamic and accurate regulation and control of a growth environment. According to the method, a two-stage collaborative optimization framework of'ant colony algorithm macroscopic optimization + particle swarm algorithm microcosmic fine tuning 'is adopted, and the ant colony algorithm globally searches'high-quality growth points' on the surface of the seed crystal, so that the problem of poor local growth caused by parameter homogenization in a traditional method is solved.

Description

Technical Field

[0001] This invention relates to the field of artificial diamond synthesis technology, specifically to an optimized method for artificial diamond synthesis based on an adaptive growth algorithm. Background Technology

[0002] Synthetic diamonds, especially those synthesized via chemical vapor deposition (CVD) or high-temperature, high-pressure (HPHT) methods, have wide applications in industry, technology, and jewelry due to their superior optical, thermal, mechanical, and electronic properties. The core of the synthesis process lies in the precise control of key parameters such as temperature, pressure, carbon source concentration, and plasma density within a sealed growth chamber to achieve ordered deposition of carbon atoms and crystal growth on the seed crystal surface, ultimately yielding high-purity, large-size, and low-defect diamond crystals.

[0003] Currently, the mainstream methods for controlling the synthesis of artificial diamonds mainly rely on the following two types of methods: The first type is open-loop control based on fixed process formulations or empirical rules. Operators preset a fixed set of process parameter curves based on historical successful experience or limited experimental data. This method is simple and easy to implement, but lacks adaptability. Due to practical factors such as equipment condition fluctuations, raw material batch differences, and uneven crystal surface conditions, a fixed set of parameters is difficult to match the optimal growth conditions consistently. This can easily lead to uneven internal quality of crystals in a single growth cycle, unstable growth rates, and even defects, making it difficult to guarantee product consistency and yield.

[0004] The second category is closed-loop control based on traditional single-objective feedback. For example, a PID controller is used to stabilize a single key parameter (such as temperature). This type of method can handle some disturbances, but it has significant limitations: First, the control objective is singular, usually only pursuing the stability of a single parameter (such as temperature), failing to comprehensively consider multiple and mutually restrictive competing objectives such as "high purity," "high growth rate," and "low energy consumption," making it difficult to achieve Pareto optimality in overall performance. Second, the control dimensions are fragmented, failing to establish a synergistic mechanism between macroscopic growth region selection and microscopic process parameter adjustment. Existing methods often perform homogenized control of the chamber, ignoring the different growth potentials of different locations on the seed surface due to differences in geometry and defect density, resulting in inefficient resource (energy, carbon source) allocation. Third, dynamic response and foresight are insufficient. Traditional feedback control makes "retroactive" adjustments based on current or historical deviations, resulting in a delayed response, and frequent adjustments can easily cause system oscillations, which is detrimental to stable crystal growth. It lacks the ability to predict growth trends and make intelligent decisions based on multimodal information (such as plasma state and surface morphology images).

[0005] In recent years, although some studies have attempted to introduce advanced algorithms (such as fuzzy control and neural networks) for parameter prediction or optimization, these optimizations are often concentrated at a single time point or in a homogeneous space, and have not yet effectively solved the problem of coordinated optimization across space and time—the challenge of "adaptive optimization of the macroscopic growth region" and "dynamic and precise matching of microscopic process parameters." Meanwhile, how to deeply integrate the holographic virtual mapping provided by digital twin technology with multimodal sensing information to construct an intelligent growth control system capable of real-time response, multi-objective trade-offs, and self-learning remains a critical technological bottleneck that the industry urgently needs to overcome.

[0006] Therefore, existing technologies suffer from problems such as weak multi-objective collaborative optimization capabilities, disconnect between macro-control and macro-control, and insufficient dynamic adaptability and intelligence, which restrict the development of synthetic diamonds in terms of higher quality, larger size, lower cost, and more stable mass production. This invention aims to address these shortcomings by proposing an innovative solution. Summary of the Invention

[0007] The purpose of this invention is to address the problems of weak multi-objective collaborative optimization capabilities, disconnect between macro-control and macro-control, and insufficient dynamic adaptability and intelligence in existing technologies, which restrict the development of artificial diamonds in terms of higher quality, larger size, lower cost and more stable mass production. Therefore, this invention proposes an optimization method for artificial diamond synthesis based on an adaptive growth algorithm.

[0008] The technical solution of the present invention to solve the above-mentioned technical problems is as follows: An optimized method for synthesizing artificial diamonds based on an adaptive growth algorithm includes the following stages: S1. Construct a digital twin model and build a Pareto front objective function; S2. Real-time acquisition of chamber data via multimodal sensors and construction of state vectors. ,in, Temperature characteristics, As a pressure characteristic, Characterized by carbon source concentration. Characteristics of plasma density, Features of the seed crystal surface; S3. Based on the ant colony algorithm, plan the macroscopic growth path of artificial diamonds to obtain high-quality growth points; S4. Using high-quality growth points as input, fine-tune the parameters based on the particle swarm optimization algorithm; S5. Output the final parameter set to achieve dynamic and precise control of the growth environment.

[0009] Based on the above technical solution, the present invention can be further improved as follows.

[0010] Preferably, the Pareto front objective function is: ; in, For purity, For growth rate, For energy consumption, For the weight set, , This serves as the lower bound for each objective.

[0011] Preferably, step S3 includes the following steps: First, the seed growth surface is dynamically divided into a fine grid to construct a virtual path map. Each grid point represents a potential growth location. By integrating surface geometric features and physical environment parameters, heuristic information is formed to comprehensively evaluate the growth potential of each point. Several virtual ants are deployed starting from the seed center. Based on the comprehensive evaluation of pheromone concentration accumulated from historical successful experiences and current environmental heuristic information, and considering growth direction constraints and frontier limitations, the ants choose a movement path according to probability. After the ants complete the path exploration, a local pheromone evaporation and global enhancement mechanism is implemented to precipitate the experience of successful paths into pheromone distribution. The pheromone concentration is calculated and compared with a set threshold. If the pheromone concentration is less than the set threshold, the iteration continues. When the pheromone concentration is greater than the set threshold, the region with a significantly higher pheromone concentration than the surrounding area is selected as a high-quality growth point. These high-quality growth points are accurately passed to the particle swarm optimization algorithm as input to the algorithm.

[0012] Preferably, the formula for calculating the heuristic information is as follows: ; in, For nodes heuristic information value, For nodes The surface curvature at that point. For nodes Temperature gradient modulus, For nodes The carbon source concentration gradient modulus, To prevent division by zero, the value is taken as... .

[0013] Preferably, step S4 includes the following steps: For each high-quality growth point identified by the ant colony algorithm, particle swarm optimization is used to optimize micro-parameters. Each particle represents a small adjustment scheme for a set of local process parameters, including the offset of temperature, pressure, and carbon source concentration. A comprehensive fitness function is constructed. The particle swarm is updated iteratively, combining the individual historical best experience with the shared knowledge of the group, and gradually converges to the optimal fine-tuning scheme. A pulse adjustment mechanism is used to ensure that parameter changes are performed only when necessary, avoiding the impact of high-frequency oscillations on growth stability. Dynamic safety boundary monitoring is implemented during the optimization process. After convergence, the optimal solution is spatially and temporally smoothed to eliminate parameter abrupt changes in adjacent regions and smoothly transition with the results of the previous cycle.

[0014] Preferably, the pulse adjustment mechanism includes the following steps: When the adjustment conditions are met, an adjustment assessment is triggered. After the CNN-LSTM network security verification confirms that there is no defect risk, the step size coefficient is dynamically adjusted according to the current growth stage. Then, the parameter adjustment amount is calculated based on the adjusted step size coefficient. The adjustment execution mechanism is to accumulate adjustment requirements. Only when the accumulated parameter amount exceeds the threshold will a discrete pulse signal be generated to change the parameters.

[0015] Preferably, the adjustment conditions are: First, the Pareto front objective function is approximated using the finite difference method. gradient vector , the gradient vector gradient magnitude With modulus threshold Comparison, if the gradient magnitude Less than or equal to the modulus threshold If the gradient modulus is not analyzed, then real-time growth data will not be analyzed. Greater than the modulus threshold If real-time growth data is analyzed, and a critical change in growth status is detected, an adjustment assessment is triggered; otherwise, no adjustment assessment is triggered.

[0016] Preferably, the formulas for calculating the adjusted step size coefficient and parameter adjustment amount are as follows: Adjusted step size coefficient: ; in, To adjust the step size coefficient, This is the current step size coefficient. For the current thickness, For target thickness; Parameter adjustment amount: ; in, To ensure that the adjustment direction is aligned with the target optimization direction.

[0017] Preferably, the modulus threshold The acquisition process is as follows: By analyzing several historical growth data points, the relationship was determined through linear regression. Energy consumption The relationship between the modulus and the threshold value. The calculation process is as follows: ; in, , These are the first and second coefficients, respectively, both of which can be calculated using linear regression. This represents the maximum permissible energy consumption.

[0018] Compared with the prior art, the technical solution of this application has the following beneficial technical effects: 1. This invention adopts a two-stage collaborative optimization framework of "ant colony algorithm for macroscopic optimization + particle swarm optimization for microscopic fine-tuning". The ant colony algorithm searches globally for "high-quality growth points" on the seed surface, solving the problem of poor local growth caused by parameter homogenization in traditional methods. The particle swarm optimization algorithm then fine-tunes the parameters at these key points, realizing "macroscopic guidance and local optimization". This collaborative mechanism can effectively reconcile the contradictions of multiple objectives such as purity, growth rate, and energy consumption, and find a better solution at the Pareto front, thereby ensuring high quality while taking into account production efficiency and economic benefits.

[0019] 2. By constructing a digital twin model and integrating real-time data from multimodal sensors, this invention enables the system to build a state vector that reflects the true physical state of the chamber. Combined with a Pareto front-based objective function, the system can not only perform static optimization but also dynamically evaluate the growth state based on real-time acquired features such as temperature, pressure, plasma density, and images. Through a "pulse-type adjustment mechanism," precise and discrete parameter adjustments are triggered only when necessary (such as when gradient changes are significant or critical state changes occur), avoiding high-frequency and blind adjustments. This achieves a leap from "open-loop preset" to "closed-loop adaptive," significantly enhancing the robustness of the process to equipment fluctuations and external interference.

[0020] 3. This invention utilizes an ant colony algorithm to simulate biological intelligence, efficiently identifying the regions with the greatest growth potential on the crystal seed surface and guiding resources to these regions preferentially, thereby accelerating the growth of high-quality diamond layers. Real-time fine-tuning of micro-parameters ensures that each growth point operates within the optimal process window, effectively suppressing defect generation. The smoothing mechanism further guarantees the continuity of parameters in space and time, avoiding crystal structure stress or uneven growth layer caused by parameter mutations, ultimately improving the size, quality consistency, and yield of single-batch synthesized crystals.

[0021] 4. This invention explicitly incorporates "energy consumption" into a multi-objective optimization function, automatically balancing growth rate and energy consumption during the optimization process. Through precise environmental control and efficient synthesis only at high-quality growth points, it avoids ineffective energy dissipation and carbon source waste. The pulse adjustment mechanism reduces unnecessary actuator actions and also lowers energy consumption. Therefore, while achieving the goal of high-quality synthesis, this invention can achieve a lower unit production cost than traditional constant parameter or simple PID control, resulting in significant economic benefits.

[0022] 5. The entire optimization process of this invention is highly automated. The pheromone update mechanism enables the system to accumulate and inherit historical successful experiences, and the CNN-LSTM network is used to predict and prevent defect risks, giving the system self-learning and predictive capabilities. This reduces the high dependence on the operator's personal experience and real-time judgment, making the complex artificial diamond synthesis process easier to standardize, stabilize, and scale up. Attached Figure Description

[0023] Figure 1 This is a schematic diagram of the process of the present invention; Figure 2 This is a logic diagram for the adjustment conditions of the pulse-type adjustment mechanism of the present invention. Detailed Implementation

[0024] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0025] An optimized method for synthesizing artificial diamonds based on an adaptive growth algorithm includes the following stages: S1. Construct a digital twin model and build a Pareto front objective function; S2. Real-time acquisition of chamber data via multimodal sensors and construction of state vectors. ,in, Temperature characteristics, As a pressure characteristic, Characterized by carbon source concentration. Characteristics of plasma density, Features of the seed crystal surface; S3. Based on the ant colony algorithm, plan the macroscopic growth path of artificial diamonds to obtain high-quality growth points; S4. Using high-quality growth points as input, fine-tune the micro parameters based on the particle swarm optimization algorithm; S5. Output the final parameter set to achieve dynamic and precise control of the growth environment.

[0026] The construction of the digital twin model includes the following steps: The three-dimensional space of the reaction chamber is discretized into a non-uniform adaptive grid, wherein the grid resolution of the seed surface region is ≤1μm and the grid resolution of the chamber edge region is ≥100μm; the physical quantities of each grid element are calculated in real time by solving a set of coupled partial differential equations, which includes the carbon source concentration diffusion equation.

[0027] The Pareto front objective function is: ; in, For purity, For growth rate, For energy consumption, For the weight set, , This serves as the lower bound for each objective.

[0028] Step S3 includes the following steps: First, the seed growth surface is dynamically divided into a fine grid to construct a virtual path map. Each grid point represents a potential growth location. By integrating surface geometric features and physical environment parameters, heuristic information is formed to comprehensively evaluate the growth potential of each point. Several virtual ants are deployed starting from the seed center. Based on the comprehensive evaluation of pheromone concentration accumulated from historical successful experiences and current environmental heuristic information, and considering growth direction constraints and frontier limitations, the ants choose a movement path according to probability. After the ants complete the path exploration, a local pheromone evaporation and global enhancement mechanism is implemented to precipitate the experience of successful paths into pheromone distribution. The pheromone concentration is calculated and compared with a set threshold. If the pheromone concentration is less than the set threshold, the iteration continues. When the pheromone concentration is greater than the set threshold, the region with a significantly higher pheromone concentration than the surrounding area is selected as a high-quality growth point. These high-quality growth points are accurately passed to the particle swarm optimization algorithm as input to the algorithm.

[0029] The formula for calculating the heuristic information is as follows: ; in, For nodes heuristic information value, For nodes The surface curvature at that point. For nodes Temperature gradient modulus, For nodes The carbon source concentration gradient modulus, To prevent division by zero, the value is taken as... .

[0030] S4 includes the following steps: For each high-quality growth point identified by the ant colony algorithm, particle swarm optimization is used to optimize micro-parameters. Each particle represents a small adjustment scheme for a set of local process parameters, including the offset of temperature, pressure, and carbon source concentration. A comprehensive fitness function is constructed by integrating carbon atom adsorption efficiency assessment and defect risk prediction to achieve multi-dimensional evaluation of growth quality. The particle swarm is updated iteratively, combining individual historical best experience with shared knowledge of the group, and gradually converges to the optimal fine-tuning scheme. A pulsed adjustment mechanism is used to ensure that parameter changes are performed only when necessary, avoiding the impact of high-frequency oscillations on growth stability. Dynamic safety boundary monitoring is implemented during the optimization process. After convergence, the optimal solution is spatially and temporally smoothed to eliminate parameter abrupt changes in adjacent regions and smoothly transition with the results of the previous cycle. Finally, the optimization result is transformed into precise execution instructions, working in conjunction with macro-control to achieve ordered stacking control at the micro-scale. This process completes one full iteration per second, forming a "perception-optimization-execution-feedback" closed loop, effectively suppressing the formation of micro-defects while maintaining efficient growth, enabling the diamond crystal to achieve dual optimization of quality and rate at the nanoscale.

[0031] The comprehensive fitness function is: ; in, For fitness value, The adsorption rate of carbon atoms. This represents the defect probability. For safety constraint penalties, , , These are the corresponding weighting coefficients.

[0032] The pulse-type adjustment mechanism includes the following steps: When the adjustment conditions are met, an adjustment assessment is triggered. After the CNN-LSTM network security verification confirms that there is no defect risk, the step size coefficient is dynamically adjusted according to the current growth stage. Then, the parameter adjustment amount is calculated based on the adjusted step size coefficient to ensure more precise adjustments in the later stages. The adjustment execution mechanism is based on the accumulation of adjustment needs. Only when the accumulated amount exceeds the threshold will a discrete pulse signal be generated to change the parameters, effectively reducing unnecessary parameter fluctuations.

[0033] The adjustment conditions are as follows: First, the Pareto front objective function is approximated using the finite difference method. gradient vector , the gradient vector gradient magnitude With modulus threshold Comparison, if the gradient magnitude Less than or equal to the modulus threshold If the gradient modulus is not analyzed, then real-time growth data will not be analyzed. Greater than the modulus threshold If real-time growth data is analyzed, and a critical change in growth status is detected, an adjustment assessment is triggered; otherwise, no adjustment assessment is triggered.

[0034] Growth status assessment includes: Growth step height change rate ; Predicted defect probability change rate ; The secondary adsorption rate of the carbon source deviates from the target value by more than [a certain amount]. ; If any of the above indicators exceeds the threshold, a discrete pulse signal will be generated.

[0035] The formulas for calculating the adjusted step size coefficient and parameter adjustment amount are as follows: Adjusted step size coefficient: ; in, To adjust the step size coefficient, This is the current step size coefficient. For the current thickness, For the target thickness, where This includes step size factors for temperature, pressure, carbon source concentration, and plasma density. Parameter adjustment amount: ; in, To ensure that the adjustment direction is aligned with the target optimization direction.

[0036] The modulus threshold The acquisition process is as follows: By analyzing several historical growth data points, the relationship was determined through linear regression. Energy consumption The relationship between the modulus and the threshold value. The calculation process is as follows: ; in, , These are the first and second coefficients, respectively, both of which can be calculated using linear regression. This represents the maximum permissible energy consumption.

[0037] This approach can also be applied to the artificial synthesis of colored diamonds. Firstly, the core optimization objective needs to shift from pursuing ultimate purity to a multi-faceted game balancing "color quality," "defect concentration," "crystal integrity," "growth rate," and "energy consumption." The Pareto front objective function must incorporate target concentrations of color parameters (such as chroma, saturation, and uniformity) and specific defects (such as NV centers and isolated nitric oxide), achieving a comprehensive optimization of both "beauty" and "quality." To achieve this goal, online spectral diagnostics and hyperspectral imaging need to be integrated into the basic physical data to capture defect types, concentrations, and color distribution in real time, forming an enhanced state vector that integrates "physical state" and "color state." Correspondingly, the digital twin model needs to be upgraded from macroscopic growth simulation to a cross-scale model coupling microscopic defect dynamics and color prediction, simulating how process parameters are ultimately mapped to color through defect engineering.

[0038] The intelligent decision-making logic also needs to evolve in tandem. The heuristic information from the ant colony algorithm needs to incorporate color uniformity guidance, ensuring that macroscopic growth point planning balances efficiency and color distribution. The optimization parameters of the particle swarm optimization algorithm need to be expanded to include key variables such as dopant flux, with its fitness function directly guided by color quality, driving parameters to converge towards the target hue. The adjustment mechanism of the entire "perception-decision-control" closed loop will be more refined and cautious to prevent over-adjustment of sensitive colors. The process safety boundary has also been redefined from preventing abnormal growth to ensuring that the color remains within the "target color gamut."

[0039] In summary, applying the above solution to colored diamonds transforms the original method from an "environmental stabilizer" to a "color designer" by reconstructing the target, enhancing perception, deepening the model, and adapting the algorithm, thus achieving predictable, controllable, and optimizable intelligent synthesis of diamond colors.

[0040] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0041] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An optimized method for synthesizing artificial diamonds based on an adaptive growth algorithm, characterized in that, Includes the following stages: S1. Construct a digital twin model and build a Pareto front objective function; S2. Real-time acquisition of chamber data via multimodal sensors and construction of state vectors. ,in, Temperature characteristics, As a pressure characteristic, Characterized by carbon source concentration. Characteristics of plasma density, Features of the seed crystal surface; S3. Based on the ant colony algorithm, plan the macroscopic growth path of artificial diamonds to obtain high-quality growth points; S4. Using high-quality growth points as input, fine-tune the parameters based on the particle swarm optimization algorithm; S5. Output the final parameter set to achieve dynamic and precise control of the growth environment.

2. The method for optimizing artificial diamond synthesis based on an adaptive growth algorithm according to claim 1, characterized in that, The Pareto front objective function is: ； in, For purity, For growth rate, For energy consumption, For the weight set, , This serves as the lower bound for each objective.

3. The method for optimizing artificial diamond synthesis based on an adaptive growth algorithm according to claim 1, characterized in that, Step S3 includes the following steps: First, the seed growth surface is dynamically divided into a fine grid to construct a virtual path map. Each grid point represents a potential growth location. By integrating surface geometric features and physical environment parameters, heuristic information is formed to comprehensively evaluate the growth potential of each point. Several virtual ants are deployed starting from the seed center. Based on the comprehensive evaluation of pheromone concentration accumulated from historical successful experiences and current environmental heuristic information, and considering growth direction constraints and frontier limitations, the ants choose a movement path according to probability. After the ants complete the path exploration, a local pheromone evaporation and global enhancement mechanism is implemented to precipitate the experience of successful paths into pheromone distribution. The pheromone concentration is calculated and compared with a set threshold. If the pheromone concentration is less than the set threshold, the iteration continues. When the pheromone concentration is greater than the set threshold, the region with a significantly higher pheromone concentration than the surrounding area is selected as a high-quality growth point. These high-quality growth points are accurately passed to the particle swarm optimization algorithm as input to the algorithm.

4. The method for optimizing artificial diamond synthesis based on an adaptive growth algorithm according to claim 3, characterized in that, The formula for calculating the heuristic information is as follows: ； in, For nodes heuristic information value, For nodes The surface curvature at that point. For nodes Temperature gradient modulus, For nodes The carbon source concentration gradient modulus, To prevent division by zero, the value is taken as... .

5. The method for optimizing artificial diamond synthesis based on an adaptive growth algorithm according to claim 1, characterized in that, S4 includes the following steps: For each high-quality growth point identified by the ant colony algorithm, particle swarm optimization is used to optimize micro-parameters. Each particle represents a small adjustment scheme for a set of local process parameters, including the offset of temperature, pressure, and carbon source concentration. A comprehensive fitness function is constructed. The particle swarm is updated iteratively, combining the individual historical best experience with the shared knowledge of the group, and gradually converges to the optimal fine-tuning scheme. A pulse adjustment mechanism is used to ensure that parameter changes are performed only when necessary, avoiding the impact of high-frequency oscillations on growth stability. Dynamic safety boundary monitoring is implemented during the optimization process. After convergence, the optimal solution is spatially and temporally smoothed to eliminate parameter abrupt changes in adjacent regions and smoothly transition with the results of the previous cycle.

6. The method for optimizing artificial diamond synthesis based on an adaptive growth algorithm according to claim 5, characterized in that, The pulse-type adjustment mechanism includes the following steps: When the adjustment conditions are met, an adjustment assessment is triggered. After the CNN-LSTM network security verification confirms that there is no defect risk, the step size coefficient is dynamically adjusted according to the current growth stage. Then, the parameter adjustment amount is calculated based on the adjusted step size coefficient. The adjustment execution mechanism is the accumulation of adjustment requirements. Only when the accumulated parameter amount exceeds the threshold will a discrete pulse signal be generated to change the parameters.

7. The method for optimizing artificial diamond synthesis based on an adaptive growth algorithm according to claim 6, characterized in that, The adjustment conditions are as follows: First, the Pareto front objective function is approximated using the finite difference method. gradient vector , the gradient vector gradient magnitude With modulus threshold Comparison, if the gradient magnitude Less than or equal to the modulus threshold If the gradient modulus is not analyzed, then real-time growth data will not be analyzed. Greater than the modulus threshold If real-time growth data is analyzed, and a critical change in growth status is detected, an adjustment assessment is triggered; otherwise, no adjustment assessment is triggered.

8. The method for optimizing artificial diamond synthesis based on an adaptive growth algorithm according to claim 6, characterized in that, The formulas for calculating the adjusted step size coefficient and parameter adjustment amount are as follows: Adjusted step size coefficient: ； in, To adjust the step size coefficient, This is the current step size coefficient. For the current thickness, For target thickness; Parameter adjustment amount: ； in, To ensure that the adjustment direction is aligned with the target optimization direction.

9. The method for optimizing artificial diamond synthesis based on an adaptive growth algorithm according to claim 7, characterized in that, The modulus threshold The acquisition process is as follows: By analyzing several historical growth data points, the relationship was determined through linear regression. Energy consumption The relationship between the modulus and the threshold value. The calculation process is as follows: ； in, , These are the first and second coefficients, respectively, both of which can be calculated using linear regression. This represents the maximum permissible energy consumption.

Images